Part One: Processes and Boundaries
Chapter 1 The Place of Casting in the Process Sequence: Forming Precedes Machining
The forming routes for a mechanical part fall broadly into four: removing excess material from a solid block; reshaping by plastic deformation; pressing metal powder into shape and sintering it; and melting metal into a liquid, filling it into a mould cavity and letting it solidify. Casting is the last of these. Its product is called a casting, and most castings leave the foundry as blanks that still need machining before they become parts. The precision of a casting therefore carries an extra layer of meaning: it determines how much allowance later machining must leave and how much material must be cut away.
This position brings two characteristics. The shape of a casting comes from the mould cavity: the casting stops at whatever precision the cavity achieves, and every subsequent machining line must carry allowance for the error in that cavity. The performance of a casting comes from solidification: the shrinkage, feeding and cooling rate as the metal cools from liquid to solid determine whether the interior is dense or porous. The first is a geometric question, the second a microstructural one; neither lies within the scope of the machining operation, and neither can be recovered by later processing. The closer a casting is to the finished part, the fewer machining hours it needs and the higher the investment in tooling and process. This trade-off runs through all five process families.
According to the China Foundry Association, China's casting output in 2024 was 50.75 million tonnes (5075 万吨), about 45% or more of global casting output, and it has ranked first in the world for 25 consecutive years since 2000 (China Foundry Association). That volume places casting upstream of almost every machinery industry, and the technical span within the single word "casting" is wider than the span between casting and forging or machining.
The five process families can be separated by three criteria: whether the mould is reused, what force fills it with liquid metal, and what material forms the cavity.
Sand casting uses a single-use sand mould, filled by gravity, and the mould breaks up once the box is opened and the part removed. Investment casting also uses a single-use mould, with the cavity formed by a ceramic shell whose precision comes from the wax pattern and the shell-building sequence; the shell is broken open to remove the part. Permanent mould casting uses a reusable metal mould, still filled by gravity, with mould life counted in pours. Die casting also uses a metal mould, filled at high speed under high pressure, with filling times usually measured in milliseconds. Centrifugal casting pours liquid metal into a rotating mould so that centrifugal force presses it against the mould wall to form the part, and its products are mainly bodies of revolution.
Beneath the five lie crossovers and derivatives. In lost-foam casting the mould is dry sand and the pattern is expanded plastic; the pattern vaporises during pouring, making this a combination of single-use mould and gravity filling. In shell moulding the mould is a thin shell bonded from resin-coated sand, backed with sand shot. Low-pressure die casting uses a metal mould and fills by gas pressure, lower than in die casting but higher than gravity pouring; counter-pressure casting adds a pressure vessel around the mould so that filling and solidification take place under back pressure. The position of these derivatives is fixed by the three criteria together; reading any one of them in isolation leads to misclassification.
The criteria are useful because each corresponds to a cost structure. A single-use mould has low upfront investment, and its unit cost falls slowly with batch size; a metal mould has high upfront investment, and its unit cost falls quickly with batch size. Gravity filling needs simple equipment and fills smoothly, but thin sections are prone to misruns; high-pressure, high-speed filling can produce thin walls and fine detail, at the cost of entrained gas and mould wear. Process choice is therefore rarely decided by a single indicator. One batch of parts may fall within the capability of three processes at once, and the final choice also depends on tooling amortisation, lead time and the supplier's spare equipment capacity.
For a buyer, a process name supplies a classification; what an enquiry must pin down is four things: which range the piece weight falls into, how thick the thinnest section is, which dimensions carry a tolerance requirement, and how many pieces are needed in a year. Once these four variables are fixed, the range of candidate processes largely narrows.
A 2023 document from the Ministry of Industry and Information Technology and two other ministries listed integrated die casting, patternless casting and 3D sand printing among the directions to be pushed towards industrial application (guiding opinions of the Ministry of Industry and Information Technology and two other ministries). These three sit at the two ends of die casting and sand casting, which shows that specific processes are still moving. The criteria provide a classification framework; the placement of specific equipment must be reconfirmed against current capability.
Chapter 2 Sand Casting: Hand Moulding, Machine Moulding and No-Bake Sand
A sand mould is mixed from base sand, binder and additives. The base sand is mainly silica sand, and the binder determines how the mould gains strength: clay plus water gives green sand, resin cross-links through a hardener, and sodium silicate hardens through carbon dioxide or an organic ester. Additives include coal dust, starch and wood flour, used to improve collapsibility and surface quality. Moulding sand must satisfy several indicators that pull against one another at the same time: green compressive strength must be sufficient, or the mould collapses during pattern withdrawal; permeability must be sufficient, or gases in the cavity cannot escape during pouring; flowability must be sufficient, or compaction varies across the mould. These three constrain one another in the formulation, and most of the work in the sand-preparation section is finding a balance among them. Used sand is reclaimed after crushing, screening and cooling; the reclamation ratio affects the stability of the moulding sand and also determines the volume of waste sand discharged.
The mould is divided into a cope and a drag, and the parting line is the surface where the two meet. Internal cavities of the casting are formed by cores, which are broken up and cleared out after pouring. The parting line and the cores set the upper limit on casting shape: the more parting lines, the greater the risk of mismatch and flash; the more cores, the greater the accumulated assembly error and the more complex the venting path. The gating system must also be laid out within the mould; the positions of the sprue, runner and ingate affect the filling sequence and the temperature distribution of the liquid metal.
Hand moulding relies on manual work for sand filling, ramming, pattern withdrawal and patching, and suits one-off parts, small batches and large castings. Its patterns are mostly made of wood, which is cheap and easy to modify; the drawback is that mould compaction depends on the operator's experience, so dimensional consistency between batches is poorer. Hand-moulded sand moulds are usually not dried, which is called green sand; the moisture vaporises during pouring and is one source of gas porosity. Pit moulding for large castings builds the mould directly in a pit, dispensing with flasks and raising the requirement for lifting capacity accordingly.
Machine moulding breaks sand filling, compaction and pattern withdrawal into machine motions; common methods are jolt-squeeze, micro-jolt squeeze, air impulse and static-pressure moulding. The pace of a moulding line is counted in moulds per hour, and compaction is guaranteed by equipment parameters, giving clearly better consistency than hand moulding. Horizontal-parting lines require flasks, while vertical-parting lines can mould without flasks; the latter holds a place in high-volume production of small and medium parts. The upfront investment in machine moulding is the pattern plate and flasks; patterns must be fixed to the plate, so changing the pattern costs more than in hand moulding. Once the capacity of a moulding line exists, it must be supported by continuous orders, and downtime for pattern changes goes straight into cost.
No-bake sand is the third route. A hardener is added when the sand is mixed, and the mould hardens by itself after being filled into the flask, requiring no drying and no compacting motion from a moulding machine. Furan resin no-bake sand is acid-catalysed, alkaline phenolic resin no-bake sand is ester-hardened, and sodium silicate sand can use carbon dioxide or an organic ester. No-bake sand has good collapsibility and higher dimensional accuracy than clay-bonded sand, and suits medium-to-large castings and one-off or small-batch production. One sand-casting plant's public capability page gives a range of 0.3 kilograms to 20 tonnes per sand casting, with the resin-sand line covering 60 to 15,000 kilograms per piece (sand casting capability page). This is the equipment basis of a single enterprise and cannot be taken as industry statistics; it is used here to show the weight coverage of sand processes.
The precision of sand moulds is measured by the parameters in an industry encyclopaedia entry (neither a national standard nor a handbook): surface roughness Ra of 50 to 12.5 micrometres, with a dimensional tolerance of about ±1.0 millimetres on a 100 millimetre basis (industry encyclopaedia process entry); the minimum wall thickness varies with material, above 3 millimetres for aluminium alloy, 3 to 4 millimetres for cast iron and above 5 millimetres for cast steel, and the wall thickness figure belongs to a common industry process range for which this article has not obtained a citable specific source. These numbers reflect the normal level of sand casting; precision machine moulding and no-bake sand can do better, with the extent of improvement limited by sand grain and mould stiffness.
The output base of sand casting lies in the material mix. According to a report by China Diecasting Magazine on Sina, sourced from the China Foundry Association's "2025 China Foundry Industry Data", grey iron castings accounted for 40.2% of total casting output in China in 2025 and ductile iron castings for 27.2% (Sina reprint); together they approach 70%, and the main production method for iron castings is sand casting. Sand casting applies to cast iron, cast steel and non-ferrous alloys alike, with low pattern cost and a wide size range, which makes it hard to replace for small batches and large parts.
The inherent defects of sand moulds are equally stable: sand inclusion comes from insufficient moulding-sand strength or sand wash, scab from the mould wall spalling at high temperature, gas porosity from moisture in the moulding sand and poor venting of cores, and shrinkage porosity from insufficient feeding. All these defects arise in the mould cavity and the gating system, have nothing to do with the machining operation, and can only be solved at the casting stage. Writing defect names and acceptance methods into the contract is more enforceable than a general requirement of "smooth surface, free of defects".
Chapter 3 Precision Casting: Investment, Lost-Foam and Shell Moulding
"Precision casting" is a collective name for a group of processes whose common points are a single-use mould and a cavity more precise than that of sand casting. The three main approaches differ completely in how the cavity is formed, and in the length of the process chain.
Investment casting begins with a wax pattern. Wax is injected into a metal die on a wax-injection press to give a wax part identical in shape to the casting; the wax part is trimmed and welded onto a sprue bar to form a pattern cluster; the cluster is repeatedly dipped in ceramic slurry and stippled with sand, drying layer by layer to build a shell of a certain thickness; after dewaxing and firing, the shell becomes a hollow, heat-resistant cavity, which is poured with liquid metal, and the shell is broken open after cooling to remove the casting. The temperature and pressure of the dewaxing autoclave affect whether the shell cracks, and the firing temperature determines the residual strength of the shell. Along the whole chain, dimensional accuracy is transmitted through three links: the accuracy of the wax-injection die, the shrinkage of the wax, and the thermal expansion of the shell during firing and pouring.
The precision of investment castings depends on the binder system. Sodium silicate shells are cheap, with dimensional tolerance grades roughly CT7 to CT9; silica-sol shells reach CT5 to CT7 with low-temperature wax and CT3 to CT5 with medium-temperature wax. Weight coverage is graded in the same way: about 0.1 to 100 kilograms for sodium silicate and combined processes, and about 0.1 to 60 kilograms for the silica-sol process (investment casting process description). These ranges come from an enterprise technical page and can serve as a reference for common industry ranges, but should not be taken as industry-wide statistics. The choice of binder system is decided jointly by material and precision: silica sol costs more, while sodium silicate gives a strong shell but lower precision.
On the basis of an industry encyclopaedia entry (neither a national standard nor a handbook), investment castings have a surface roughness Ra of 12.5 to 1.6 micrometres and a dimensional tolerance of about ±0.3 millimetres on a 100 millimetre basis (industry encyclopaedia process entry); the minimum wall thickness is usually 0.7 millimetres and the minimum cast hole diameter 1.5 to 2 millimetres, a set of figures that falls within common industry process ranges and for which this article has not obtained a citable specific source. Investment casting can produce thin walls and fine holes, at the cost of a long process chain, a long cycle and a high unit cost, and its upper limit on piece weight is clearly below that of sand casting. It suits small precision parts in hard-to-machine materials such as stainless steel, heat-resistant steel and high-alloy steel. The shell is single-use, so every casting consumes one shell, and the labour hours of shell making carry a higher share of cost than the moulding step in sand casting.
In lost-foam casting the cavity comes from an expanded-plastic pattern. Foam beads are pre-expanded, moulded and matured, then fused into a pattern; the surface is coated with a refractory coating and dried, the pattern is placed in a flask and compacted by vibrating dry sand, and during pouring the liquid metal vaporises the pattern, the gas escapes through the coating and the dry sand, and the metal takes the place the pattern occupied. Lost-foam casting has no parting line and no cores, produces no flash or mismatch, and can form complex internal cavities in a single piece.
The difficulty with lost-foam casting lies in the products of foam vaporisation. Vaporising the pattern absorbs heat, so the pouring temperature must be raised accordingly; if the vaporisation products cannot escape freely they leave wrinkles on the casting surface and form gas porosity inside, and for low-carbon steel they also carry a risk of carbon pick-up. Vacuum-assisted pouring is a common countermeasure, drawing a vacuum in the flask to extract the vaporisation products. Lost-foam casting suits small and medium parts with complex internal cavities such as boxes, housings and pipe fittings, and at medium batch sizes the pressure of tooling amortisation is lower than in die casting.
In shell moulding the cavity is a thin shell bonded from resin-coated sand. Resin-coated sand is made by coating silica sand with phenolic resin; it is shot onto a heated metal pattern plate, where the resin cures to give the two half-shells either side of the parting line, which are clamped together, placed in a flask, backed with sand shot or steel shot and then poured. Shell moulds have better surface roughness than clay-bonded sand moulds and dimensional accuracy between sand casting and investment casting, and suit medium-batch production of small and medium parts. Shell moulding cost is set by the resin-coated sand and the pattern plate: the plate has a long life and the shell is used once. Shooting and curing take place on the heated plate, whose temperature together with the shooting pressure determines shell thickness; a shell that is too thin will burst during pouring.
What the three processes share is that the mould is destroyed when the part is removed, so every casting needs a new mould. This common point determines their cost curves: the larger the batch, the more repetitive labour in mould making and the more valuable automated mould making becomes; the smaller the batch, the heavier the amortisation of pattern and tooling. Investment casting has the longest process chain and high dimensional accuracy, lost-foam casting has the greatest freedom of shape and the hardest defect control, and shell moulding sits between the two; the choice among them depends on the combination of precision requirement and batch size.
Chapter 4 Die Casting: High Pressure, Low Pressure and Counter-Pressure
In die casting the liquid metal fills the cavity under pressure. The complete high-pressure die casting cycle is: die closing, shot, intensification, cooling, die opening, ejection, and removal of the gate and overflow. The shot has two stages: the slow shot pushes the metal up to the gate, the fast shot drives it into the cavity at high speed, and intensification then builds solidification pressure so that shrinkage cavities are compensated under pressure.
The process window of high-pressure die casting is wide. Industry technical material gives filling velocities of 0.5 to 120 metres per second, filling times of 0.01 to 0.2 seconds, and specific injection pressures from a few megapascals to 500 megapascals (die casting technical material page). Filling times are measured in milliseconds, the flow of metal in the cavity is close to turbulent, and gas in the cavity has no time to escape through the vent, so it is entrained in the metal and forms gas porosity. Gas porosity is an inherent feature of die castings and the reason they are generally not quenched and are difficult to weld: during heat treatment the gas pores expand and raise blisters on the casting surface. Vacuum die casting draws the cavity to a negative pressure before filling, which reduces the entrained gas and raises the sealing requirements for equipment and die accordingly.
High-pressure die casting ranks near the front of the five process families for precision. Surface roughness Ra is 6.3 to 1.6 micrometres and dimensional tolerance about ±0.3 millimetres on a 100 millimetre basis, with some textbooks giving a best surface of Ra 0.8 micrometres (industry encyclopaedia process entry). Minimum wall thickness depends on the alloy: 0.3 millimetres can be achieved in zinc alloy and 0.5 millimetres in aluminium alloy; the minimum cast hole diameter is 0.7 millimetres and the minimum pitch 0.75 millimetres. These figures fall within common industry process ranges, and this article has not obtained a citable specific source for them. Thin walls and small cast holes reduce subsequent machining, which is why die castings hold the advantage in housing-type parts.
Die casting has a relatively high economic batch threshold. One foundry's process page gives a rule of thumb of annual demand above 5,000 pieces (casting process comparison page). The threshold comes from tooling amortisation: a die casting die works under high pressure and thermal cycling, costs a great deal, and when the batch is too small the per-piece amortisation is too heavy. The piece weight of die castings is limited by equipment and die, and industry material lists two ranges for aluminium alloy die casting capability, 0.1 to 5 kilograms and 0.5 to 50 kilograms, with zinc alloy die castings at 0.02 to 2 kilograms (die casting capability introduction page). That page is commercial promotional material, so the ranges can only be treated as examples of common industry practice.
Low-pressure die casting drops the filling pressure to a different order of magnitude. The crucible is sealed inside the furnace body, compressed air presses the metal up a riser tube and into the cavity from below, the metal fills smoothly from bottom to top, and the pressure is held until solidification ends. Smooth filling reduces entrained gas, feeding is supplied continuously by the metal in the riser tube, and the density of the casting is higher than with gravity pouring. An industry process page gives wall thickness of 3 to 15 millimetres, tolerance grades CT5 to CT7 and annual demand above 1,000 pieces; one plant's low-pressure line covers aluminium alloy pieces of 0.5 to 25 kilograms (casting process comparison page). On an industry encyclopaedia entry basis, low-pressure die castings have a surface roughness Ra of 12.5 to 3.2 micrometres and a tolerance of ±0.4 millimetres on a 100 millimetre basis (industry encyclopaedia process entry); the minimum wall thickness is usually 2 millimetres and 0.7 millimetres at the thinnest, which falls within common industry process ranges and for which this article has not obtained a citable specific source.
Counter-pressure casting adds a pressure vessel to low-pressure die casting. The mould is placed inside a sealed vessel, the vessel and the crucible are pressurised together, the metal fills under the pressure difference, and after filling the pressure in the vessel continues to rise so that the casting solidifies in an environment above atmospheric pressure. The pressure improves feeding capacity and also suppresses gas evolution, suiting large, thin-walled complex parts with high density requirements. Counter-pressure casting is more complex than low-pressure die casting, with a narrower process window and stricter requirements on sealing and pressure control. The pressure curve in the vessel must be synchronised with the pouring temperature, and the cost of trial and error is higher than in low-pressure die casting. High-pressure die casting trades speed for thin walls and precision; low-pressure and counter-pressure casting trade smooth filling for density; the trade-off depends on what the part demands of wall thickness and density.
Gravity permanent mould casting sits between sand casting and die casting. The metal fills a metal mould by gravity, the mould is reusable, cooling is fast and the grain is finer. The industry encyclopaedia entry gives a surface roughness of Ra 12.5 to 6.3 micrometres and a tolerance of ±0.4 millimetres on a 100 millimetre basis (industry encyclopaedia process entry). Typical parts are intake manifolds, pump bodies and thick-walled parts, which require internal density but demand less in thin walls and fine detail than die castings.
The 2023 three-ministry document lists integrated die casting among advanced process directions (guiding opinions of the Ministry of Industry and Information Technology and two other ministries). Integrated die casting forms in one shot a large structural part previously assembled from several pieces, which places higher demands on the clamping force of the die casting machine, on die size and on filling control, and also pushes up the upper limit on the piece weight of die castings.
Chapter 5 Parameter Comparison: Surface Roughness, Dimensional Tolerance, Minimum Wall Thickness and Weight Range
| Process | Surface roughness Ra (micrometres) | Dimensional tolerance (millimetres, 100 millimetre basis) | Minimum wall thickness (millimetres) | Typical piece weight |
|---|---|---|---|---|
| Sand casting | 50 to 12.5 | ±1.0 | Above 3 for aluminium, 3 to 4 for cast iron, above 5 for cast steel | 0.3 kilograms to 20 tonnes |
| Investment casting | 12.5 to 1.6 | ±0.3 | 0.7 | 0.1 to 100 kilograms (silica sol 0.1 to 60) |
| Gravity permanent mould casting | 12.5 to 6.3 | ±0.4 | — | — |
| High-pressure die casting | 6.3 to 1.6 | ±0.3 | 0.3 for zinc, 0.5 for aluminium | Aluminium 0.1 to 5 kilograms |
| Low-pressure die casting | 12.5 to 3.2 | ±0.4 | Usually 2, 0.7 at the thinnest | 0.5 to 25 kilograms |
| Centrifugal casting | Determined by mould material | Determined by mould material | — | DN40 to DN3000 |
The table is compiled from industry encyclopaedia entries (neither national standards nor handbooks) and enterprise process pages. The roughness and tolerance columns are taken from the industry encyclopaedia process entry; the minimum wall thickness column is neither a national standard nor a single source and serves only for cross-comparison; the weight ranges for investment casting are taken from the investment casting process description; the weight ranges for sand casting from the sand casting capability page; the piece weights for die casting and low-pressure casting from the casting process comparison page and the die casting capability introduction page; and the size range for centrifugal cast iron pipe from the GB/T 13295-2026 extract page.
The three parameters of roughness, tolerance and wall thickness each describe a different thing. Surface roughness Ra is the arithmetic mean deviation of the profile; the smaller the value, the smoother the surface. It affects the seating of sealing faces, the adhesion of coatings and the running-in of friction pairs, and is not equivalent to dimensional accuracy. Dimensional tolerance describes the permitted range by which the actual size may deviate from nominal; the values in the table use a 100 millimetre basis, and the tolerance changes as the reference length changes. Minimum wall thickness describes the limit at which the process can fill reliably, and is decided jointly by the fluidity of the metal, the venting capacity of the cavity and the cooling rate.
The values in the table come from industry encyclopaedia entries and enterprise process pages; they can serve as a reference for cross-comparison but cannot be written as provisions of a national standard. The national standard system for dimensional tolerance speaks another language: GB/T 6414-2017 was withdrawn on 1 December 2025 (National Standards Full-Text Public System) and replaced by GB/T 42124.3-2025, which was issued on 30 May 2025 and came into force on 1 December 2025 (National Standards Full-Text Public System). The new standard is a modified adoption of ISO 8062-3:2023 (standard information page) and expresses casting dimensional tolerances in CT grades. A CT grade on a drawing constrains a supplier better than the words "precision casting".
The parameters are coupled. The thinner the wall, the faster the cooling, the more concentrated the shrinkage stress and the harder it is to hold the tolerance; the same tolerance band gives a smaller relative error on a large part and may consume the entire machining allowance on a small one. Weight and wall thickness together determine the size of the gating system, which in turn affects process yield. Roughness is related to mould temperature: when die temperature control in permanent mould casting and die casting is unstable, the same die produces surfaces that differ from batch to batch. The table lists the parameters line by line, but the real capability boundary is the joint constraint of several parameters.
Four points need to be stated when reading the table. The tolerance column uses a 100 millimetre basis; when a part's size departs from that length, the tolerance band is adjusted under the standard's stepped rules and cannot be copied directly. The weight column gives the net weight of a single casting, excluding the gating system and risers. Batch size is not listed, because the batch threshold depends on tooling amortisation, and tooling cost varies with the number of cavities and the degree of automation, so the batch threshold for the same process can differ several-fold. The weight and minimum wall thickness cells for gravity permanent mould casting are left blank because its applicable range varies with mould and equipment, public material offers no consistent basis, and any number inserted would be misleading.
The ranking of minimum wall thickness follows the filling method. Sand moulds fill by gravity, the metal cools quickly and meets high flow resistance in thin sections, and the risk of misrun rises as the wall gets thinner, so cast steel has a greater minimum wall thickness than cast iron and cast iron greater than aluminium alloy. Die casting fills at high speed and can form walls of 0.3 to 0.5 millimetres, at the cost of harder control of gas porosity and cold shuts in thin sections. Low-pressure die casting fills smoothly, and its minimum wall thickness lies between sand casting and die casting.
The span of the weight ranges likewise comes from equipment and mould. The 20 tonne upper limit of sand casting corresponds to pit moulding and heavy lifting equipment, the 5 kilogram upper limit of die casting to clamping force and shot capacity, and the upper limit of investment casting to shell strength and furnace size. Centrifugal casting produces mainly bodies of revolution, with size expressed as nominal diameter. The specification range of GB/T 13295 covers DN40 to DN3000 (standard extract page, which is an English-language summary of GB/T 13295-2026). GB/T 13295-2026 was issued on 25 May 2026 and comes into force on 1 March 2027, replacing the 2019 edition; the DN range in this article is taken from the 2026 edition summary page.
The value of the parameter table is that it translates "a bit more precise" into quantities that can be inspected. A purchasing drawing that states the process, the standard number and the CT grade, and then lists tolerances separately for critical dimensions, comes closer to an enforceable agreement than a general label of "precision part".
Chapter 6 How Batch Size and Tooling Life Affect Process Selection
Among casting costs there is one item different from other processes: tooling. Tooling is a one-off investment amortised over pieces, and the amortised amount equals tooling cost divided by batch size. Double the batch and the per-piece amortisation halves; with too small a batch, tooling cost outweighs material and labour and becomes the main part of the quotation. The first level of process selection therefore falls on batch size.
Tooling takes a different form in each process. Sand casting patterns are reusable — wood, aluminium and resin patterns rise in cost in that order — while the sand mould itself is consumed. Investment casting's wax-injection die is reusable, while the wax pattern and the shell are consumed, and tooling cost is amortised over wax parts. Die casting's steel die is reusable, with life counted in shots, and tooling cost is amortised over every die casting. Lost-foam casting's foaming die is reusable, while the foam pattern is consumed. All of these are described as "tooling", yet the one-off investment differs in magnitude and is amortised against different objects.
Tooling life affects three things. The first is amortisation: the longer the life, the lower the tooling cost each piece carries. The second is downtime: changing and repairing tooling occupies machine time, and small-batch orders spend a high proportion of time on die changes. The third is consistency: tooling wears in service, cavity dimensions drift slowly, and a systematic deviation appears between later parts and the first part. Die casting dies endure high pressure and thermal cycling over long periods, and thermal fatigue cracks and die sticking appear on the cavity surface; wax-injection dies for investment casting wear more slowly, and the dimensional stability of the shell is affected more by slurry and drying conditions. The difference in tooling life ultimately shows up in the dimensional distribution of the middle and later part of a batch, not only in unit cost.
Tooling design life should match order batch size. Tooling whose life far exceeds the batch leaves the investment idle for long periods, with insufficient utilisation; tooling whose life is shorter than the batch forces a die change mid-run, and the lead time for a second die enters the delivery date. Orders with steady volumes allow tooling and backup tooling to be designed on annual output, while parts with intermittent orders tend towards processes with low tooling cost that can be changed quickly. Tooling maintenance and backup tooling are costed too: cores and ejector pins in die casting dies are wearing parts, and wax-injection dies for investment casting need periodic reworking of the cavity.
The economic batch threshold can be quantified. One foundry's process page gives rules of thumb of annual demand above 5,000 pieces for die casting and above 1,000 pieces for low-pressure die casting (casting process comparison page). These two numbers come from a single enterprise's quotation logic and are used to illustrate the different batch requirements of the two processes: die casting has a large tooling investment and short unit labour time, needing a larger batch to spread the tooling; low-pressure die casting has a smaller tooling investment, so its batch threshold moves down accordingly.
A worked example shows the order of judgement. Suppose an aluminium alloy housing weighs 2 kilograms, has a maximum outline of 300 millimetres and a wall thickness of 4 millimetres, with annual demand of 800 pieces. By piece weight and wall thickness, die casting, low-pressure casting and sand casting are all within capability; by batch size, die casting's tooling amortisation is too heavy and sand and low-pressure casting are more suitable. If annual demand for the same part becomes 80,000 pieces, tooling amortisation falls to a very low share of unit cost, die casting's short labour time and small machining allowance begin to dominate, and the process choice turns accordingly. The geometry of the part has not changed; only the batch has.
Beyond batch size, three other constraints pull the choice back. The first is piece weight and outline: sand casting can cover up to the 20 tonne class, while die casting's usual piece weight stops at tens of kilograms. The second is material: the melting points of high-chromium cast iron and heat-resistant steel lie beyond what a die casting die can withstand, so only sand or investment casting will do. The third is precision and surface: die casting's Ra and tolerance beat sand casting, and demanding sealing faces are hard to cast directly in sand. The three constraints and batch size together form an intersection in which usually only one or two processes remain.
Lead time also takes part in the choice. Making tooling takes time, and die casting dies and wax-injection dies have longer build cycles than wood patterns; shell making requires layer-by-layer drying, and investment casting has a longer production cycle than sand casting. Small-batch trial orders use sand or lost-foam casting to obtain castings without cutting a steel die, then switch to die casting once the design is fixed, a path common at the sample stage. Batch size decides whether tooling can be amortised, tooling structure sets the floor of unit cost, and process selection is the intersection of the four constraints of batch, weight, material and precision.
Chapter 7 Typical Downstream Sectors by Process
The downstream distribution of castings can be read from two angles: by demand industry and by material. The conclusions from the two angles point to the same set of processes.
By demand industry, automotive is the largest user. In 2025 the automotive industry's demand for castings was 15.8 million tonnes, 30.2% of total casting demand, an increase of 5% (2025 foundry industry data). In 2024 the share was 29.6% (2024 foundry industry data). The main sectors after automotive are cast iron pipe and fittings, mining and metallurgical heavy machinery, construction machinery, internal combustion engines and agricultural machinery, machine tools, and power generation equipment and electric power (2025 foundry industry data). This list covers heavy equipment, energy, agriculture and construction, whose common points are modest batch sizes per order and a wide spread of piece weight and material.
By material, China's total casting output in 2025 was 52.3 million tonnes, up 3.1% year on year, resuming growth after negative growth and flat output between 2022 and 2024 and marking the second-highest level on record (2025 foundry industry data). Grey iron castings accounted for 40.2%, ductile iron castings for 27.2%, cast steel for 12.3% and aluminium-magnesium alloy castings for 17.3%. Aluminium-magnesium alloy casting output was 9.05 million tonnes, up 8.8%, with its share rising from 16.4% in 2023 to 17.3% in 2025; compared with 6.8 million tonnes in 2020, this is growth of 33.1% over five years, a compound annual growth rate of 5.88% (2025 foundry industry data). Consumption of magnesium alloy castings was about 180,000 tonnes, 0.3% of casting output (2025 foundry industry data). Total output in 2024 was 50.75 million tonnes, down 2.2% year on year (2024 data as reported elsewhere, a page whose data are presented as images, with these three figures not listed in the text page), and the 2020 base was 51.95 million tonnes (the 14th Five-Year Plan for the Foundry Industry).
Putting the two angles together makes the correspondence between process and downstream clear.
Sand casting corresponds to large, heavy parts in cast iron and cast steel. Machine tool beds, pump and valve bodies, wind turbine hubs, structural parts for construction machinery and diesel engine cylinder blocks range from tens of kilograms to tens of tonnes in piece weight and from single pieces to thousands of pieces in batch; sand casting's low pattern cost and wide size range cover exactly this segment. Cast steel accounted for 12.3% in 2025. Cast steel has a high pouring temperature and large shrinkage, which sand moulds can accommodate through their refractoriness and give, while investment casting and die casting are limited by shell strength and die life.
Investment casting corresponds to small precision parts in stainless steel, heat-resistant steel and high-alloy steel. Pump and valve bodies and trim, automotive precision parts, hardware and medical device components mostly weigh 0.1 to 100 kilograms, require higher precision than sand casting and have surface requirements close to the finished part. The investment shell is single-use and can produce thin walls and fine holes that sand casting cannot, at the cost of high unit cost, so it suits parts with higher added value. Cast steel output in 2024 was 6.28 million tonnes, 12.37% of the national casting total (2024 data as reported elsewhere); these figures are second-hand reporting and can only be used to indicate the scale of cast steel.
Mining and metallurgical heavy machinery, construction machinery and power generation equipment correspond to another segment. Crusher jaw plates and hammers, excavator bucket teeth and track shoes, hydraulic turbine runners and steam turbine cylinders are mostly made of high-manganese steel, heat-resistant steel and alloy cast steel, weigh from hundreds of kilograms to tens of tonnes, and are often ordered in sets. This segment is served by sand casting and partly by investment casting, and its demands on material and internal density exceed those on dimensional accuracy, so inspection agreements weigh more than tolerance grades. Castings for coal machinery, cement machinery and marine equipment also fall in this segment.
Die casting and low-pressure die casting correspond to thin-walled housings and structural parts in aluminium and zinc alloys. Automotive gearbox housings, engine blocks and heads, motor housings, filter cavities for communications equipment and mid-frames for consumer electronics mostly weigh 0.1 to 50 kilograms and are produced in batches counted in tens of thousands. The five-year growth curve of 33.1% for aluminium-magnesium alloy castings runs in the same direction as capacity expansion in die casting and low-pressure die casting, and the three-electric-system housings and integrated structural parts of new energy vehicles are the main drivers of that curve. The 2023 three-ministry document lists integrated die casting among advanced process directions (guiding opinions of the Ministry of Industry and Information Technology and two other ministries), which corresponds precisely to this class of part.
Centrifugal casting corresponds to pipe and bodies of revolution. GB/T 13295 covers ductile iron pipe from DN40 to DN3000 (GB/T 13295 extract page; see Chapter 5 for the edition note). Cast iron pipe and fittings are one of the main demand sectors after automotive (2025 foundry industry data); water supply and gas pipelines are its stable market, and process parameters are expressed as nominal diameter and wall thickness, differently from general mechanical castings.
The material mix downstream determines the process mix: the volume of cast iron and cast steel supports sand casting, and the growth of aluminium-magnesium alloys supports die casting and low-pressure die casting. At order level this correspondence acts as a filter: material and piece weight rule out one group of processes, precision and batch size rule out another, and the remaining candidates are usually no more than two.
Part Two: Scale and Geography
Chapter 8: The 56,600 Casting Entities in This Database
The Tianxia Gongchang enterprise database contains about 4.8 million manufacturing entities, and its industry field is recorded only down to the 31 major categories of manufacturing. In the Industrial Classification for National Economic Activities (GB/T 4754-2017), casting falls under two small classes: C3391 ferrous metal casting and C3392 non-ferrous metal casting, both within the medium class C339 (casting and other metal products manufacturing) under the major category C33, metal products. The database holds no codes for C339, C3391 or C3392; to draw figures at small-class granularity, keyword matching on three text fields is the only option: company_service (summary of main products), business_scope (business scope) and title (enterprise name).
The match terms fall into two groups. The general terms are "casting" and "castings"; the process and material terms include sand casting, investment casting (lost wax), lost foam, die casting, low-pressure die casting, gravity casting, centrifugal casting, cast steel, cast iron, ductile iron, grey iron, malleable iron, cast aluminium, cast copper and cast magnesium. An entity is counted if any term hits in any field, deduplicated by unique enterprise identifier. This yields 56,563 entities.
The nature of this figure needs to be stated first. It counts legal entities whose name, business scope or main-product summary mentions casting or its processes and materials, and it belongs to a different system from the stock registered by industry code. Wherever this article uses the term "this database", it refers to this basis; the shares, multiples and averages that follow all use these 56,563 entities as the denominator and cannot be interchanged with the industry statistics of the National Bureau of Statistics.
Insured persons total 2,710,793, of which 35,283 entities report more than zero, or 62.4%. Across all entities the average is 47.9 persons per entity; across entities with an insurance record it is 76.8. Insured persons are taken from the items publicly disclosed in enterprise annual reports; a zero does not mean no employees, since it may reflect failure to file in that year, non-participation in insurance or the use of dispatched labour, so 62.4% indicates only the coverage of filing and cannot be read as an employment coverage rate.
The table below compiles casting-related entities in this database by province (keyword-match basis, October 2026). Only the eight provinces with the largest number of entities are listed.
| Province | Entities | Share of national total |
|---|---|---|
| Shandong Province | 8,602 | 15.2% |
| Jiangsu Province | 7,357 | 13.0% |
| Guangdong Province | 6,440 | 11.4% |
| Zhejiang Province | 6,204 | 11.0% |
| Hebei Province | 4,243 | 7.5% |
| Liaoning Province | 2,782 | 4.9% |
| Henan Province | 2,761 | 4.9% |
| Fujian Province | 2,322 | 4.1% |
| Total of the eight provinces | 40,711 | 72.0% |
Shandong, Jiangsu, Guangdong and Zhejiang together account for 28,603 entities, or 50.6%; adding Hebei gives 32,846, or 58.1%. The top eight provinces total 72.0%, and the remaining provincial-level units divide 15,852 entities among them. The centre of gravity of the distribution lies on the eastern seaboard: seven of the top eight provinces are coastal, and Henan is the only inland province.
The top eight county-level units together account for 8.6% nationally, lower than the 72.0% of the top eight provinces; the two levels have different bases, so the two shares cannot be compared directly. The table below compiles the eight county-level units with the largest number of casting-related entities in this database (including prefecture-level cities without districts, whose statistical level is the same as that of counties).
| County-level unit | Province | Entities | Share of province | Share of national total |
|---|---|---|---|---|
| Dongguan | Guangdong | 971 | 15.1% | 1.7% |
| Boluo County | Guangdong | 829 | 12.9% | 1.5% |
| Botou | Hebei | 692 | 16.3% | 1.2% |
| Wujin District | Jiangsu | 639 | 8.7% | 1.1% |
| Nanhai District | Guangdong | 483 | 7.5% | 0.9% |
| Beilun District | Zhejiang | 464 | 7.5% | 0.8% |
| Zhuji | Zhejiang | 406 | 6.5% | 0.7% |
| Boshan District | Shandong | 394 | 4.6% | 0.7% |
| Total of the eight | — | 4,878 | — | 8.6% |
The eight county-level units total 4,878 entities, or 8.6% nationally. Dongguan leads with 971, followed by Boluo with 829; both are in Guangdong and together they account for 1,800 entities, or 28.0% of the province. Botou's 692 entities represent 16.3% of Hebei, the highest contribution to its own province among the eight places in the table.
There are at least four directions of undercounting. Plants that use only colloquial terms, such as fansha (sand casting) or jiaozhu (pouring), cannot be identified; plants whose name and business scope mention only specific products, such as machine tool beds, manhole covers, flanges or pump bodies, cannot be identified; where the casting shop is internalised in a finished-machine plant, the legal entity is the parent plant and the process name does not necessarily appear in the text; and entities that write only "metal products processing" or "machinery parts processing" without naming the process likewise fall outside.
The directions of overcounting are equally clear. Enterprises that manufacture and sell casting equipment, casting materials, refractory materials, trading in pig iron and scrap steel, dies, cleaning equipment and testing instruments all carry the word "casting" in their business scope; hardware plants that use a negative formulation in their business scope, such as "hardware processing (excluding casting)", still match; downstream finished-machine plants that purchase castings, which write "casting procurement" or "casting processing and assembly", are also counted; industry associations, research institutes and park management companies with "casting" in their names are included as well; and entities that have ceased production or been deregistered but remain in the database are counted if their name contains "casting".
Placing the database figure back against official statistics allows only an order-of-magnitude reference. The Fifth National Economic Census reports that at the end of 2023 there were 4.236 million industrial enterprise legal entities nationwide, of which 455,000 were in metal products (C33), with 7.651 million persons employed (Fifth National Economic Census bulletin). The 56,563 entities here equal 12.4% of the C33 legal entities, and the insured persons equal 35.4% of C33 employment. Both ratios cross statistical objects and industry boundaries: this database's set includes entities outside C33, and insured persons are not equal to persons employed. They indicate order of magnitude, not coverage.
A 2020 reply from the Ministry of Ecology and Environment put the number of foundry enterprises nationwide at about 23,000 (Ministry of Ecology and Environment), and this database's basis is 2.46 times that figure. The 23,000 is the number of foundry enterprises as given in the original text, which does not state its statistical boundary or year; this database's basis is keyword hits and is broader, so the difference cannot be decomposed precisely. The two figures cannot be used to infer industry growth. The database figure describes the structure of the entity distribution and is not used to assert an industry total.
Chapter 9: Shandong and Hebei: Botou, Laizhou and Linqing
Shandong has 8,602 entities, 15.2% of this database's casting entities, the highest of any province; Hebei has 4,243, or 7.5%, ranking fifth. With Liaoning's 2,782, the three Bohai Rim provinces total 15,627 entities, or 27.6%. The centre of gravity in this belt lies at county level: Boshan in Shandong enters the national top eight county-level units with 394 entities, and Botou in Hebei ranks third with 692.
Botou has a complete official basis. In August 2025 the Hebei Daily reported that the city has 322 foundry enterprises, 79 of them above designated size; cluster enterprises have an annual output value of RMB 26 billion (260 hundred million yuan); annual casting output is 2 million tonnes (200 ten thousand tonnes), 3.94% of national casting output (Hebei Daily). The 3.94% is the ratio Botou calculated against the 2024 national output of 50.75 million tonnes (China Foundry Association), and the 2025 figure of 52.3 million tonnes cannot be used instead.
Botou's present scale is what remains after remediation. The same report gives cumulative figures: 288 low-end enterprises converted to other production or exited, the number of foundry enterprises fell by nearly half, and production efficiency rose by more than a factor of 2; the share of mid- and high-end products in the cluster exceeded 50%; cumulative investment in transformation and upgrading reached RMB 6.806 billion; and 87 enterprises were recognised as China Green Foundry Demonstration Enterprises or China Green Foundry Enterprises. The halving of the enterprise count and the attainment of RMB 26 billion in output value occurred at the same time, which indicates that what exited was mainly low-end capacity. Whether output changed in step cannot be calculated, because public materials do not give the pre-remediation base.
Botou also has a group of figures that are easily confused. In 2024 the industrial cluster recorded operating revenue of RMB 26 billion, tax payments of RMB 381 million, and 28,000 persons employed (industry data); the annual output value of cluster enterprises given by Hebei Daily in August 2025 is likewise RMB 26 billion. Operating revenue and output value are two different indicators, and 2024 and 2025 are two different years; that the values coincide is a coincidence and cannot be merged into the sentence "RMB 26 billion for two consecutive years". What can be confirmed is that in September 2024 Xinhua News Agency gave a basis of 322 enterprises and RMB 24.7 billion (Xinhua News Agency), and a year later the enterprise count was unchanged while output value had increased by RMB 1.3 billion. The cluster's title evolved from "China's Hometown of Casting" and "China's Renowned Casting City" to "China's Renowned Green Casting City", and the change of title kept pace with remediation and green upgrading.
On this database's basis Botou has 692 entities, 2.15 times the official 322. The difference is consistent with the directions of overcounting listed in Chapter 8: casting equipment and materials, dies, trading, and downstream plants that purchase castings may all be counted; Botou also has a large number of tool-and-gauge and machining entities that supply castings. The two figures should not substitute for each other: the official figure describes enterprises whose main business is casting, while the database figure describes the density of entities carrying the word "casting".
Another cluster in Shandong with complete data is Laizhou. The Laizhou Municipal Bureau of Industry and Information Technology announced in 2022 that the city has 245 foundry enterprises, 30 above designated size and 215 below; planned annual capacity is 3 million tonnes, annual casting output is more than 2.2 million tonnes, annual operating revenue is more than RMB 9 billion, and employment is more than 14,000 persons (Laizhou Municipal Bureau of Industry and Information Technology). Equipment holdings are more than 500 medium-frequency electric furnaces (sets) and more than 350 main moulding installations (sets) such as moulding lines and core shooters, with a capacity of 0.25 to 4 tonnes per furnace.
Dividing the two figures gives about 2.0 medium-frequency furnaces and 1.4 moulding lines per enterprise. More than 500 furnaces against 245 enterprises, with a capacity of 0.25 to 4 tonnes per furnace, indicate that Laizhou's melting equipment consists mainly of small and medium-sized medium-frequency furnaces, with no large-scale centralised melting. More than 350 moulding lines against 245 enterprises, or 1.4 per enterprise, indicate that some enterprises run several lines at the same time. Both averages are estimates; official materials give only totals.
Laizhou's 3 million tonnes is planned capacity and its more than 2.2 million tonnes is actual output, a difference of about one quarter. Capacity and output appear side by side in the same material, which precisely illustrates that design capacity cannot be used as output. If more than RMB 9 billion in operating revenue is set against more than 2.2 million tonnes of output, the average price per tonne is about RMB 4,100, a figure that can serve only as an order-of-magnitude reference: operating revenue may include business other than castings, and both figures carry "more than", so their precision is limited.
Linqing is a section that this chapter must leave blank. Public materials from the Linqing municipal government cover only the bearing industry: more than 5,100 manufacturing enterprises and more than 10,000 self-employed businesses, annual operating revenue of more than RMB 50 billion, and annual output of 9 billion finished bearings covering more than 7,000 models (China Economic Net). Retrieval of multiple public documents did not obtain any summary of enterprise numbers, output or output value for a casting cluster. The bearing industry and casting belong to different industries, and the two bases cannot substitute for each other. This chapter therefore lists no summary figures for casting in Linqing.
Leaving a blank does not mean there is no casting. Public statistics do not list Linqing's casting entities separately, so their scale and composition cannot be confirmed. Completing this section would require at least three types of material: a municipal special plan for casting or a cluster cultivation programme; the environmental impact assessment and pollutant discharge permit lists for foundry enterprises held by the environmental protection department; and industrial legal-entity data grouped by medium category in the statistical yearbook. None of the three was obtained this time, and any attempt to convert bearing-industry figures into a casting scale would introduce unverifiable assumptions.
Chapter 10: Jiangsu: Changzhou, Zhuzhe in Liyang, and Wuxi
Jiangsu has 7,357 entities, 13.0% of this database's casting entities, ranking second. Jiangsu's 7,357 equals 85.5% of Shandong's 8,602, and the two provinces together account for 15,959 entities, or 28.2% of this database's casting entities. Within the province the single point of concentration is Changzhou: Wujin District has 639 entities, 8.7% of the province, and enters the national top eight county-level units.
A May 2024 reply from the Changzhou Municipal Bureau of Industry and Information Technology states that the city has more than 600 foundry enterprises, of which about 240 are industrial enterprises above designated size and 2 are A-share listed companies; leading enterprises include Hengli Hydraulic, Kehua Holdings, Sanxin Heavy Industry and Juling Foundry; and 10 enterprises, including Kaida Heavy Industry and Jingleng Casting and Forging, were selected in the first batch of enterprises meeting the Normative Conditions for Foundry Enterprises recognised by the China Foundry Association (Changzhou Municipal Bureau of Industry and Information Technology).
Of the more than 600 enterprises, about 240 are above designated size, a rate of about 40%; 10 selected in the China Foundry Association's first batch of enterprises meeting the normative conditions represent 1.7% of the more than 600. The Normative Conditions for Foundry Enterprises is group standard T/CFA 0310021 of the China Foundry Association, an industry self-governance instrument rather than a mandatory standard (Ministry of Industry and Information Technology). The two ratios indicate that Changzhou's casting entities are mainly small and medium-sized enterprises and that those meeting the group standard's normative conditions are a minority. "Above designated size" and "meeting the normative conditions" are two different thresholds: the former is drawn by operating revenue, the latter assessed against the conditions of a group standard, so 40% and 1.7% cannot explain each other.
On this database's basis Wujin District alone has 639 entities, almost equal to the official figure of more than 600 for the whole of Changzhou. This apparently contradictory result comes from the basis: the official figure counts foundry enterprises, whereas the database's keywords also cover die casting plants, casting-related entities and processing plants that write only "castings". When the two sets of figures are placed side by side, each should be labelled with its basis, and they cannot be used to validate each other. If Wujin's 639 are deducted from Jiangsu's 7,357 in this database, the remaining 12 prefecture-level cities divide 6,718 entities, an arithmetic average of 560 per city. This average serves only to show how high Wujin's share is and does not represent the actual distribution among the cities.
Zhuzhe in Liyang exemplifies the park-based approach to casting in Jiangsu. In 2023 the Xinhua Daily reported that the Zhuzhe Green Casting Park gathers the town's 12 foundry enterprises, with annual casting output of more than 300,000 tonnes, over 60% of local casting output (Xinhua Daily). In 2023 the town's casting output was 300,000 tonnes and sales revenue RMB 3.8 billion (Yangzhou Net, republishing Xinhua Daily). Zhuzhe was once recognised as the country's first "China Green Casting Town" and in 2020 was selected as a National Torch Special Industrial Base (the Liyang Green Casting Special Industrial Base) (China Jiangsu Net). Twelve enterprises producing more than 300,000 tonnes average 25,000 tonnes each; that the park accounts for more than 60% of local casting output indicates that the park carries the main body of local output.
Wujin's 639 and Zhuzhe's 12 represent two forms. Wujin has many entities on this database's basis, but no public data on its casting output, so no inference about plant size can be drawn from that; Zhuzhe's average of 25,000 tonnes across 12 enterprises is a per-plant average on the park basis. The two sets of figures allow comparison only of entity density and park concentration, not of plant size. The specific scale of urban supporting operations requires separate investigation.
Dividing RMB 3.8 billion by 300,000 tonnes gives an average price of about RMB 12,700 per tonne for Zhuzhe, three times Laizhou's roughly RMB 4,100 in 2022. This gap cannot be read directly as a difference in technology: the two places have different product structures and different statistical years, and it is not stated whether sales revenue includes business other than castings. What it does show is that "average price per tonne" cannot be compared directly between clusters, and this article does not rank clusters on that basis.
Wuxi is the second gap in this chapter. Retrieval did not obtain a municipal basis for Wuxi's foundry enterprise count and output after 2023. The available higher-level figure is Jiangsu Province as a whole: according to incomplete statistics, Jiangsu has more than 3,000 foundry enterprises, and the operating revenue of the province's foundry enterprises ranks among the highest in the country (Jiangsu Economic News). This database's 7,357 is 2.45 times that figure and, like Botou's 2.15 times, more than double, which indicates that the gap between the two bases appears at both levels. A 2010 industry article once grouped Suzhou, Wuxi and Changzhou as an area dense with die casting plants; that article is remote in time and can serve only as background, and this article does not cite its plant-count figures (industry data).
The gap at Wuxi affects the within-province breakdown for Jiangsu. Of Jiangsu's 7,357 entities, the only known county-level high point is Wujin's 639; Wuxi and other places have no usable municipal casting basis, so the province's internal centre of gravity cannot be determined. The difference between the official provincial basis of more than 3,000 and this database's 7,357 can likewise be explained only by basis and cannot be apportioned by city.
Chapter 11: Zhejiang: Ningbo Die Casting and Zhuji
Zhejiang has 6,204 entities, 11.0% of this database's casting entities, ranking fourth. Zhejiang's 6,204 equals 96.3% of Guangdong's 6,440, a difference of 236 entities. Among county-level units, Beilun District in Ningbo has 464 and Zhuji in Shaoxing has 406; together they account for 870 entities, or 14.0% of the province. Zhejiang has only these two units in the database's county-level top eight, namely the two high points of Beilun, Ningbo and Zhuji, Shaoxing; the process mix differs between the two, and the database data cannot be broken down further.
Ningbo's official basis is capacity share. A 2021 document from the Ningbo Municipal Bureau of Economy and Information Technology states that the city's die casting capacity accounts for about 10% of the national total and 42% of the provincial total; products cover key core automotive components, luminaires, home appliances, hardware tools and electronic communications (Ningbo Municipal Bureau of Economy and Information Technology). The 10% is a share of national capacity and the 42% a share of provincial capacity, so the two ratios have different denominators; the 2021 capacity basis likewise cannot be divided directly into the 2025 output of 52.3 million tonnes.
The same document gives targets for normative upgrading. Targets and stock are two different concepts: by 2023, 500 die-casting-related enterprises of all kinds were to complete normative upgrading, and by 2025, 800; and by 2025 the park entry rate of above-designated-size die casting enterprises was to reach more than 80%. The object of normative upgrading is "die-casting-related enterprises of all kinds", a wider scope than die casting producers. The two targets of 500 and 800 are two years apart, an average of 150 additional upgrades a year, which is a decomposition of the target; actual progress is not disclosed in public materials. Reading the 800 as the number of existing die casting enterprises in Ningbo would mistake the target for the stock; the 80% park entry rate is likewise a target value and cannot be taken as the current situation. The threshold for new, modified and expanded projects is an investment of more than RMB 10 million in main equipment such as die casting machines and melting furnaces, or a total of more than 5 die casting machines or more than 1,000 tonnes of total tonnage. This threshold applies to new and expanded projects and does not apply retroactively to existing enterprises. The total number of die casting enterprises in Ningbo does not appear in this document, and other statements in public materials use inconsistent bases, so this article does not adopt them. This set of data was published in July 2021 and the year should be retained when citing it.
On this database's basis Beilun has 464 entities and Zhuji 406, a difference of 58, representing 7.5% and 6.5% of Zhejiang respectively and falling in the same order of magnitude. Ningbo's 42% is the share of provincial die casting capacity while Beilun's 7.5% is the share of provincial entity count; numerator and denominator both differ, so the two cannot be multiplied or divided to obtain Beilun's share within Ningbo. Multiplying 42% by Zhejiang's 6,204 gives 2,606, a meaningless result, because the denominator is an entity count and the numerator is capacity. Beilun is a district under Ningbo, and the database data cannot separate die casting from sand casting, nor assign a municipal capacity share to the district level.
Zhuji has 406 entities, 6.5% of Zhejiang, the province's second unit to enter the national county-level top eight. Retrieval did not obtain any summary of enterprise numbers, output or output value for Zhuji's casting cluster; this chapter lists only the database entity count and does not supply official figures. The 406 is close to Beilun's 464, which indicates that Zhejiang's casting entities are not all concentrated in Ningbo; the two together still account for only 14.0% of the province, and the remaining 5,334 are dispersed across other counties, cities and districts. Zhuji is a county-level city under Shaoxing with a long history in hardware and machining, and casting exists mostly as an upstream step for hardware parts. This layer is industrial background; the database data cannot verify the composition of the 406 among die casting, sand casting and precision casting.
The share of aluminium alloy castings is rising. According to China Foundry Association data republished by China Die Casting magazine, output of aluminium (magnesium) alloy castings in 2025 was 9.05 million tonnes, up 8.8%, with its share rising from 16.4% in 2023 to 17.3% in 2025; compared with 6.8 million tonnes in 2020, this is growth of 33.1% over five years, a compound annual growth rate of 5.88%; consumption of magnesium alloy castings was about 180,000 tonnes, 0.3% of casting output (2025 casting industry data). The association's data are not broken down by process, so the 9.05 million tonnes cannot all be attributed to die casting. Ningbo's die casting capacity corresponds precisely to this type of demand.
Zhejiang's 11 prefecture-level cities total 6,204 entities; the two known county-level high points account for only 14.0%, and the remaining entities are mainly dispersed across counties. Ningbo's die casting capacity is 42% of the province, yet its entity count cannot be broken down at municipal level; capacity concentration and entity dispersion can coexist within the same province. Taking Ningbo's capacity share as an entity share would overstate Ningbo's position among Zhejiang's casting entities; conversely, taking Beilun's 464 as the whole of Ningbo would understate Ningbo's die casting scale. Zhejiang's casting geography therefore has two threads: one is the concentration of scale represented by Ningbo, with 42% of provincial die casting capacity; the other is the dispersed distribution represented by Zhuji and other places, measured by county-level entity counts. The former is supported by the downstream density of the automotive and electromechanical industries, the latter by local hardware and machinery supply; the two threads run in parallel within one province and cannot be summarised by a single figure.
Chapter 12: Guangdong: Dongguan, Boluo and Nanhai in Foshan
Guangdong has 6,440 entities, 11.4% of this database's casting entities, ranking third. Concentration within the province is high: Dongguan has 971, Boluo 829 and Nanhai 483, together 2,283 entities, or 35.4% of the province. The 2,283 equals 31.0% of Jiangsu's 7,357. Outside these three, the remaining counties and districts total 4,157 entities, or 64.6% of the province. Guangdong's county-level high points fall on the eastern shore of the Pearl River Estuary and in Huizhou: Dongguan and Boluo rank first and second among national county-level units, Nanhai sits at the junction of Guangzhou and Foshan, and no county or district in eastern, western or northern Guangdong enters the top eight.
Dongguan has 971 entities, 15.1% of the province and 1.7% of the country, the county-level unit with the most casting entities nationwide on this database's basis. Dongguan's industrial structure is dominated by electronics, home appliances, hardware and automotive parts, and die casting and zinc-aluminium alloy part production are common processes in these industries. This connection is industrial background; the database data cannot give the downstream affiliation of each Dongguan entity, nor can the 971 all be read as die casting plants.
Boluo has 829 entities, 12.9% of the province and 1.5% of the country, ranking second among national county-level units. Boluo is a county under Huizhou, and its entity count exceeds that of Nanhai District in Foshan by 346. Dongguan's 971 is 1.17 times Boluo's 829, a difference of 142. The top two county-level units are both in Guangdong, consistent with the density of the province's finished-machine and component industries. Boluo's composition likewise cannot be broken down from the database data; all that can be confirmed is that its entity density is higher than that of every county and district in the province except Dongguan.
Nanhai has 483 entities, 7.5% of the province, equal to 49.7% of Dongguan. Nanhai is Guangdong's third unit to enter the national county-level top eight; the other four districts of Foshan all fall outside, with fewer than 394 entities. Foshan is the gap in this chapter: public statistics for Foshan city and Nanhai District do not list casting separately, and the city's industrial map does not give a casting basis (Foshan industrial map). Nanhai's aluminium processing is dominated by extruded profiles, and extrusion is not casting, so Nanhai's entity count on a casting basis is lower than its aluminium processing scale would suggest. This chapter does not list enterprise numbers, output or output value for Foshan's casting and die casting clusters.
Guangdong's casting entities are concentrated in the Pearl River Delta and supply downstream electronics, home appliances and automotive parts. Die castings in electronics and home appliances are mostly thin-walled small parts, while sand castings in machine tools and heavy machinery are mostly medium-to-large parts; the database data cannot be broken down by process or part weight, and can only confirm that entity density moves in the same direction as downstream density. Foshan does not list casting separately, so the within-province breakdown for Guangdong can reach only the three points of Dongguan, Boluo and Nanhai; the remaining 4,157 have no county-level detail.
Guangdong's casting exports can be observed through national figures. From January to September 2025, exports under 20 tariff lines totalled 4.763 million tonnes, up 4.4% year on year, with a total value of USD 13.23 billion, up 3.7%; exports to the United States were 1.034 million tonnes, 21.7% of the total, down 11.9% year on year, with a value of USD 3.16 billion, down 5.4%; exports of aluminium alloy wheels were 763,000 tonnes, up 8.7%; automotive parts 2.148 million tonnes, down 1.1%; ferrous castings 1.662 million tonnes, up 9.4%; and rolls 190,000 tonnes, up 12.0% (China Foundry Association data). Total volume grew 4.4% while exports to the United States fell 11.9%, which indicates that the export increment came from markets other than the United States. The 5.4% fall in the value of exports to the United States is smaller than the 11.9% fall in volume; dividing the two gives a rise of about 7.4% in average price per tonne, an arithmetic result that does not represent an upgrading of product structure. This set of judgments uses the national basis and cannot be applied directly to Guangdong.
This set of figures must carry qualifications. The 20 tariff lines for customs "cast articles" are not equal to all casting exports: automotive parts move with complete vehicles and most non-ferrous castings are excluded, and the association includes this explanation in every monthly report. The 4.763 million tonnes can serve only as a trend observation on a consistent basis, not as China's total casting exports. The comparable figures for January to November 2024 are 5.57 million tonnes and USD 15.6 billion, of which ferrous castings were 2.065 million tonnes, up 20.9%; automotive parts 2.654 million tonnes, up 11.4%; and aluminium alloy wheels 851,000 tonnes, down 1.2% (2024 export data). The two periods cannot be added directly or annualised.
The centre of domestic demand is automotive. In 2025 demand for castings from the automotive industry was 15.8 million tonnes, 30.2% of all castings, an increase of 5% (2025 casting industry data); in 2024 the automotive share was 29.6% (2024 data), an increase of 0.6 percentage points in one year. Guangdong is a major province both for complete vehicles and parts and for electronics and home appliances; the demand for die castings from these two downstream sectors moves in the same direction as entity density in Dongguan, Boluo and Nanhai. The density of casting entities follows the density of downstream finished machines and parts, and Guangdong's three-point distribution is consistent with this relationship.
Chapter 13: Ningguo in Anhui and Fujian
Ningguo is the only casting cluster in Anhui for which this retrieval found a complete official basis. The Ningguo Municipal Bureau of Economy and Information Technology announced in June 2023 that, as of 2022, the city had more than 170 wear-resistant casting enterprises, including 97 above designated size and 23 with output value above RMB 100 million; 40 enterprises had obtained foundry industry access qualification, and 9,700 persons were employed; annual production capacity was 2 million tonnes, output value was RMB 12 billion, up more than 34% year on year (Ningguo Municipal Bureau of Economy and Information Technology). Calculated on 170 enterprises, the 97 above designated size account for 57%, the 23 with output value above RMB 100 million for 13.5%, and the 40 with access qualification for 23.5%. The three ratios indicate that Ningguo's cluster is dominated by above-designated-size enterprises and that fewer than one quarter hold foundry industry access qualification.
The basis at the end of 2025 changed its statistical object. A June 2026 report in China Quality News states that Ningguo's high-end equipment wear-resistant industry gathers more than 170 upstream and downstream enterprises, with 97 industrial enterprises above designated size and annual output value exceeding RMB 12 billion; in 2025 the city's wear-resistant casting exports were USD 142 million, up 13.6% year on year (China Quality News). The two instances of "more than 170" mean different things: in 2022 it referred to wear-resistant casting producers, in 2025 to upstream and downstream enterprises. The two figures cannot be placed side by side as one growth curve. The number of above-designated-size enterprises is 97 in both years, but the statistical objects are not entirely the same, so this shows only that the above-designated-size count did not change markedly and cannot be treated as the same group of enterprises. Wear-resistant castings are wear parts, and their downstream is the grinding stage in mining, cement and power generation; demand varies with mill operation and ore throughput, a different cycle from that of automotive and home-appliance castings.
The title "Capital of China's Wear-Resistant Castings" passed re-evaluation by a China Foundry Association expert panel in March 2023. From 21 to 23 March 2023 the re-evaluation panel visited Ningguo and concluded with unanimous approval (Ningguo municipal government). Title and output value are two different things: the former indicates the industry's recognition of the cluster, the latter its scale. Ningguo's output value went from RMB 12 billion in 2022 to "exceeding RMB 12 billion" in 2025; taking 2022 as the base, nominal growth over three years is very small. Growth falling from more than 34% in 2022 to marginal increase indicates that the expansion of wear-resistant castings slowed after 2022. This judgment is affected by the change of basis: the more than 170 in 2025 are upstream and downstream enterprises, and output value may include non-casting business. Exports of USD 142 million, converted at about 7 yuan to the dollar, come to about RMB 1 billion, equal to 8% of the RMB 12 billion output value; this is a rough calculation, and rounding the exchange rate introduces error.
Ningguo's 2 million tonnes is annual production capacity and its RMB 12 billion is output value for the year; the two are not on the same basis. Dividing one directly by the other would give an "average price per tonne" of RMB 6,000, a meaningless figure: capacity is design capability, while output value corresponds to actual output and that year's prices, so denominator and numerator do not match. Chapter 14 explains that conflating capacity with output is the most common error in casting cluster data. Dividing the 2022 output value of RMB 12 billion by more than 170 enterprises gives an average of about RMB 71 million (7,100 ten thousand yuan) per enterprise in Ningguo; dividing Botou's RMB 26 billion by 322 gives about RMB 81 million per enterprise. The two averages are close, which indicates that "few and specialised" and "many and comprehensive" can fall in the same range of plant size. This comparison is subject to three conditions: both output values are on a cluster basis, both enterprise counts are on a casting basis, and the two years differ. The average is meaningful only when the first two hold, and the third means that the difference contains price and product-structure factors.
On this database's basis Anhui does not enter the top eight provinces, which means the province's casting-related entities number fewer than Fujian's 2,322; Ningguo likewise does not enter the top eight county-level units, with fewer than 394 entities. These two ceilings are not inconsistent with the official figure of more than 170: Ningguo is a single-product cluster with a modest number of enterprises and a high concentration of output value and exports. Ningguo represents a "few and specialised" cluster form, different from the forms of Botou and Changzhou, which have larger numbers of enterprises.
Fujian has 2,322 entities, 4.1% of this database's casting entities, ranking eighth, equal to 27.0% of Shandong and 31.6% of Jiangsu. No county-level unit in Fujian enters the national top eight, which means the province has no county or district with more than 394 entities at a single point. Fujian's gap from Henan is 439 entities and from Liaoning 460, both far smaller than the 6,280 gap from Shandong. Retrieval did not obtain a provincial or municipal basis for Fujian's casting cluster enterprise count, output or output value; this chapter lists only the database entity count and does not supply official figures. Completing a Fujian section would require dedicated casting materials from the provincial industry and information technology department or cluster statistics from prefecture-level cities; this retrieval obtained neither. Fujian's casting entities can therefore serve only as an aggregate observation: it sits last among the top eight provinces, more than 3 times behind Shandong and Jiangsu and about 2.7 times behind Guangdong and Zhejiang.
Chapter 14: Formation and Migration of Industrial Clusters
Placing the figures from the preceding six chapters together, the conditions for the formation of a casting industrial cluster can be reduced to three: the density of downstream demand, the accessibility of raw materials and energy, and the remaining room in land and environmental capacity. Where all three are met, entity counts and output concentrate together; where only one is met, a place often remains confined to a single product or a single process.
Botou belongs to the downstream-support type. Large machine tool castings account for 25% of the national total, pump and valve castings for 35%, castings for automotive panel dies for 16%, tool and gauge platform castings for 60% and full-mould castings for more than 50% (industry data). This structure indicates that Botou's casting follows three downstream sectors — machine tools, pumps and valves, and dies — and that its products are mainly medium-to-large, single-piece and small-batch. Ningguo belongs to the single-product type: more than 170 wear-resistant casting enterprises organise production around the grinding stage in mining, cement and power generation and exported USD 142 million in 2025. Changzhou and Ningbo belong to the finished-machine support type: the former's more than 600 enterprises cover hydraulics, automotive parts and machinery supply, while the latter serves automobiles, luminaires, home appliances, hardware tools and electronic communications with die casting capacity equal to 10% of the national total.
Concentration has two measures. By enterprise count, the top eight provinces account for 72.0% of this database's 56,563 entities and the top eight counties for 8.6%; by output, the basis cited by the Ministry of Ecology and Environment in 2020 is that the top 4,500 enterprises produced more than 70% of national castings (Ministry of Ecology and Environment). The 4,500 equals 19.6% of the 23,000 foundry enterprises at that time. The two sets of figures show that output is concentrated in a few enterprises while enterprise counts are dispersed across more places. In 2025 national casting output was 52.3 million tonnes, up 3.1% on 2024's 50.75 million tonnes, so the scale is still expanding (2025 casting industry data). The 50.75 million tonnes of 2024 accounts for more than 45% of global casting output, and China has ranked first worldwide for 25 consecutive years since 2000 (China Foundry Association).
Botou provides a sample in which a falling enterprise count and a rising output value occur together. From 2017 to 2025, 288 low-end enterprises converted to other production or exited, the number of foundry enterprises fell by nearly half, production efficiency rose by more than a factor of 2, annual cluster output value reached RMB 26 billion, and the share of mid- and high-end products exceeded 50% (Hebei Daily). The pre-remediation output base is not public, so the effect of the exiting enterprises on total output cannot be calculated. What can be confirmed is that the exit took place at the level of enterprise counts, and that output value reached RMB 26 billion in the same period.
Park-based organisation is the second common direction. Zhuzhe gathered the town's 12 foundry enterprises into one park, whose output accounts for more than 60% of local casting output; Ningbo proposed that by 2025 the park entry rate of above-designated-size die casting enterprises should exceed 80%, and set the threshold for new, modified and expanded projects at more than RMB 10 million in main equipment investment, or more than 5 die casting machines or more than 1,000 tonnes of total tonnage (Ningbo Municipal Bureau of Economy and Information Technology). The threshold serves to restrict small-scale additions and expansions, while the park entry rate serves to gather the existing stock into parks; together they raise the lower limit of scale for individual entities.
Conflating capacity with output is the most common data error in this part. Botou has design capacity of 4 million tonnes and actual output of 2 million tonnes; Laizhou has planned capacity of 3 million tonnes and actual output of more than 2.2 million tonnes. Treating Botou's 4 million tonnes as output would produce a conclusion of 8% of the national total, whereas calculation on 2 million tonnes gives 3.94%. The capacity utilisation rates of the two clusters are about 50% and 73% respectively; these are estimates obtained by dividing the public figures, and official materials do not give capacity utilisation rates.
The direction of migration can be inferred from the stock structure but cannot be proved from cross-sectional data. On this database's basis the three Bohai Rim provinces account for 27.6%, Jiangsu and Zhejiang together for 24.0%, and Guangdong for 11.4%, the three areas together for 63.0%. Casting is distributed across the eastern seaboard and the industrial provinces of central China, and neither raw-material sites nor finished-machine sites have formed a single pole. This pattern is consistent with the three formation conditions: downstream density determines demand, ports and railways determine the accessibility of pig iron, scrap steel and coke, and environmental capacity determines where melting capacity can still be added.
Traces of all three forms of migration can be found in public materials. The first is park concentration, to which Zhuzhe's 12 enterprises and Ningbo's park entry target belong; the second is a falling enterprise count with rising plant size, to which Botou's 288 exits belong; the third is cross-regional relocation, for which public materials contain only scattered enterprise cases and no statistical basis above provincial level, so this article does not list it. The first two forms change the number and size of entities within a county; the third changes their distribution across provinces. This database's data are a cross-section from October 2026 and show only the result of migration, not the process.
The geography of casting clusters is determined jointly by downstream density and environmental capacity, and a fall in enterprise numbers can occur at the same time as a concentration of output. This judgment explains why, after 288 enterprises exited, Botou still has RMB 26 billion in output value, and why enterprise count, output and output value can rank the same cluster differently. The four denominators — enterprise count, output, capacity and output value — must be distinguished before reading any figure about a casting cluster.
Part Three: Market Access and Compliance
Chapter 15 The Repeal of the Foundry Industry Market-Entry Conditions and the Gaps in Current Rules
On 20 May 2013, the Ministry of Industry and Information Technology issued the Foundry Industry Market-Entry Conditions (《铸造行业准入条件》) by Announcement No. 26 of 2013, the text being dated 10 May 2013 (announcement reproduced on the State Council website). The circulation of this document within the foundry industry reaches far beyond its period of validity: in local environmental impact assessment reports, park investment-promotion materials and customer supplier audit forms, its scale and energy-consumption thresholds can still be read to this day.
It has been repealed. In June 2019, the Ministry of Industry and Information Technology issued Announcement No. 19 of 2019, repealing, with effect from 3 June 2019, 4 documents including the Foundry Industry Market-Entry Conditions, and with them three batches of announced enterprise lists, namely Announcement No. 15 of 2014, No. 13 of 2015 and No. 33 of 2016 (repeal announcement of the Ministry of Industry and Information Technology). The reason given in the announcement is "to implement the spirit of the Party Central Committee and the State Council on transforming government functions and deepening the reform of 'delegating power, improving regulation and upgrading services'". The way the lists existed says much about the design of this system: thresholds were set first, then qualifying enterprises were announced batch by batch, and the lists themselves constituted a market signal. With the lists repealed, the signal disappeared with them.
To understand the scale of the gap, one must first look at what it once governed. The quantitative provisions of the 2013 version are concentrated in a single scale table (full text of the Conditions reproduced on the Anyang Municipal Bureau of Industry and Information Technology website).
The table is compiled from Table 1 of the 2013 Foundry Industry Market-Entry Conditions, on scale thresholds.
| Product category | Existing enterprises (Group One of the Beijing, Tianjin, Shanghai, Jiangsu, Zhejiang, Shandong, Fujian, Guangdong, Hebei, Liaoning and Hainan areas listed in Table 1 / Class Three areas) | New enterprises |
|---|---|---|
| Iron castings | 5,000 tonnes / 4,000 tonnes | 10,000 tonnes or output value ≥RMB 70 million (7,000 万元) |
| Steel castings | 4,000 tonnes / 3,000 tonnes | 8,000 tonnes |
| Aluminium alloy castings | 1,200 tonnes / 1,000 tonnes | 3,000 tonnes |
| Copper alloy castings | 600 tonnes / 400 tonnes | 1,000 tonnes |
| Centrifugal ductile iron pipe | 100,000 tonnes or output value ≥RMB 500 million (50,000 万元) | 200,000 tonnes or output value ≥RMB 1 billion (100,000 万元) |
For the areas listed in Table 1, the threshold for new enterprises is twice the existing threshold for iron castings, steel castings and centrifugal ductile iron pipe, 2.5 times for aluminium alloy and about 1.7 times for copper alloy; for other areas the multiples are higher. Table 1 is graded by area; in other provinces and municipalities the existing thresholds are 4,000 / 3,000 for iron, 3,000 / 2,000 for steel, 1,000 / 800 for aluminium and 500 / 300 for copper, while the thresholds for new enterprises are the same nationwide. The intent of this gradient is clear: a way out for the stock, a raised threshold for the increment. The same document also contains a set of technical provisions: the sand reclamation rate is graded by sand type — ≥95% for clay sand, ≥90% for furan resin no-bake sand (reclaimed), ≥70% for alkaline phenolic resin no-bake sand (reclaimed) and ≥60% for sodium silicate sand (reclaimed); the melting rate of cupolas must be greater than 3 tonnes/hour for existing enterprises and greater than 5 tonnes/hour for new (expanded) construction; coreless line-frequency induction furnaces, aluminium-shell medium-frequency induction furnaces of 0.25 tonnes and above without a magnetic yoke, and oil-fired heating furnaces for foundry use must not be used; comprehensive energy consumption per tonne of iron castings is ≤0.44 tonnes of standard coal, and per tonne of steel castings ≤0.56 tonnes of standard coal. In Class One areas — scenic and historic interest areas, nature reserves, water source areas and other areas requiring special protection — foundry enterprises are not recognised.
These figures have only one use today: they describe a historical threshold. Presenting them as "the current national market-entry conditions" is the most common factual error in this industry.
The gap after the repeal has three layers. The scale threshold has gone: a new iron foundry with an annual output of two thousand tonnes no longer faces a national tonnage floor. The technical threshold has gone: the sand reclamation rate and the comprehensive energy consumption per tonne of castings are no longer items of market-entry review. The lists have gone: customers and local governments have lost a verifiable register of compliant enterprises.
The forces filling the gap come from four directions, none of which is called market access. The first is environmental impact assessment and pollutant discharge permitting; it governs emissions, and is covered in Chapter 16. The second is energy consumption and environmental performance, covered in Chapters 18 and 20. The third is workplace safety conditions, covered in Chapter 19. The fourth is industry self-regulation standards: the China Foundry Association group standard T/CFA 0310021, Specification Conditions for Foundry Enterprises (《铸造企业规范条件》), 2019 version, whose revision a 2023 document of the Ministry of Industry and Information Technology calls for, and which "encourages localities to refer to these conditions in guiding the standardised development of foundry enterprises" (2023 guiding opinions of the Ministry of Industry and Information Technology and two other ministries and commissions). Group standards are issued by social organisations and adopted by enterprises voluntarily; they carry no binding force. Their difference from market-entry conditions lies in the source of authority: market-entry conditions rest on administrative licensing, group standards on market choice.
Voluntary adoption does not mean nobody uses them. A reply from the Changzhou Municipal Bureau of Industry and Information Technology in May 2024 mentions that "10 enterprises including Kaida Heavy Industry and Jingleng Casting and Forging were selected in the first batch of enterprises meeting the Specification Conditions for Foundry Enterprises of the China Foundry Association" (reply of the Changzhou Municipal Bureau of Industry and Information Technology). Materials from the Ningguo Municipal Bureau of Economy and Information Technology in 2023 still record on the old basis: "40 enterprises have obtained foundry industry access qualification" (Ningguo Municipal Bureau of Economy and Information Technology), referring to the stock in 2022. A qualification repealed four years earlier still appears in local industry introductions, which shows that the inertia of a register outlasts the life of the document.
For a foundry, the practical conclusion of this chapter is that no piece of paper called a "national foundry access certificate" exists. Its compliance identity is assembled from four documents: the environmental impact assessment approval, the pollutant discharge licence, the workplace safety conditions, and the industrial policy and park admission requirements of its location. The first three have legal effect; the fourth determines whether it can be sited at all. If a counterparty still cites the 5,000-tonne threshold of 2013 when discussing a project, it must first be established whether the reference is to history or to a currently effective local document. Quite a few local documents did copy this set of figures verbatim, in which case it is the local document that takes effect; the ministerial announcement itself no longer produces any effect.
The compliance threshold has not disappeared; it has dispersed from a single nationwide market-entry table into four independent reviews. To judge whether a plant can stay in business, the tables to be examined have gone from one to four.
Chapter 16 Pollutant Discharge Permitting and Air Pollutant Emission Standards
GB 39726—2020, the Emission Standard of Air Pollutants for the Foundry Industry (《铸造工业大气污染物排放标准》), was issued on 24 December 2020, took effect on 1 January 2021 and is currently in force (metadata of the National Standards Full-Text Public System). The standard covers eight processes: metal melting, moulding, core making, pouring, shakeout and cleaning, sand preparation, casting heat treatment and surface coating (interpretation of the standard by the Ministry of Ecology and Environment). The eight items determine that a foundry's emission outlets are set up separately by process and form a set: melting has its own melting stack, sand preparation its own sand-preparation stack, and coating has a separate volatile organic compound outlet.
The compliance date for existing enterprises is 1 July 2023: "existing enterprises shall, from 1 July 2023, comply with the air pollutant emission limits set out in Table 1". New enterprises have complied since 1 January 2021. This time lag means that the same set of limits took effect in the industry in two batches, with the period from 2021 to the first half of 2023 a transition period in which old and new regimes coexisted.
The table is compiled from Table 1 of GB 39726—2020, on particulate matter emission limits (unit: mg/m³). No publicly accessible link to the original text was obtained for this article; the values were cross-checked against environmental impact assessment approvals from several places.
| Emission source | Limit |
|---|---|
| Cupola | 40 |
| Gas-fired furnace | 30 |
| Electric arc furnace, induction furnace, refining furnace | 30 |
| Moulding, core making, pouring | 30 |
| Shakeout and cleaning, sand preparation | 30 |
| Heat treatment, surface coating | 30 |
The difference between 40 and 30 has a basis. Cupolas burn coke, produce a large volume of flue gas and carry a high dust concentration, so the standard gives them a limit one grade looser: 40 is one third looser than 30. This difference conceals an easily overlooked consequence of upgrading: when a plant replaces its cupolas with induction furnaces, the corresponding row in the standard changes from "cupola 40" to "induction furnace 30", and the limit tightens accordingly. The furnace is changed, and so is the yardstick. The cupola melting-rate provision in the 2013 market-entry conditions (greater than 3 tonnes/hour for existing enterprises, greater than 5 tonnes/hour for new construction) is a scale basis, while the 2023 three-ministry document uses the basis of "encouraging the foundry industry to convert cupolas (10 tonnes/hour and below) to electric furnaces", which draws the line at 10 tonnes/hour. The two lines each have their own source and cannot be mixed.
The rules on the volatile organic compound side are different. The interpretation of the standard states: "where the initial emission rate of NMHC is ≥3 kg/h, the treatment efficiency of VOCs treatment facilities shall not be lower than 80%". Where the initial emission rate of non-methane total hydrocarbons (NMHC) is below 3 kg/h, the efficiency requirement is not triggered, but the concentration limit must still be met. This dividing line splits foundries into two classes: plants with a large coating volume using solvent-based coatings compete on treatment facility efficiency, with activated carbon adsorption, catalytic combustion and regenerative combustion each requiring an investment; plants with a small coating volume compete on meeting the concentration limit, and can pass with water-based coating substitution and local capture. One standard, two technical routes, two sets of accounts.
Compliance determination cannot rest on instrument readings alone. The measured concentration must be multiplied by the actual exhaust volume and then converted to the reference basis; the converted figure is the one against which enforcement is checked. Diluting the concentration by increasing exhaust volume does not work under this algorithm: exhaust volume and concentration are roughly inversely proportional, so the converted value is unchanged while the electricity bill rises. The correct action for an enterprise under this rule is to reduce fugitive emissions, raise capture efficiency and concentrate the exhaust volume at the main emission points.
The emission reduction estimate given when the standard was formulated is: particulate matter emissions to fall by about 30% in key areas and by more than 50% in other areas, with a total reduction of about 50,000 to 80,000 tonnes (5 万至 8 万吨) (interpretation of the standard by the Ministry of Ecology and Environment). This is a calculation made at the drafting stage, used to illustrate the strength of the standard; it cannot be treated as a statistical figure for actual emission reductions.
One gap must be stated. The revision status of this standard, and whether localities have issued stricter foundry air pollutant emission limits, could not be established from citable original texts within the scope of this verification, so this article lists no revised versions and no local limits. To confirm which tier applies, an enterprise should check three places: whether the province where the plant is located has issued a local standard, whether the environmental impact assessment approval for the project imposes separately stricter requirements, and the permitted emission concentration stated in the pollutant discharge licence. The strictest of the three applies.
The role of the pollutant discharge licence is to attach the limits in the standard to specific emission outlets. The standard specifies what a given process emits and the limit it must meet; the licence specifies how much this stack is permitted to emit, what the total quantity is, how monitoring is carried out and how often execution reports are submitted. Two plants in the same industry applying the same standard may differ several-fold in permitted emissions; the difference comes from the total-quantity allocation in the environmental impact assessment approval and the local allocation of environmental capacity. The air compliance cost of a foundry is ultimately determined by its permitted emissions; the limit in the standard is only a floor.
For buyers, the implication of this chapter is verifiability. Pollutant discharge licences are public information; the number of emission outlets, the pollutant items and the period of validity on the licence reflect the completeness of the plant's processes and its level of treatment. The absence of a surface-coating outlet at a plant usually means that it does not perform the coating process, so castings must be sent to another plant for processing and the lead time gains an extra step.
Chapter 17 Compliance Boundaries for Foundry Pig Iron, Scrap Steel and Recycled Metals
The raw-material arithmetic of casting is not complicated: pig iron, scrap steel, foundry returns and ferroalloys, plus coke or electricity. The compliance boundary does not lie in the raw materials themselves but in three questions: what is melted, what is used to melt it, and what the melt is turned into.
The first question is what the melt is turned into. The 2023 three-ministry guiding opinions contain a hard phrase: "it is strictly prohibited to add steel production capacity in violation of regulations under the name of casting and forging, or to produce steel billets and ingots in violation of regulations and sell them on the market" (2023 guiding opinions of the Ministry of Industry and Information Technology and two other ministries and commissions). The meaning of this sentence for a plant is very concrete: the output of the melting shop can only be castings. If a medium-frequency furnace has spare capacity and continuous-casts the molten iron into steel billets for outside sale, that falls within the prohibited items; establishing a project as a casting project while actually building steelmaking capacity likewise falls within the prohibited items. Casting and steelmaking overlap heavily in equipment — a medium-frequency induction furnace can melt cast iron and also make steel — and the boundary is drawn by the output and proved by the records.
The second question is what is used to melt. The same document proposes "promoting short-process casting and encouraging the foundry industry to convert cupolas (10 tonnes/hour and below) to electric furnaces". 10 tonnes/hour is where the encouragement line starts. The equipment provisions in the 2013 market-entry conditions were harder and listed elimination items outright: coreless line-frequency induction furnaces, aluminium-shell medium-frequency induction furnaces of 0.25 tonnes and above without a magnetic yoke, and oil-fired heating furnaces for foundry use must not be used. That document has been repealed, yet it is not rare for its equipment list to be cited in local documents; when replacing equipment, an enterprise needs to check the currently effective documents of its location and confirm whether the list remains valid. The choice of furnace type also affects another line, namely electricity consumption: for the same tonne of liquid metal, a 1-tonne furnace and a 5-tonne furnace differ by 40 kWh, and the scale effect shows most directly in melting, with the data given in Chapter 18.
The third question is what is melted. No mandatory national market-entry standard on the raw-material side was obtained in this verification, so it is not listed here. The practical constraints come from three places. The first is incoming inspection. The 2013 market-entry conditions required "chemical composition analysis and molten metal temperature measurement equipment at the furnace"; as a historical provision this shows the industry's basic requirements for raw-material composition control and remains the standard configuration of foundries today. The second is the economics of the charge mix. the Mysteel spot index page was captured on 4 October 2026; the page carries live quotations whose values change with each trading day, and no archivable point-in-time value was obtained for this article; at the time, the page showed scrap steel priced below foundry pig iron. The price gap points toward raising the scrap ratio, and the composition constraint pulls back against that direction: scrap steel comes from miscellaneous sources, with large fluctuations in sulphur, phosphorus and trace elements; ductile iron is sensitive to elements that interfere with nodularisation, and cast steel is sensitive to residual elements. Every percentage point of scrap added to the charge mix must be bought back with furnace-side analysis.
The difference between pig iron and scrap steel is not only in price. Pig iron has a high carbon content and few non-metallic inclusions, and provides good conditions for graphite nucleation, so a high proportion makes the graphite morphology of cast iron easier to control; scrap steel has a low carbon content and must be supplemented with a recarburiser, whose absorption rate and method of addition directly affect composition stability. The pig iron proportion in the charge mix therefore has a lower limit determined by the casting grade and section thickness, and price can only act above that limit.
The third is the record-keeping for foundry returns and purchased recycled metals. Gating systems, risers and reject castings that circulate within the plant are foundry returns; purchased scrap steel and purchased recycled aluminium ingots require records of origin and composition. There is no directly citable provision of a national standard on this point; in practice it follows customer audit requirements: customers in automotive, rail transit and pressure vessel sectors generally require suppliers to provide records of the origin and composition of each batch of raw material, and export orders are traced further back to the furnace batch number. A plant with incomplete records loses marks first at the customer audit stage.
The share of recycled metals is rising. In 2025 the output of aluminium (magnesium) alloy castings was 9.05 million tonnes (905 万吨), up 8.8% year on year, and their share of total casting output rose from 16.4% in 2023 to 17.3% (report in China Die Casting Magazine on Sina). The growth of aluminium and magnesium castings comes mainly from automotive lightweighting and die-cast structural parts, a line on which the proportion of recycled aluminium used is higher than in gravity casting. The rising share brings a process consequence: recycled aluminium has a broader impurity spectrum than primary aluminium, the window for melt treatment (refining, degassing, filtration) is narrower, and quality consistency depends more on furnace-side testing and pre-treatment.
One further constraint, often overlooked, comes from the status of foundry returns. While gating systems and risers circulate within the plant they are raw material; once transported off site they enter solid waste management, and commissioning an outside unit to handle them must follow the rules for solid waste. For a foundry, clarifying the destinations of three streams — gating systems and risers, dust removal ash and waste sand — determines its compliance status more than saving a few fen on raw material purchasing.
The compliance anchors on the raw-material side are three sets of records: the destination record for the output of melting, the equipment list for furnace types, and the composition record for each batch of raw material. Where the three do not agree, the problem usually surfaces first at the stage of capacity determination.
Chapter 18 Energy Consumption and the "Dual Carbon" Constraints
The energy consumption indicators for casting were once written into the market-entry conditions and lost their binding force with the 2019 repeal: comprehensive energy consumption ≤0.44 tonnes of standard coal per tonne of iron castings and ≤0.56 tonnes of standard coal per tonne of steel castings (full text of the Conditions reproduced on the Anyang Municipal Bureau of Industry and Information Technology website). The ratio of the two figures is 1.27: steel castings are about 27% higher than iron castings. The reason lies in the difference in melting temperature: the tapping temperature of cast steel is significantly higher than that of cast iron, and each step up in temperature increases both heat loss through the lining and the superheating time.
The same document contains a more detailed electricity consumption table, graded by electric furnace capacity. It has more engineering value than the comprehensive energy figures because it ties scale directly to energy consumption.
The table is compiled from Tables 3 and 4 of the 2013 Foundry Industry Market-Entry Conditions, on electricity consumption indicators (unit: kWh/t of liquid metal).
| Furnace capacity | Melting cast iron | Steelmaking |
|---|---|---|
| ≤0.5 tonnes | — | 730 |
| ≤1.0 tonnes | <630 | 720 |
| 1.5 tonnes | <620 | — |
| 2 tonnes | <610 | 710 |
| 3 tonnes | <600 | 700 |
| ≥5 tonnes | <590 | 690 |
The cast iron column falls from 630 to 590, with the 5-tonne furnace 6.3% lower than the 1-tonne furnace. The steelmaking column falls from 730 to 690, a decline of 5.5%. The two curves have the same shape and are offset by about 100 kWh: at the same capacity, steelmaking consumes about 14% more electricity than melting cast iron (720 against 630 at the 1-tonne grade).
The mechanism by which electricity consumption falls with capacity determines the choice between "more small furnaces" and "fewer large furnaces". Heat loss in an induction furnace has two parts: heat dissipation from the lining and the furnace mouth, proportional to the furnace's surface area; and the heating and holding time per melt, inversely proportional to the batch weight. Two 1-tonne furnaces have a larger lining area than one 2-tonne furnace and twice as many batch changes, so electricity consumption is naturally higher. This is why the 2013 table used capacity as the basis for grading. Today the table is no longer applied as a market-entry condition, but it remains a usable engineering reference.
The "dual carbon" constraints on casting are currently transmitted through three existing channels; no separate set of carbon indicators has been established.
The first is electrification. The 2023 three-ministry document encourages the conversion of cupolas of 10 tonnes/hour and below to electric furnaces, and the decision to convert or not requires two calculations: one is the price gap between coke and electricity — on 4 October 2026 the metallurgical coke spot index was RMB 1,390 per tonne (Mysteel); the other is that after conversion the particulate matter limit tightens from 40 to 30, as covered in Chapter 16. Where coke is cheap and electricity expensive, the payback period for conversion is long; where coke is expensive and electricity has an advantage, the payback period is short. Differences in electricity prices between parks within the same province are enough to change the conclusion.
The second is reduction targets. The same 2023 document lists five indicators to be achieved by 2025: build more than 10 industrial clusters with demonstration effects, develop more than 30 intelligent manufacturing demonstration factories, cultivate more than 100 green factories, reduce particulate matter emissions from the foundry industry by more than 30% compared with 2020, and achieve annual reclamation and recycling of more than 8 million tonnes (800 万吨) of foundry waste sand.
Set against the size of the industry, the order of magnitude of this group of targets becomes clear. On the Ministry of Ecology and Environment's 2020 basis there were about 23,000 foundries nationwide (interpretation of the standard by the Ministry of Ecology and Environment), so 100 green factories correspond to a proportion of about 0.4%. The role of demonstration targets is to set benchmarks; they do not reach most enterprises, which face the two hard constraints of emission limits and energy costs.
The third is waste sand. The historical thresholds for the sand reclamation rate were graded by sand type: ≥95% for clay sand, ≥90% for furan resin no-bake sand (reclaimed), ≥70% for alkaline phenolic resin no-bake sand (reclaimed) and ≥60% for sodium silicate sand (reclaimed). The difference between 95% for clay sand and 60% for sodium silicate sand is 35 percentage points, and the gap is determined by the reclamation process: clay sand can return to the moulding line with mechanical reclamation plus dust removal, whereas sodium silicate sand reclamation must address collapsibility and strength loss after reuse, with equipment investment an order of magnitude higher. The China Foundry Association's industry plan sets the target for waste sand reclamation and recycling at more than 8 million tonnes per year, and the same plan separately proposes "cultivating more than 200 green foundries" (the foundry industry's 14th Five-Year Development Plan). The 200 here and the "more than 100 green factories" of the 2023 three-ministry document come from two issuing bodies; the bases differ, and neither can substitute for the other, nor can the two be added together.
On the carbon accounting layer, no public document on carbon emission accounting methods or carbon allowance allocation for the foundry industry was obtained for this article, so no carbon emission factor per tonne of castings is listed. Only energy consumption and emissions indicators can be verified; the rest is left blank.
One further cost detail is worth remembering. The energy cost of casting is sensitive to furnace type and equally sensitive to batch organisation. On the same furnace, continuous production and frequent grade changes can differ by a considerable margin in electricity consumption per tonne; a grade change requires cleaning the furnace, adjusting the composition and reheating, and all of these actions register on the meter. The 2013 grading table was graded by capacity alone and did not incorporate batch organisation; an enterprise that reads the table directly when benchmarking energy consumption will reach an over-optimistic conclusion.
The meaning of energy consumption for a foundry has changed: it has moved from a market-entry review to a cost spread out year by year. No one checks the 0.44 and 0.56 in the market-entry conditions any longer, but electricity prices, coke prices and production restriction plans still affect gross margin year after year.
Chapter 19 Safety: Molten Metal, Dust and Confined Spaces
The core risks of casting come from three things: high-temperature liquid metal, combustible dust, and enclosed spaces that people must enter. The Criteria for Determining Major Accident Hazards in Industrial and Trade Enterprises (《工贸企业重大事故隐患判定标准》), Order No. 10 of the Ministry of Emergency Management, in force since 15 May 2023, breaks the first of these into four determinable situations, set out in Article 7 (the criteria published in the State Council Gazette).
The first: "assembly places of 5 types, namely conference rooms, activity rooms, rest rooms, changing rooms and shift handover rooms, are located within the floor area of a molten metal handling bay or a pouring bay". A molten metal handling bay is the longitudinal passage along which an overhead crane travels carrying a ladle of molten iron; a pouring bay is the bay in which the pouring stations are located. This determination item falls on the plant layout drawing: once a ladle overturns or spills, there is no place to take cover within the floor area. In many old plants, conference rooms, rest rooms and shift handover rooms sit right beside the pouring bay, for the sake of proximity to the workstations, and this rule turns "close" into a hazard.
The second: "melting furnaces, refining furnaces and holding furnaces used for casting are not equipped with emergency discharge and emergency storage facilities". When a furnace lining burns through, the power fails or cooling is interrupted, the liquid metal in the furnace needs somewhere to go. Emergency discharge directs the liquid metal into an accident pit or an emergency ladle; emergency storage is the vessel that catches it. The common precondition for both facilities is that they be kept dry, empty and accessible in normal times. An accident pit piled with moulding boxes or waste is equivalent, for the purpose of determination, to having no accident pit at all.
The third: "standing water is present in 8 types of areas, including furnace bottoms, furnace pits and accident pits, as well as moulding pits, pouring operation pits and molten metal transfer passages". Water in contact with molten metal vaporises instantly and expands sharply in volume, producing an explosion. This item makes "standing water" a hazard because water has too many sources: the groundwater table, roof leakage, cooling water leaks, floor washing and fire-fighting test water. The plum rain season and cleaning operations are the two high-incidence occasions. The engineering response is to raise the furnace foundation, build drainage channels and sumps, and install level alarms; the management response is to write the standing-water check into the pre-shift inspection.
The fourth: "the cooling water system is not equipped with monitoring and alarm devices for outlet water temperature and the difference between inlet and outlet flow, or the monitoring and alarm devices are not interlocked with the molten metal heating and transfer control systems". Water-cooled furnace walls, water-cooled cables and water-cooled crystallisers are common configurations on electric furnaces, and leakage into the molten bath directly triggers the explosion described in the previous item. The technical point of this item lies in the word "interlock": monitoring and alarm only give a warning, whereas an interlock can automatically cut off heating and transfer when a limit is exceeded. Instruments installed without an interlock make the devices no more than decoration.
The four determination items share one feature: all of them point to facilities and layout, and not one points to operating practice. This orientation determines the nature of the rectification. Installing an interlock requires shutting down production to run wiring; adjusting the layout of the pouring bay requires moving plant; an accident pit must be excavated and made waterproof. Safety investment at this level is capital expenditure, booked in the same category of accounts as environmental investment.
The lifting link is another concentration point of molten metal accidents. The criteria do not list it separately, but the risk remains: overhead cranes carrying molten iron ladles need reliable braking arrangements, lifting gear must undergo periodic non-destructive testing, and ladle linings must be checked on a per-ladle basis. These practices are common requirements in the industry; this article has not verified the currently effective regulatory provisions corresponding to each of them, so they are offered only as engineering practice guidance.
Dust risk divides into two lines. One is occupational health: dust generated by sand preparation, shakeout, cleaning and grinding contains free silica, and long-term inhalation leads to silicosis; protection rests on three things — enclosure, dust removal and personal protective equipment. The other is dust explosion: grinding and polishing dust from aluminium and magnesium castings is combustible metal dust, a dust cloud meeting an ignition source will deflagrate, and the damage from a secondary explosion is usually greater than from the first. Engineering measures on this line include wet dust removal, explosion-proof electrical equipment, equipment earthing and explosion venting facing a safe direction. Foundry dust removal systems are generally designed for dust removal efficiency, while combustible dust scenarios also require explosion-proof design; the two sets of requirements apply to the same system, and the latter is easily missed at the selection stage.
Confined spaces are densely distributed in a foundry: moulding pits, pouring pits, sand preparation trenches and bucket elevator bottoms, dust collector hoppers, sand silos, and furnace and flue inspection openings. The common features of such spaces are a narrow access, poor ventilation, and possible oxygen deficiency or accumulation of toxic gases. The basic sequence of work is isolation, ventilation, testing, standby supervision and then the work itself; if any of the five is missing, no one may enter. Cleaning dust collector hoppers and inspecting furnace interiors are the two high-incidence operations: the former may accumulate carbon monoxide, the latter may retain unburnt coke.
The relationship between safety and compliance has a particular feature in the foundry industry: a determination of major accident hazard is at the same time a veto item in customer audits. When a plant is ordered to suspend production for rectification, the notice the customer receives will not state the reason; the customer sees only that the delivery schedule has been interrupted.
To judge the safety level of a foundry, looking at the plant layout and the facility records is more accurate than looking at training records. All four major accident hazards can be confirmed against drawings and on site; there is no need to wait for an accident to verify them.
Chapter 20 The Impact of Heavy Pollution Weather Performance Grading on Capacity
Emergency emission reduction during heavy pollution weather lists casting as one of 39 key industries, on the basis of the Technical Guide for Formulating Emergency Emission Reduction Measures for Key Industries during Heavy Pollution Weather (2020 Revision) (《重污染天气重点行业应急减排措施制定技术指南（2020 年修订版）》) (public reply to a proposal by the Yizheng Municipal Bureau of Ecology and Environment). Grading evaluation looks at eight dimensions: equipment level and production process, pollution treatment technology, emission limits, fugitive emissions, monitoring and surveillance level, environmental management level, transport mode and transport supervision.
Of the eight dimensions, the first five are hardware and the last three are management. This structure means that performance grading cannot be obtained through a single upgrade; it requires an enterprise to maintain a verifiable set of records continuously: operating logs of treatment facilities, inspection logs of fugitive emissions, the connection status of monitoring data and entry and exit registers of transport vehicles. Hardware determines the ceiling of the rating; records determine whether the rating can be retained.
Grades are divided into four levels, A, B, C and D, with leading enterprises listed separately. Warnings are divided into four levels, tightening progressively from Level IV to Level I. Different grades have different emission reduction proportions under the same warning level.
The table is compiled from the Technical Guide for Formulating Emergency Emission Reduction Measures for Key Industries during Heavy Pollution Weather (2020 Revision), on differentiated emission reduction proportions for the foundry industry.
| Performance grade | Level IV warning | Level III warning | Level II warning | Level I warning |
|---|---|---|---|---|
| Grade B | ≥20% | ≥30% | ≥40% | ≥50% |
| Grade C | ≥30% | ≥40% | ≥50% | ≥60% |
| Grade D | ≥40% | ≥50% | ≥60% | ≥70% |
Enterprises rated as leading or Grade A in the performance grading evaluation may adopt emission reduction measures at their own discretion. This provision draws a substantive dividing line between Grade A and Grade B: Grade B must reduce emissions in proportion, while Grade A decides for itself how to do so.
The particularity of the foundry industry lies in the position of the melting process. Casting production begins with melting; once melting stops, all downstream processes have nothing to work on. Local contingency plans deal with this point directly: during an orange warning, the melting process of Grade C enterprises is suspended; for Grade D enterprises, all gas-related processes are suspended and road transport is halted (emergency response measures for heavy pollution weather published by a local government). The practical severity of the two measures differs: for Grade C, stopping melting amounts to stopping the whole plant, though the wording attaches to a process; Grade D adds a transport ban on top, so finished castings cannot be shipped either, inventory piles up in the plant, and delivery moves from delayed to impossible to promise.
The cost impact of a production stoppage on a foundry is greater than on most manufacturing. Restarting a cold medium-frequency furnace requires re-sintering the lining, and repeated stopping and starting shortens lining life; once a cupola's blast is cut, the coke and charge in the furnace must be cleared out, and reignition and heating on restart take several hours; prepared charges and finished moulds must be rescheduled, resin-bonded sand has a limited usable time, and moulds left overnight may be scrapped. These losses do not appear in the emission reduction proportion; they appear in the cost per tonne.
The direction of the incentive is clear. Shandong's formulation is: "enterprises that have not reached environmental performance Grade B or above before the autumn and winter of 2025 will in principle no longer be included in the list of guaranteed enterprises" (ibid.). Guaranteed enterprises may reduce production less, or not at all, during warnings. The practical meaning of this sentence is that Grade B becomes the threshold for entering the guaranteed list, and Grades C and D lose production continuity during the autumn and winter heating season. For foundries oriented to northern markets with tight delivery schedules, the performance grade is directly equivalent to delivery capability.
The specific emission limit values for the four grades were not obtained from an original text in this verification, and several second-hand accounts contradict one another, so this article does not list them. To confirm the grading indicators for its own industry, an enterprise should check the performance grading technical guide and the annual rating results published by the ecology and environment department of its location, together with its own entry in the heavy pollution weather emergency emission reduction list.
For buyers, this chapter offers an operational judgment: incorporate the supplier's performance grade into lead-time risk assessment. A Grade C foundry in the north during autumn and winter may stop melting because of a single orange warning, delaying delivery by several weeks; the exact duration depends on the warning level and the restart arrangements, and no statistics were obtained for this article. A Grade A plant in the same region may reduce emissions at its own discretion, so its lead time is less affected. On the same order, where quotations differ by a few percentage points, lead-time risk may differ by an order of magnitude.
Performance grading also changes the way capacity is measured. The emission reduction proportion in the emergency emission reduction list is calculated on the baseline emissions verified in the list; the baseline is usually taken as emissions or output under normal production conditions. For an enterprise already operating below capacity, the absolute reduction during a warning is small, and complying with the proportion is relatively easy; a plant running at full capacity must genuinely stop lines to reduce emissions in proportion. This arithmetic puts plants with high capacity utilisation under greater pressure during warnings, contrary to intuition.
Performance grading has turned "whether continuous production is possible" into a rateable attribute. A foundry's capacity is no longer determined only by equipment and orders; in autumn and winter it is also determined by a grade certificate.
Chapter 21 How Compliance Reshapes the Geography of Capacity
The previous six chapters dealt with rules; this chapter deals with their spatial consequences.
The most telling sample is Botou. A report in Hebei Daily in August 2025 gives the figures: the city has 322 foundries, 79 of them above designated size; cluster enterprises have an annual output value of RMB 26 billion (260 亿元); annual casting output is 2 million tonnes; and this accounts for 3.94% of national casting output (Hebei Daily report). The same report contains a more critical sentence: "288 low-end enterprises converted or exited, and the number of foundries fell by nearly half, yet production efficiency more than doubled", over the period from 2017 to 2025 cumulatively. 322 plus 288 equals 610, an exit proportion of 47.2%, consistent with "nearly half".
The mechanism these figures illustrate is screening. The bulk of compliance cost is fixed cost: dust removal systems, performance grading upgrades, monitoring connections and relocation into parks — the absolute amount of each has little to do with output. Suppose a given fixed compliance investment is RMB 500,000 (50 万元); spread over 5,000 tonnes it is RMB 100 per tonne, and spread over 500 tonnes it is RMB 1,000 per tonne — this is only an arithmetic example illustrating the dilution mechanism and does not represent the actual investment of any enterprise. The smaller the plant, the higher the unit compliance cost and the earlier it exits. Botou's 288 exits and 322 survivors are two outlets of the same sieve.
One basis confusion must be guarded against. Botou's designed capacity is 4 million tonnes and its actual output 2 million tonnes, a factor of two apart (Botou cluster data published on an industry platform). Dividing 4 million tonnes by national output would yield a conclusion such as "Botou accounts for 8% of the country", more than double the 3.94% in the report. The 3.94% is calculated from 2 million tonnes and the national figure of 50.75 million tonnes in 2024 (China Foundry Association 2024 output data). The distinction between capacity and output carries a further meaning in the compliance context: idle capacity is the denominator for the next round of environmental and energy indicator allocation, and also a bargaining chip in local investment promotion.
In Botou too, the other side of the account is technological upgrading: "cumulative investment in transformation and upgrading of RMB 6.806 billion (68.06 亿元)", and "87 enterprises have been rated as China Green Casting Demonstration Enterprises or China Green Casting Enterprises". The enterprises that stayed and completed the transformation obtained order concentration and qualification endorsement. The cluster's share of mid- and high-end products broke through 50%, a figure that occurred in the same seven years as the 288 exits.
Park-based clustering is the second spatial consequence. The sample of Zhuzhen town in Liyang, Jiangsu, is: the green casting park gathers 12 foundries from the whole town, with annual casting output of more than 300,000 tonnes, accounting for more than 60% of local casting output (Xinhua Daily report). Twelve enterprises producing sixty percent of output — that proportion is itself concentration. The park has two functions: centralised treatment of pollutants, and centralised allocation of environmental capacity and land quotas. For local governments, parks make total emissions controllable; for enterprises, entry into a park is a precondition for obtaining a pollutant discharge licence and new capacity.
Ningbo took another route, writing park entry into a timetable and investment thresholds. A 2021 document of the Ningbo Municipal Bureau of Economy and Information Technology proposed "completing the standardisation and upgrading of 500 die-casting-related enterprises of all kinds", and by 2025 "standardising and upgrading 800 die-casting-related enterprises", and set the target of "a park entry rate of more than 80% for die-casting enterprises above designated size"; the threshold for new construction, alteration or expansion is "investment in main equipment such as die-casting machines and melting furnaces of more than RMB 10 million (1,000 万元); or a total of 5 die-casting machines or a total tonnage of more than 1,000 tonnes" (Ningbo Municipal Bureau of Economy and Information Technology). The RMB 10 million equipment investment line limits the entities of new capacity to those with capital expenditure capability; the 80% park entry target moves existing enterprises out of villages and into parks. Together, the two actions change the industrial land structure of a county.
Differences between localities are equally clear. Laizhou in Shandong had 245 foundries in 2022, of which 30 were above designated size and 215 below (Laizhou Municipal Bureau of Industry and Information Technology), with below-scale enterprises accounting for 87.8%. Calculated from the equipment holdings in the same material, the city has more than 500 medium-frequency electric furnaces and more than 350 moulding lines, averaging about 2 furnaces and 1.4 moulding lines per enterprise; this is a calculation made by this article from the original data and is not an official statistic. Laizhou's pattern is many enterprises each of small size, in contrast with Botou's halved count and enlarged units. The two patterns correspond to two stages of compliance pressure: Botou has been through the sieve, while Laizhou's screening is still under way.
One direction of screening that is easily overlooked is order transfer. When Grade D or Grade C enterprises suspend production during warnings, the orders do not disappear; they flow to Grade A and Grade B enterprises in the same region that can produce continuously. This layer of transfer does not appear in enterprise count statistics; it appears in the seasonal fluctuation of capacity utilisation. For buyers, this means that lead-time risk differs among suppliers within the same cluster, and assessment must be done enterprise by enterprise, not region by region.
Park-based clustering also brings new constraints. After entering a park, the prices of land, steam and wastewater treatment are set by the park, and the enterprise's fixed cost structure changes from "own plant, self-built treatment" to "rent, service fees, charges by volume". The impact of this conversion on cash flow often only becomes visible in the second year after entry.
Compliance has raised the minimum economic scale of casting, and the increase comes mainly from the cost of stoppages. Equipment can be bought on credit; orders lost and customers switched during a stoppage cannot be made good. This constraint explains why capacity is concentrating from administrative villages into parks, and why the number of enterprises in a cluster falls faster than output.
Part Four: Cost and Equipment
Chapter 22 Cost Structure: Metal Charge, Coke and Electricity, Sand, Labour, Moulds and Environmental Protection
Casting cost items are collected by operation and can be divided into eight blocks: metal charge; coke and electricity consumed in melting; moulding materials (base sand, resin, binder, coating); labour; moulds and tooling; environmental treatment; depreciation; and outsourced processing. The proportions of these eight blocks vary widely between processes, and the proportions for the same process vary just as widely between products. Public materials contain no cost-structure statistics covering the whole industry; the only verifiable complete breakdown is the precision-component basis disclosed by three listed companies in their responses to enquiry letters.
Table 1 Cost Composition of Three Precision-Component Enterprises (Share of Operating Cost)
| Item | Lianhe Precision (2019–2021 average) | Huaxiang (2021–2023 average) | Baida Precision (2021–2023 average) | Three-company average |
|---|---|---|---|---|
| Direct materials | 36.72% | 50.65% | 45.58% | 44.32% |
| Direct labour | 15.52% | 17.64% | 17.28% | 16.82% |
| Manufacturing overhead (including power) | 47.76% | 31.70% | 37.14% | 38.87% |
Table compiled from Nante Technology's public offering enquiry response (2025-04-29). The three-company ranges for direct labour and manufacturing overhead are 15.52%–17.64% and 31.70%–47.76% respectively.
Adding the three rounded values in the table gives 100.01%; for Huaxiang alone the three items total 99.99%. Both show that the disclosed denominator is operating cost, and that period expenses such as selling, administrative and research and development expenses are not included. This determines how the table can be used: it describes the division within manufacturing cost and cannot be discussed together with gross margin or net margin.
The most likely misuse of this table lies in its basis. All three are precision-component enterprises, whose products include machining operations; direct materials include blanks and some purchased parts, and manufacturing overhead includes depreciation of machining equipment. Using 44.32% to infer the material share of a pure sand-casting plant would go in the wrong direction. The conversion for Laizhou below indicates the order of magnitude of the deviation.
The order of magnitude of the metal charge can be read directly from the spot index. The Mysteel spot index page was captured on 4 October 2026; the page carries live quotations whose values change with the trading day, and this article did not obtain a time-stamped value that can be archived; at the time, the page showed scrap steel prices below casting pig iron. As the proportion of scrap in the charge increases, the metal charge cost falls accordingly, and the extent of the fall depends on the price difference between pig iron and scrap on that day.
Fuel and power for melting are two separate cost lines. Cupolas use coke; on the same day the metallurgical coke spot index stood at RMB 1,390 per tonne; electric furnaces use electricity, which on the basis of the electricity-consumption ceilings in Chapter 23 amounts to 590–630 kWh per tonne of molten metal. The relative price of coke and electricity determines which route is cheaper, and that ratio varies with region and season. Public materials contain no verifiable basis for fuel-and-power cost collected per tonne of castings, so this article gives no percentage for this item.
Laizhou provides a comparison on a different basis. In 2022 the city's foundries had an annual output of nearly 2.2 million tonnes (220 余万吨) of castings and operating revenue of more than RMB 9 billion (90 余亿元); dividing one by the other gives operating revenue of about RMB 4,100 per tonne of castings. Substituting the index at the time of capture, and ignoring burning loss and remelting, the metal charge alone equals 60%–74% of revenue per tonne. The two years differ, so this proportion serves only as an order-of-magnitude reference and cannot be taken as the actual cost in 2022; operating revenue is on the cluster-enterprise basis and may include non-casting business, which is counted in full here. This range is far above the 44.32% given above, because Laizhou is dominated by sand-cast iron parts with little machining, whereas the precision-component disclosure includes machining. The two figures each hold on their own terms, but applying one on the other's basis makes both wrong.
The cost of moulding materials is determined jointly by consumption and make-up quantity. Base sand circulates within the system, and the quantity of new sand actually purchased depends on the used-sand reclamation rate. The 2013 Foundry Industry Access Conditions (《铸造行业准入条件》) once required a clay-bonded sand reclamation rate of no less than 95%, furan resin no-bake sand no less than 90%, alkaline phenolic resin no-bake sand no less than 70%, and sodium silicate sand no less than 60%; that document was repealed by Announcement No. 19 of 2019 of the Ministry of Industry and Information Technology (MIIT), and this set of values can only be cited as a historical threshold. Converted by reclamation rate, for the same 100 tonnes of moulding material in circulation, clay-bonded sand requires 5 tonnes of new sand to be added and sodium silicate sand requires 40 tonnes, a difference of 8 times. Resin and hardener are added in a certain proportion of sand weight, and consumption varies with core complexity and the sand-to-metal ratio; public materials contain no verifiable unit price or unit-consumption basis, so this article gives no figures.
Labour cost accounts for 16.82% on the disclosed basis, which corresponds to precision-component enterprises whose degree of automation is above the industry average. In sand casting, the three stages of manual moulding, core making and cleaning employ the most workers and offer the greatest scope for automation; in die casting, part removal, deburring and inspection likewise depend on labour, but the cycle is set by the die-casting machine, so labour mainly affects equipment utilisation.
Moulds and tooling are amortised items; the amortised amount equals the total mould price divided by the batch over which it is spread, so the larger the batch, the lower the amortisation per piece. On the basis given on a company's process page, the economic batch size for die casting is 5,000 pieces per year or more, and for low-pressure die casting 1,000 pieces per year or more; behind these two thresholds is the combined effect of mould amortisation and die-change time. Public price ranges for casting moulds were not obtained, so this article gives no unit prices.
Environmental costs include dust removal, desulphurisation, volatile organic compound treatment, waste-sand disposal and online monitoring. The limits for particulate matter and volatile organic compounds are given in Chapter 16, and the effect of performance grading on output in Chapter 20; only cost items are counted here. The running cost of a dust-removal system is determined by the air volume and the filtration velocity, and the air volume in turn by how well the exhaust hood is sealed; the degree of sealing is the most elastic item in environmental cost. The comparison between disposal and reclamation costs for waste sand depends on the local landfill price and the investment in reclamation equipment; public materials contain no uniform basis.
Cost structure changes with process, material, batch size and whether machining is included; no general table of proportions is sufficient to support a quotation. The verifiable approach is to calculate separately the metal charge on the day's index, melting on electricity consumption or coke consumption per unit, sand on the reclamation rate, and labour on hours, then compare the result with peers in the same cluster.
Chapter 23 Melting: Cupola, Medium-Frequency Furnace and Duplex Melting
The first operation in a casting shop is melting. Melting equipment determines three things: what materials can be melted, the temperature and compositional stability of the molten metal, and the energy consumption per tonne of molten metal.
A cupola uses coke as fuel; the charge is added in layers from the top, hot blast is blown in through the tuyères, and the iron charge is preheated, melted and superheated as it descends. Its strengths are continuous tapping, high productivity and high thermal efficiency, making it suitable for large-batch production of a single grade; its weaknesses are a weak ability to adjust composition, marked carbon and sulphur pick-up, and a heavy burden of flue-gas treatment. The 2013 Foundry Industry Access Conditions required existing cupolas to have a melting rate above 3 tonnes per hour and new or expanded ones above 5 tonnes per hour; that document was repealed by MIIT Announcement No. 19 of 2019, and the values serve only as a historical comparison. In 2023, the Guiding Opinions on Promoting the High-Quality Development of the Casting and Forging Industry (《关于推动铸造和锻压行业高质量发展的指导意见》), issued by three ministries and commissions, proposed promoting short-process casting and encouraging cupolas of 10 tonnes per hour and below to be converted to electric furnaces. The direction is that small-capacity cupolas are at a disadvantage on both environmental protection and energy consumption.
A medium-frequency induction furnace converts mains-frequency current to a higher frequency and feeds it to the induction coil; the charge heats up through eddy currents generated in the alternating magnetic field. The strengths of the electric furnace are controllable composition and temperature, the ability to start and stop at any time, and no sulphur pick-up from coke; its weaknesses are high electricity consumption, a limited lining life, and sensitivity to charge cleanliness. Table 3 of the 2013 Access Conditions gives the electricity-consumption ceiling for melting cast iron in a coreless induction furnace (including medium frequency), and Table 4 gives the electricity-consumption ceiling for melting steel in an induction furnace.
Table 2 Electricity Consumption Ceilings for Melting in Induction Furnaces (kWh per tonne of molten metal)
| Capacity | Melting cast iron | Melting steel |
|---|---|---|
| 0.5 tonnes and below | — | 730 |
| 1.0 tonnes and below | 630 | 720 |
| 1.5 tonnes | 620 | — |
| 2 tonnes | 610 | 710 |
| 3 tonnes | 600 | 700 |
| 5 tonnes and above | 590 | 690 |
Table compiled from Tables 3 and 4 of the 2013 Foundry Industry Access Conditions (page reproducing the full text). The cast-iron column has no 0.5-tonne tier and the steelmaking column has no 1.5-tonne tier; the capacity tiers in the two columns do not correspond exactly.
These values are all ceilings; in measurement, a figure below the ceiling counts as compliant. In actual operation, affected by charge cleanliness, the number of melt starts and holding time, the electricity consumption of a single heat can exceed the ceiling, and this excess enters the electricity bill only.
Duplex melting links a cupola and an induction furnace in series: the cupola handles melting, and the induction furnace handles superheating, holding and fine adjustment of composition. It retains continuous melting capacity while gaining compositional control. It applies to situations of a single grade, a large batch and continuous pouring; the investment is higher than for a cupola alone, and the running cost depends on the relative price of coke and electricity.
The Institute makes an illustrative calculation with the above parameters to show the relationship between electricity consumption and yield. The inputs are: 1 tonne of acceptable iron castings; a process yield (net casting weight ÷ poured molten metal weight) of 70%; a reject rate of 5%; a melting loss of 3%; and an electricity consumption of 600 kWh per tonne of molten metal for the medium-frequency furnace, taking the ceiling value for a 3-tonne furnace. The calculation gives poured molten metal of 1÷0.70÷0.95≈1.504 tonnes, melting electricity consumption of 1.504×600≈902 kWh, and new metal charge make-up of 1×1.03＝1.03 tonnes; runners, risers and rejected parts are remelted in-house and not counted as new charge. The sensitivities are: with the process yield falling to 65%, electricity consumption rises to 972 kWh, an increase of 7.7%; with it rising to 75%, electricity consumption falls to 842 kWh, a decrease of 6.7%; with the reject rate rising from 5% to 10%, electricity consumption rises to 952 kWh, an increase of 5.6%. All three assumptions are set by the Institute and are used only to demonstrate the method; they do not represent the actual level of any enterprise. The electricity price is entered as the local catalogue tariff, and this article gives no uniform unit price.
On this example, 1.03 tonnes of new charge costs RMB 2,530 if all priced on the scrap steel index and RMB 3,135 if all priced on the pig iron index, a difference of RMB 605. Actual charge is a mixture of pig iron, scrap steel and return material, and the ratio is determined jointly by material, furnace type and material source; public materials contain no nationally uniform basis for the ratio.
Melting loss comes from oxidation of elements and from carry-over in the furnace gas. An electric furnace has a lower melting loss than a cupola; the cleaner the charge and the smaller the proportion of return material, the lower the loss. It is compensated by increasing the proportion of pig iron or carburiser in the charge, and this part enters the metal charge cost directly.
Short-process casting means that molten iron from a blast furnace goes directly into an electric furnace for temperature and composition adjustment and then to pouring, without ingot casting or remelting; what is saved is the 590–630 kWh per tonne consumed in the remelting of pig iron. The conditions for putting it into practice are the physical distance between the foundry and the blast furnace, and control of the temperature drop during transport of the molten iron. The 2023 guiding opinion lists the short process as a direction for promotion; actual siting is constrained by the layout of steel capacity.
The comprehensive energy-consumption indicators given in the 2013 Access Conditions are no more than 0.44 tonnes of standard coal per tonne of cast iron and no more than 0.56 tonnes of standard coal per tonne of cast steel, and these are historical thresholds. The current basis for industry self-regulation is the China Foundry Association group standard T/CFA 0310021, Specification Conditions for Foundry Enterprises; it is a group standard and not mandatory, and the revision requirements are given in the MIIT document of 2023. Where local environmental impact assessments and park admission still cite the 2013 figures, enterprises need to check whether the basis for that citation has already lapsed.
Melting stability rests on three things: pre-furnace chemical composition analysis, molten metal temperature measurement and furnace lining management. The 2013 Access Conditions required pre-furnace chemical composition analysis and molten metal temperature measurement equipment, which was the minimum line of more than ten years ago. Enterprises that still judge temperature by eye from the spark today can absorb compositional fluctuation only by increasing margins, and those margins ultimately become metal charge and machining hours.
The equipment stock in Laizhou shows the melting profile of one cluster: 245 foundries together have more than 500 medium-frequency electric furnaces (sets), with capacities between 0.25 and 4 tonnes. That is about 2 furnaces per enterprise on average (our estimate). The upper limit of 4 tonnes means that the molten metal from a single heat is limited; large castings require several heats to be poured together, and the temperature drop and compositional deviation must be handled in the process.
Choosing a melting route is a choice of whether to stake cost on coke or on electricity, and whether to stake quality control on continuous production or on adjustment heat by heat. The relative price of coke and electricity, local emission requirements, and the material and unit weight of the product together determine the route; no single route is superior under all conditions.
Chapter 24 Moulding Lines and Core Making
Moulding compacts moulding sand around the pattern to form the mould cavity; core making produces the internal cavities and complex passages of the casting. These two operations determine the dimensional accuracy, surface roughness and incidence of internal defects of the casting.
Moulding methods fall into three classes by degree of mechanisation. Manual moulding uses a pit or a moulding box, with ramming, pattern drawing and mould repair all done by hand; it suits single pieces, small batches and very large castings. Machine moulding compacts the sand with a jolt-squeeze or jolt-ram mechanism, together with mechanical actions for rollover, pattern drawing, core setting and closing. No-bake sand moulding uses resin and hardener to harden the sand in the box by itself, requires no drying, and gives a mould of high strength and stable dimensions. The process differences between the three are given in Chapter 2; only equipment and capacity are discussed here.
The capacity of a moulding line is determined by four parameters: moulding rate (boxes per hour), pieces per box, net weight per piece, and effective operating hours per year. Written as a formula:
Annual casting output (tonnes) ＝ moulding rate × pieces per box × net weight per piece (tonnes) × effective operating hours per year × first-pass yield
An illustrative calculation by the Institute: a moulding line running at 60 boxes per hour, with 2 pieces per box, a net weight of 8 kilogrammes per piece, 4,000 effective operating hours per year and a first-pass yield of 95%, gives hourly output of 0.96 tonnes and annual output of about 3,648 tonnes. All five inputs are assumptions used to illustrate the method. On an actual line, pieces per box are set by the process layout, net weight per piece by the product, and effective operating hours per year by order saturation and maintenance scheduling; first-pass yield is covered in Chapter 26.
Investment in a moulding line and changeover time are in tension. The higher the rate, the more complex the tooling adjustment required for a changeover: pattern plates must be changed, moulding boxes must be changed, the core-setting mechanism must be adjusted, and pick-up positions must be retaught. High-rate lines are therefore suited to long runs of a single product, while multi-product small batches are better served by manual moulding or single-box no-bake moulding. The correspondence is different in die casting: a die change is measured in hours and the thermal balance of the die must be re-established, and on the company process-page basis the economic batch size is 5,000 pieces per year or more; this threshold is likewise set jointly by die changing and die amortisation.
Core making falls into three classes by curing method. A hot box uses a heated core box to cure the resin, with high energy consumption and short core-box life. A cold box uses triethylamine gas as a catalyst to cure, giving high efficiency and good dimensional accuracy of the core, but the exhaust gas must be treated. A no-bake core hardens through resin and hardener at ambient temperature, suiting large cores and single-piece production. Core quality directly determines the dimensions and surface of the internal cavity and also the workload of the cleaning operation. The more complex the core, the more core-making stations are needed and the harder it is to match the rates of the moulding line and the core line; this is the most common bottleneck position on an automated moulding line.
Sand preparation is the supporting system of a moulding line, covering used-sand cooling, magnetic separation, crushing, reclamation and mulling. The used-sand reclamation rate determines the make-up quantity of new sand and the quantity of waste sand discharged. The historical thresholds given in the 2013 Access Conditions are a reclamation and reuse rate of no less than 60% for sodium silicate sand, no less than 90% for furan resin no-bake sand, no less than 70% for alkaline phenolic resin no-bake sand, and no less than 95% for clay-bonded sand; that document was repealed by MIIT Announcement No. 19 of 2019, and this set of values serves only as a comparison. Reclamation methods divide into mechanical reclamation and thermal reclamation; sodium silicate sand is the hardest to reclaim, which is why its reuse threshold is the lowest. The 2023 guiding opinion proposed that by 2025 the reclamation and recycling of foundry waste sand reach more than 8 million tonnes (800 万吨); this target turns reclamation capacity from a cost problem inside individual enterprises into a regional supporting-infrastructure problem.
The precision boundary of sand casting determines where it can be applied. According to a comparison in an industry encyclopaedia entry, the surface roughness of sand casting is Ra 50–12.5 micrometres, and the dimensional tolerance on a 100 mm datum is ±1.0 mm; the minimum wall thickness for cast iron is 3–4 mm (a common process range in the industry, for which this article did not obtain a citable specific source), and the full comparison of the five processes is given in Chapter 5. These three values mean: machining allowance must be left according to a tolerance of ±1.0 mm plus a layer for surface defects, and thin-walled structures below 3 mm are difficult to form reliably. Where the precision requirement of a casting exceeds this range, the process must be changed or machining added.
Lost-foam casting replaces the wooden or metal pattern with a foam pattern that remains in the mould and is vaporised by the molten metal; there is no parting line and no draft angle, and dimensional accuracy is better than with ordinary sand moulding, with the process principle given in Chapter 3. In the 2024 cluster data for Botou, annual output of full-mould castings accounts for more than 50% of the national total, which indicates that the area is the country's main producing region for lost-foam castings; there is no public breakdown of the local process mix, and this article makes no inference. The price of lost-foam casting is the one-time consumption of the foam pattern and the collection and treatment of the vaporised products.
Sand mould 3D printing combines the manufacture of pattern and mould into one step, eliminating tooling, and suits first articles of new products and small-to-medium batches. The 2023 guiding opinion of three ministries and commissions includes patternless casting and sand mould 3D printing among the directions for industrial application. Its cost structure is tied to batch size: unit cost is high and falls rapidly as the batch increases, and the point at which its cost curve intersects the mould-amortisation curve determines the batch range in which it is economical.
Equipment stock can be seen in Laizhou. In 2022 the city's 245 foundries had more than 350 major moulding installations of all kinds, including moulding lines and core shooters, or about 1.4 per enterprise (our estimate). This density indicates that most enterprises run a single line, with one line carrying all products, frequent changeovers and rates that are hard to run at maximum. There is no fixed conversion between equipment count and capacity: the same number of lines can produce different output, and the difference lies in rate utilisation and yield.
The equipment level of moulding and core making ultimately comes down to three measurable quantities: whether dimensional tolerance can be held consistently within the standard grades, whether changeover time can be compressed, and whether the used-sand reclamation rate can meet the threshold. Improving any of the three requires investment in equipment and also requires stopping production, which is the hardest stage to schedule in a foundry's technical upgrading.
Chapter 25 Cleaning, Heat Treatment and the Degree of Machining Integration
Cleaning is the whole set of operations after the casting is taken out of the mould: shakeout, removal of runners and risers, sand removal, grinding, weld repair and shot blasting. This stage has the highest dust, noise and physical workload in a casting shop, which is why GB 39726—2020 lists shakeout and cleaning as one of eight controlled processes. Table 1 of that standard sets the particulate matter limit for the shakeout and cleaning stage at 30 mg/m³ and for cupolas at 40 mg/m³; the values were cross-checked against environmental impact assessment approvals from several places, and this article did not obtain a publicly accessible link to the original text; the standard's metadata are available from the National Standards Full-Text Public System.
Cleaning equipment is graded by degree of automation. Manual cleaning uses angle grinders, cutting discs and pneumatic chisels, with scattered workstations where dust is hard to collect. Mechanical cleaning uses shot-blasting machines, classified by workpiece form into tumble-belt, hook-type, roller-conveyor and rotary-table types, and shot blasting removes sand and strengthens the surface at the same time. Robotic grinding concentrates grinding stations in an enclosed booth with a dust-extraction system, and dust capture efficiency is higher than at scattered stations. The degree of automation in cleaning directly affects two cost items: grinding hours and the running cost of the dust-extraction system.
Heat treatment of castings has three purposes: to relieve casting stress, to improve microstructure and machinability, and to meet mechanical-property requirements. Grey iron castings are usually only stress-relieved by annealing; ductile iron castings are annealed or normalised to obtain the required matrix structure; cast steel is annealed, normalised or quenched and tempered according to grade; aluminium alloy castings are solution-treated and aged. Heat-treatment furnaces are classified by operating mode as bogie-hearth, pit and continuous, and by heat source as electric, natural gas and fuel oil. The cost items of heat treatment are energy consumption and hours, and the quality problems are distortion and uneven hardness. A casting with a thick section has a large temperature difference between inside and outside during heat treatment, so residual stress is not fully released and distortion appears only during subsequent machining; the rework cost of such problems is far higher than the cost of the heat treatment itself.
Determining the machining allowance is a technical detail in integration negotiations. If the allowance is left too small, the decarburised layer and the sand-adhered layer on the casting surface cannot be machined away, and defects are exposed after machining; if it is left too large, metal charge and machining hours are both wasted, and the density of a thick section may be insufficient. The allowance must be set jointly by the dimensional tolerance of the casting, surface roughness, heat-treatment distortion and the clamping method; the drawings and process documents of both parties must agree, and where the bases differ, the drawing callout governs.
Machining follows cleaning. The relationship between a foundry and machining takes three forms.
The first is to supply the blank only. Castings are cleaned, inspected and shipped, and machining is done by the customer. In this form the foundry's responsibility boundary is the dimensions and material of the blank, and the machining allowance is left according to a standard or drawing.
The second is the blank plus rough machining. The foundry provides rough machining of datum surfaces and main bores, and the customer does the finish machining. The value of rough machining is that it absorbs within the foundry the allowance uncertainty created by casting tolerances, reducing one clamping operation for the customer.
The third is delivery of the finished part. Casting, cleaning, heat treatment, machining, surface treatment and assembly are all completed in-house. In this form, direct materials account for 36.72%–50.65% of operating cost, and labour and equipment depreciation are added to cost. The disclosed basis of the three precision-component enterprises belongs to the third form, which is why Chapter 22 warns that it cannot be treated as a pure casting cost structure.
Greater integration brings three benefits. Transport cost falls, especially for castings with a high unit weight; second-clamping error is eliminated and the datum is unified; and delivery is controlled by one party, with a clear boundary of responsibility. The price is higher capital expenditure and reduced flexibility: machining equipment is specified according to the largest casting size, machine adjustment takes longer when products are switched, and the unit machining cost of small-batch orders is higher than at a specialist machining plant.
The unit weight of a casting determines its transport radius. The capabilities page of a sand-casting enterprise gives a range of 0.3 kilogrammes to 20 tonnes, and a unit range of 60 to 15,000 kilogrammes for its resin-sand line. Transporting a casting of the 20-tonne class requires special vehicles, freight is charged as out-of-gauge cargo, and the sales radius is constrained by transport conditions. In Botou, large machine-tool castings account for 25% of the national total, pump and valve castings for 35%, and castings for automotive panel dies for 16%; the high share of heavy parts corroborates nearby supply under transport constraints.
Weld repair is part of the cleaning operation and is also a frequent point of quality dispute. Gas porosity, shrinkage porosity and cracks in a casting are exposed after cleaning, and weld repair can rescue some rejects. Weld repair requires process qualification: filler metal, preheat temperature, interpass temperature, post-weld heat treatment and inspection method must all be fixed in the process documents. The customer's acceptance is written into the technical agreement; some customers allow weld repair but not on load-bearing locations, and some prohibit it entirely. If the technical agreement does not state this, each party acts on its own understanding at delivery, and disputes arise from this.
The degree of integration of cleaning, heat treatment and machining determines a foundry's position in the supply chain: a plant that supplies blanks only quotes by the tonne, while a plant that supplies finished parts quotes by the piece; what lies between the two quotations is operations, and also responsibility. If a buyer does not distinguish between these two forms when comparing prices, it is not comparing the same thing.
Chapter 26 Yield: Reject Rate, Rework and Batch Consistency
Casting rejects fall into two classes by where they are found. Internal rejects are found during in-house inspection, and what is lost is the metal charge already committed, melting energy, moulding hours and sand; external rejects are found during the customer's machining or assembly, and the loss also includes freight, the customer's line-stop loss and reputation. The cost difference between the two classes can reach several times, which is the economic basis for investment in inspection.
The form of a reject is related to the process. In sand casting the common defects are scab, shrinkage porosity, mould shift, core displacement and misrun; in die casting they are gas porosity, shrinkage cavity, cold shut, die sticking and distortion; in investment casting they are out-of-tolerance dimensions, surface defects, shell cracking and foreign matter in internal cavities. The forms differ, the causes differ, and so do the means of control. Scab comes from insufficient moulding-sand strength or too low a pouring temperature, shrinkage porosity from insufficient feeding, and gas porosity from poor venting of the mould cavity or a high gas content in the molten metal.
The reject rate has two bases: by weight and by piece count. On a weight basis, scrapping a 2-tonne machine-tool casting and scrapping a 2-kilogramme die casting differ by a thousand times in their contribution to the denominator; on a piece-count basis, the two carry the same weight. External quotations and internal assessment must use the same basis; otherwise the same batch of products has two reject rates, both correct and neither comparable with the other.
There is no public statistic for the reject rate of the national casting industry, and this article gives no figure. What can be given is the way the reject rate amplifies cost. Continuing the illustrative calculation of Chapter 23: with a process yield of 70% and melting electricity consumption of 600 kWh per tonne of molten metal, when the reject rate rises from 5% to 10%, the poured molten metal required to deliver the same 1 tonne of acceptable castings rises from 1.504 tonnes to 1.59 tonnes, and melting electricity consumption rises from 902 kWh to 952 kWh, an increase of 5.6%. The make-up quantity of metal charge is unchanged, since rejected parts are remelted in-house; what increases is melting energy, moulding hours and sand consumption. If the rejects are found at the customer, round-trip freight and the customer's disposal cost must be added.
Batch consistency is harder to quantify than the reject rate. Compositional fluctuation of the same grade between heats, dimensional drift of the same mould between shots, and hardness variation of the same batch of moulds between time periods all belong to batch consistency. What the customer experiences on the assembly line is often that some parts in a batch fit and others do not, while out-of-tolerance individual pieces appear only in the sampling inspection report.
Ductile iron has a special constraint on batch consistency: after inoculation the molten iron degrades with time, and the nodularity and graphite morphology deteriorate accordingly, so the time from inoculation to pouring must be held within the window specified in the process. The length of the window depends on the molten iron temperature, the type of inoculant and the treatment method. The implication for production organisation is that the rates of melting, moulding and pouring must be matched, and waiting at any one of the three consumes the inoculation effect. This is one reason duplex melting is adopted in large-batch production of ductile iron: the ability to hold and adjust temperature leaves room for matching rates.
Rework includes weld repair, straightening, re-heat treatment and re-blasting. Judging a reworked part requires standards: under what conditions it is re-inspected after rework, whether passing re-inspection is equivalent to the original product, and whether a limit is set on the number of rework cycles. Only when these clauses are written into the technical agreement is there a basis at delivery. Rework costs less than scrapping, but a reworked part may have lower fatigue performance and density than the original, which matters especially for pressure-containing and safety parts.
The carrier of batch traceability is the heat number. The heat number links chemical composition, melting records, pouring time, heat-treatment curves and inspection results together. GB/T 40805—2021, Steel Castings — General Technical Conditions for Delivery Acceptance, lists the quality certificate as a required item of delivery acceptance, and the EN 10204 3.1 certificate commonly required for export orders requires traceability to the heat number. A workshop without heat-number management can only recall an entire batch when a quality problem arises, and the cost is counted by batch.
The statistical period of the reject rate also affects judgement. A shift basis is used for process control, a heat basis for traceability, and a monthly basis for assessment. The three periods give different values: shift-based figures fluctuate most, and monthly figures have the fluctuation averaged out. Judging the effect of a single process adjustment from monthly data usually involves an insufficient sample size, and the effect of the adjustment is masked by normal shift-to-shift variation. A reject rate reported externally without a statement of its period cannot be compared as it stands.
Process control comes down to three records: pre-furnace chemical composition, moulding-sand properties (moisture, compactability, strength, permeability), and pouring temperature and time. The three records must be kept every shift; the data volume is small, but they cover the great majority of reject causes. Applying statistical process control to these data moves judgement from after the event into the process, provided that the repeatability and reproducibility of the measurement system are adequate. If records are not kept, or are kept but not analysed, no amount of inspection equipment can do more than sort after the event.
Improving yield requires visibility first: without heat numbers and process records, the reject rate is only an outcome number, and its causes cannot be traced. The order of investment in inspection should be process records first, then non-destructive testing, and automation last.
Chapter 27 Layering of Inspection Methods and Equipment Levels
Casting inspection divides into five layers by object: the molten metal before the furnace, the moulding sand, the dimensions of the casting, the interior of the casting, and the surface and mechanical properties of the casting. Each layer has different equipment, standards and methods of judgement, and the layering of equipment levels unfolds along these five layers.
The two basic contents of pre-furnace inspection are chemical composition and temperature. Chemical composition is measured with a direct-reading spectrometer or a carbon-sulphur analyser, with results in tens of seconds after sampling, and is used to adjust the charge ratio; temperature is measured with a thermocouple or an infrared thermometer, with tapping temperature and pouring temperature controlled separately. Ductile iron also requires pre-furnace determination of nodularity, commonly by thermal analysis, judging nodularisation and inoculation effects from the features of the solidification curve. The 2013 Access Conditions required pre-furnace chemical composition analysis and molten metal temperature measurement equipment, which was the minimum line of more than ten years ago. In industry practice, the direct-reading spectrometer is used for rapid composition analysis during melting and before tapping, the coordinate measuring machine for comparison with the digital model, and X-rays or industrial CT for detecting internal shrinkage cavities, gas porosity and cracks; these three types of equipment form the backbone of inspection capability.
Moulding-sand testing covers moisture, compactability, green compressive strength, permeability and effective clay content. These indicators determine the hardness of the mould, its drawability and its resistance to scab; the equipment is a sand testing machine, which is not a large investment but must be used every shift with records kept. Testing of the used-sand reclamation system also covers loss on ignition and mud content.
Dimensional inspection starts with callipers, height gauges and surface plates and extends to coordinate measuring machines and 3D scanning. CMM results can be compared directly with the computer-aided design model, producing a deviation colour map, and are suited to first articles and tool verification; online inspection in volume production relies more on dedicated gauges and vision measurement. Dimensional judgement is based on GB/T 42124.3—2025, which is a modified adoption of ISO 8062-3:2023, was issued on 30 May 2025 and takes effect on 1 December 2025, replacing GB/T 6414—2017. Where a purchase contract still cites the old standard number, the versions of the drawing and the acceptance clauses need to be checked.
Internal defects are examined by ultrasonic testing and radiographic testing. Ultrasonic testing follows GB/T 7233.1—2023, applicable to general-purpose steel castings that have undergone grain-refining heat treatment and have a wall thickness not exceeding 600 mm; it does not apply to austenitic steels or austenitic-ferritic duplex stainless steels, and martensitic stainless steels may be treated by reference to it; high-pressure steel castings follow GB/T 7233.2—2023. Radiographic testing follows GB/T 5677—2018, Castings — Radiographic Testing. Ultrasonic testing is more sensitive to planar defects (cracks, lack of fusion) and radiography to volumetric defects (gas porosity, shrinkage cavities); the two are complementary in detection capability and cannot replace each other.
Surface and near-surface defects are examined by magnetic particle testing and penetrant testing. GB/T 9444—2019, Steel and Iron Castings — Magnetic Particle Testing applies to surface and near-surface defects in ferromagnetic materials, and GB/T 9443—2019, Steel and Iron Castings — Penetrant Testing applies to surface-breaking defects in non-porous materials; stainless steel and aluminium alloy castings use penetrant testing.
Mechanical-property testing uses separately cast or integrally cast test blocks for tensile, hardness and impact tests. The sampling position, number and test method for the blocks are specified by the material standard, and the acceptance indicators are set by grade. The mechanical properties of a test block represent the block; the properties of the casting body at thick sections and critical locations must be verified separately, and on thick-walled parts the difference between the two cannot be ignored.
Combining the inspection capabilities and equipment described above yields three levels. The first level is manual moulding with a medium-frequency furnace, where inspection relies on visual examination, callipers and chemical analysis sent out to a laboratory; it can meet the dimensional and appearance requirements of general machinery parts, but cannot supply traceable chemical composition or non-destructive testing reports. The second level is a moulding line with a direct-reading spectrometer and shot-blast cleaning, with pre-furnace rapid analysis and hardness and tensile testing capability, and dedicated gauges for dimensional inspection; it can take on volume parts for automotive and general machinery and issue a quality certificate under GB/T 40805—2021. The third level is an automated line with online inspection, CMM and ultrasonic and radiographic testing, with full heat-number traceability and statistical process control; it can take on pressure-containing, export and safety parts and issue an inspection certificate under EN 10204 3.1. The difference between the three levels lies in the completeness of the quality evidence chain; the equipment list is only its outward manifestation.
Equipment counts give a rough impression of the layering. Laizhou's 245 foundries hold more than 500 medium-frequency electric furnaces (sets) and more than 350 moulding lines and core shooters of all kinds (sets); averaged over the number of enterprises, that is about 2 furnaces and 1.4 moulding installations each (our estimate). Within the same cluster there are 30 enterprises above designated size and 215 below it, and the distribution of equipment is uneven. The average indicates only the equipment density of the cluster, not the equipment level of an individual plant.
Inspection cost items include equipment depreciation, consumables, labour and downtime. The proportion of ultrasonic and radiographic testing is specified in the technical agreement; the higher the proportion, the higher the cost, and when the proportion falls from 100% to sampling inspection, the extent of the cost reduction depends on the batch size. Stating the inspection method, standard number, inspection proportion and acceptance level clearly in a quotation is better for controlling cost, and better for avoiding delivery disputes, than the general phrase "flaw detection passed".
The layering of inspection methods determines which kind of supply chain a foundry can enter: plants able to issue certificates enter the systems of original equipment manufacturers, while plants unable to do so remain in second-tier supply and the repair market. Building this capability takes longer than a moulding line, because its core is people and records, of which equipment is only one part.
Part Five: Procurement and Verification
Chapter 28 How to Verify a Foundry (Ten Things to Look at on Site)
The appearance of a casting can be copied from a sample; the process records cannot. A buyer who visits the plant once, following the sequence from material receipt to finished-goods shipment, can look at ten items, each of which has a pass line that can be judged on the spot and a question that must be pursued. The checklist below can be taken straight onto the shop floor.
- Melting and furnace-front composition analysis
What to look at: whether there is a direct-reading spectrometer beside the furnace, whether the spectrometer is switched on, and whether there are analysis records for the current shift. The common practice in the industry is to use a direct-reading spectrometer for rapid composition analysis during melting and before tapping (notes on testing configuration).
Pass line: every heat in this batch has an analysis record, and the record carries the heat number, sampling time, element contents and the person who made the judgement; the spectrometer has reference standards and records of routine checks or calibration.
Follow-up: how the composition is adjusted when it is out of specification, and whether it is re-tested after adjustment; whether a specimen is separately taken from the casting body for re-testing. More than a decade ago, the industry access threshold required furnace-front equipment for chemical composition analysis and molten-metal temperature measurement (reprint of the 2013 access conditions); a plant that still cannot meet this baseline today has essentially no process control.
- Heat number and batch traceability
What to look at: whether every heat has a unique heat number, and whether the heat number can link the pouring record, the heat-treatment charge, the finished-product marking, the dispatch note and the quality certificate. GB/T 40805-2021 includes the quality certificate among the delivery acceptance items for steel castings (standard information).
Pass line: a finished casting picked at random can, on the spot, be traced to its heat number, composition, pouring date and heat-treatment curve.
Follow-up: how heats are separated when one ladle of molten iron is poured into moulds for different grades; whether subcontracted parts are given a new heat number when they return to the plant. Traceability breaks wherever the heat number breaks.
- Sand preparation and moulding sand testing
What to look at: whether there is used-sand processing equipment and a sand mixer, and whether the moulding sand is tested for compactability, wet compressive strength and moisture content.
Pass line: there are moulding-sand test records for the day, the sand mixer has batch formula records, and the ratio of new to used sand can be checked.
Follow-up: where the waste sand goes. The 2013 access threshold set used-sand reuse rates of not less than 95% for clay-bonded sand and not less than 90% for furan resin no-bake sand (historical clause); a 2023 document from three ministries proposed that by 2025 the recycling of foundry waste sand should reach more than 8 million tonnes (800 万吨) (guiding opinion). Where waste sand goes is both an environmental risk and a cost item.
- Moulding line and mould quality
What to look at: whether the moulding line is running, whether moulding is done by hand or by machine, and whether mould hardness is spot-checked.
Pass line: the moulding line is in production during the shift, there are records of mould hardness spot checks, and pattern plates have numbers and fixed storage locations.
Follow-up: how many pattern changes are made in a day. In a plant with frequent pattern changes, the risk to dimensional consistency is high, and wear on pattern location features and fasteners shows up directly as wall-thickness deviation.
- Pouring
What to look at: how the pouring temperature is measured, whether the pouring time is recorded, and whether ladles are preheated.
Pass line: every ladle has a temperature record, the pouring operation has a process card, and the card states the grade, the temperature range and the pouring sequence.
Follow-up: where the records of spheroidisation and inoculation are kept (for ductile iron parts); when several moulds are poured from the same ladle, how the temperature of the later moulds is controlled. If there are no records at the pouring stage, the data from the two preceding operations cannot be matched to the finished product.
- Fettling, heat treatment and weld repair
What to look at: whether the heat-treatment furnace has a temperature recorder and a furnace curve for each load, and whether weld repairs of castings have approval and records.
Pass line: heat-treatment curves can be matched to the furnace batch numbers; weld repairs have records of location, number, welder and inspection.
Follow-up: whether weld-repaired parts are marked separately and whether the repairs are noted on the quality certificate. A weld repair is ground flush and matches the parent metal in appearance, which makes it the hardest class of defect to find in incoming inspection.
- Testing capability
What to look at: whether there is a universal testing machine, a hardness tester and a metallographic microscope; whether non-destructive testing is done in-house or subcontracted.
Pass line: the testing machine is within its calibration validity period, and recent test specimens and fracture surfaces are kept on file; for subcontracted testing, the subcontractor's qualification and the original report can be produced, and the report states the method, standard number, level and proportion.
Follow-up: whether the mechanical-property specimens are integrally cast test blocks or separately cast test blocks, and from which part of the casting body they are taken. The sampling location determines whether the report can represent the casting body. A report dated earlier than the start of production of this batch cannot serve as the basis for this batch; for subcontracted testing, it is also necessary to see who took the samples.
- Where rejects and non-conforming parts go
What to look at: the reject area, the rework area and the scrap register. Three scrapped parts are picked at random, and the corresponding records must be found on the spot.
Pass line: there are monthly scrap-rate statistics, broken down by defect type (shrinkage cavity, gas porosity, sand inclusion, crack, dimension); the scrapped parts reconcile with the register.
Follow-up: whether the denominator of the scrap rate is the number of parts poured or the number of conforming parts; whether scrapped parts are remelted or sold outside. If the proportion of remelt is not recorded, the source of composition drift cannot be traced.
- Environmental and safety hardware
What to look at: whether dust extraction is running at the melting, moulding, pouring, shakeout and fettling operations; whether the melting furnace has emergency discharge and emergency storage facilities; whether there is standing water at the furnace bottom, in the furnace pit and in the pouring pit. The latter two items, together with the siting of places where people gather, are listed as major accident hazards in the Standard for Determining Major Accident Hazards in Industrial and Trade Enterprises (《工贸企业重大事故隐患判定标准》) (Order No. 10 of the Ministry of Emergency Management).
Pass line: dust-extraction equipment is running and there are operating records; the furnace pit is dry; and within the floor area of the molten-metal crane bay and the pouring bay there are no meeting rooms, rest rooms or similar places where people gather.
Follow-up: which operations were suspended during the most recent heavy-pollution weather alert. Grade C enterprises in the performance grading system suspend the melting operation during an orange alert, and Grade D enterprises suspend all emission-generating operations (alert measures). Shutdown records say more about an enterprise's actual local standing than any self-description.
- Order structure and subcontracting list
What to look at: what parts are in production in the shop, how large the batches are, and which operations are subcontracted.
Pass line: the plant can produce a list of subcontracted operations with the names of the subcontractors, and melting and pouring are not subcontracted.
Follow-up: whether subcontracted parts are re-inspected when they return, and whether a change of subcontractor is notified to the customer. The subcontracting list defines the real boundary of this plant's capability; subcontracting machining and heat treatment means that both delivery time and quality gain an extra link that cannot be controlled.
Having looked at the ten items, a judgement can be made: whether this plant's processes have records, or only explanations. Records can be checked by sampling; explanations can only be trusted. What the buyer should take away is a set of records that can be checked; verbal promises cannot be checked.
Chapter 29 Standards and Certificates: Material Grades, Mechanical Properties and Non-Destructive Testing
The three things most easily left vague in a casting contract are the material grade, the acceptance standard and the testing clauses. If all three are written clearly, disputes become far fewer; if any one is vague, the judgement after delivery has to be renegotiated.
- Material grade
Writing only "cast iron", "ductile iron" or "cast steel" in the contract amounts to not specifying the material at all. A grade encodes the material category and the minimum performance requirements in its designation, and the specific code and values are governed by the material standard adopted. The correct form is to state all three items — grade, governing standard and heat-treatment condition — and to keep the drawing, the contract and the quality certificate consistent in all three. When the customer's drawing cites a foreign grade, a cross-reference note and a statement of equivalence should be attached, signed by the technical managers of both parties. If the supplier converts grades on its own, the risk remains on the buyer's side.
- Delivery acceptance standard
Steel castings have a dedicated general standard for delivery acceptance: GB/T 40805-2021, Steel castings — General technical conditions for delivery acceptance, published on 11 October 2021, which is a modified adoption of ISO 4990:2015 (standard information). The standard covers ordering requirements, technical requirements, inspection and test methods, inspection rules, marking, the quality certificate, packaging and transport. The current version of ISO 4990 is ISO 4990:2023, published on 9 November 2023 (ISO 4990:2023). The standard applies to steel castings; the delivery acceptance clauses for iron castings must be agreed item by item in the contract, and a bare "acceptance according to the national standard" cannot be executed. The official version issued by the publishing body governs the text of a standard; photocopies and web summaries are for reference only. When a contract cites a standard number, it should also state the standard's title and year.
- How to specify mechanical properties
The acceptance items for mechanical properties usually include tensile strength, yield strength, elongation and hardness, and the specific items and values are set by the material standard corresponding to the grade. The contract should not fill in values on its own; it should cite the grade and the standard, and then state the specimen form. Casting test specimens for mechanical properties come in three forms: separately cast test blocks, integrally cast test blocks and specimens cut from the casting body; the cooling conditions of the three differ, and the values measured cannot be compared directly. The contract must state which form is used, from which part it is taken, and how many specimens are taken per batch. A report that does not state the specimen form and location is treated as incomplete documentation. The report should also state the specimen number, the sampling date and the test equipment number; where several reports are issued for the same batch, the numbers must cross-reference each other, and if the specimen numbers cannot be matched to the heat batch numbers, the reports cannot be assigned to a batch.
- Revision of the dimensional tolerance standard
Dimensional tolerance is the most frequent item in casting disputes, and it has just changed edition. GB/T 6414-2017 now has the status of withdrawn, with a withdrawal date of 1 December 2025 (standard status); it is replaced by GB/T 42124.3-2025, published on 30 May 2025 and implemented on 1 December 2025, which is a modified adoption of ISO 8062-3:2023 (new standard, adoption relationship). If technical agreements, drawing annotations, gauges and CMM programs are still executed to the old edition, inspection reports issued after the change of edition will cite a withdrawn standard. The buyer must state in the agreement which edition is to be followed, and agree on the judgement rules for the transition period.
- Non-destructive testing
The division of work among the four methods follows the scope of application of the standards themselves: ultrasonic and radiographic testing detect internal defects, while magnetic particle and penetrant testing detect surface and near-surface defects.
Ultrasonic testing is based on GB/T 7233.1-2023, Ultrasonic testing of steel castings for general purposes, published on 17 March 2023 and implemented on 1 October 2023, replacing the 2009 edition (standard information). The standard applies to steel castings for general purposes that have undergone grain-refining heat treatment and have a wall thickness of not more than 600 mm; it does not apply to austenitic steels or austenitic-ferritic duplex stainless steels, and martensitic stainless steels may be tested with reference to it (scope of application). For high-pressure steel castings, GB/T 7233.2-2023 is used instead, published on 23 May 2023 and implemented on 1 December 2023.
Radiographic testing is based on GB/T 5677-2018, Radiographic testing of castings, published on 17 September 2018 and implemented on 1 April 2019 (standard information). Magnetic particle testing is based on GB/T 9444-2019, Magnetic particle testing of steel and iron castings, and penetrant testing on GB/T 9443-2019, Penetrant testing of steel and iron castings; both were published on 30 August 2019 and implemented on 1 March 2020 (magnetic particle, penetrant).
Contract clauses must cover all five items: method, standard number, testing level or acceptance level, testing proportion and testing timing. Writing only the four words "NDT passed" makes the clause unenforceable in a dispute: who tests, to what standard, how much is tested, and whether testing is done on the raw casting or after rough machining are all left unspecified. The boundary of the material must also be written into the clause: austenitic steels and duplex stainless steels are not covered by GB/T 7233.1-2023, and copying the standard number blindly invites the rebuttal that the standard does not apply. The testing proportion and acceptance level directly determine testing cost and also determine the probability of missed defects; the proportion is written in the contract and the acceptance level in the report, and the two cannot be stated in only one place.
- Certificates
Domestic orders are accepted on delivery according to GB/T 40805-2021, and the supplier provides a quality certificate with the goods. Export orders and orders from foreign companies often require a type 3.1 inspection certificate (EN 10204 or ISO 10474), which is issued by the manufacturer's independent inspection representative and can be traced by heat batch number (example certificate). When the certificate is received, three things are checked: the heat batch numbers on the certificate match the markings on the goods; the inspection items on the certificate cover the contract requirements; and the signatory and the date of issue are complete. A third-party test report is responsible only for the sample submitted, and who takes the sample determines the evidential weight of the report. The date of issue of the certificate must also be checked: a report is responsible only for the batch submitted, and a report dated earlier than the melting date of this batch has no correspondence with this batch.
- Re-inspection on arrival
There are three approaches to re-inspection on arrival: document review, sampling and submission for testing, and full inspection. Document review checks the heat batch numbers, the items on the certificate and the test reports; sampling and submission for testing sends samples to a third-party laboratory designated by the buyer; full inspection is used only for critical parts, since samples subjected to destructive testing cannot return to the finished product. The three approaches differ greatly in cost, and allocating them according to the consequences of failure is more sustainable than inspecting everything.
- Archiving of acceptance records
Acceptance records should be kept at least until the warranty period of the complete machine ends. The archive includes the heat batch number list, the quality certificate, the non-destructive testing reports, the mechanical-property reports, the heat-treatment curves and concession acceptance forms. Export orders also require the correspondence between inspection certificates and customs declaration documents to be kept. When records are incomplete, tracing a customer complaint can only rely on memory.
- What a certificate can prove
From 3 June 2019, the Foundry Industry Access Conditions (《铸造行业准入条件》) and the three batches of announced lists were abolished (abolition notice); the current basis for industry self-regulation is the China Foundry Association group standard T/CFA 0310021, Specification conditions for foundry enterprises, and group standards are not mandatory (2023 MIIT document). A certificate can prove the system and the sample submitted for testing; it cannot prove the batch that arrives. What the buyer must do is bind the certificate to the heat batch numbers and the accompanying records and check them together.
Chapter 30 Verification Workflow for First Articles, Trial Production and Small Batches
Casting verification has three stages: sample parts, first articles and small batches. Each stage answers one question: whether the part can be made at all, what it looks like when made under production conditions, and whether it can be kept consistent when made repeatedly. When the three stages are collapsed into one, the usual outcome is a qualified sample and a non-conforming production run, with no accountable checkpoint left in the contract.
- Sample stage: verifying manufacturability
Sample parts can be made with simple tooling, hand moulds or rapid prototyping, the purpose being to expose in advance the features on the drawing that cannot be cast: wall thickness below the process minimum, fillets that are too small, insufficient draft angle, insufficient machining allowance, and cores that cannot be placed or cannot be removed. For sand casting, the minimum wall thickness is more than 3 mm for aluminium parts, 3 to 4 mm for iron castings and more than 5 mm for steel castings; investment casting can usually achieve 0.7 mm. These figures are common process ranges in the industry, and this article has not obtained a specific source that can be cited; for any particular feature, size and shape must also be considered.
The sample stage should leave behind a list of problems and suggestions for drawing changes; a conclusion that merely states "sample qualified" is useless. Nor can a qualified sample serve as the basis for production, because samples are usually made by the best craftsmen under the most forgiving conditions. The number of samples and the verification items should be set out in the technical agreement; a confirmation that involves only one part and only a visual check cannot cover assembly and subsequent machining.
- First-article stage: verifying production conditions
The first article must use production tooling, production process documents, production operators and production inspection; if any one of the four is missing, the first-article report loses its meaning. The first-article inspection items must cover full dimensional inspection (to the tolerance standard and grade agreed in the technical agreement, the current edition being GB/T 42124.3-2025, see Chapter 29), material (furnace-front composition plus a specimen from the casting body), mechanical properties, hardness, metallography (where agreed in the contract), non-destructive testing (to the method, level and proportion in the contract), weight, marking and traceability.
The first-article report must include a measurement point drawing. A report without a measurement point drawing contains only conclusions and no process, and on review it is impossible to tell which surface was measured and against which datum. The report should also state the number of the measuring equipment and its calibration validity period. The measuring instruments and gauges used for first-article measurement must be the same as those used for production inspection: a first article measured on a CMM and production inspected with a calliper detect different deviations, and a qualified first article does not guarantee that production inspection can detect them. Dedicated gauges should be accepted together with the first article.
Mould acceptance is carried out together with first-article acceptance. The mould's material, parting line, gating system, venting and cooling, mould number and design mould life must all be written into the acceptance sheet. Ownership of the mould and its storage location are written into the contract, so that the mould cannot become impossible to retrieve when production moves to another plant. The first article should be made by a regular shift on the production line, and operators and equipment must not be hand-picked; a first article made specially by the sample group verifies the capability of the sample group. The handling of a non-conforming first article should also be stated in the agreement: repair, remaking, or cutting a new mould.
- Small-batch stage: verifying stability
Small-batch production should run over several consecutive batches. The measured dimensions and defect types of each batch must be recorded; looking only at pass/fail judgement will not reveal drift in the distribution. For each batch, record the heat number, the moulding shift, the pouring temperature, the heat-treatment charge and the operating team; only by viewing these variables together with the measured dimensions and defect types can the source of variation be identified.
Three kinds of problem are most easily exposed at the small-batch stage: dimensional drift after a shift change, composition variation after a material change, and wall-thickness change after mould repair. All three are process changes, and a notification obligation should be agreed in the contract. The size of a small batch is determined by the order and by process stability; the contract can specify a minimum number of verification batches and a quantity per batch, so that the supplier cannot respond with a single conforming part.
- Change control
The following changes should be notified in writing and re-verified: a change in the proportions of pig iron, scrap steel and remelt; a change in the moulding sand and binder system; mould repair or replacement; a change of subcontractor for a critical operation; and a change in heat-treatment parameters. Mould repair is the item most easily overlooked, because appearance, marking and packaging do not change and only dimensions move. Change records must leave a trail: who proposed the change, when it took effect, which batches were affected, and the conclusion of re-verification. A change with no trail cannot be used in a dispute to prove that the supplier has fulfilled its notification obligation.
- Sampling and judgement rules
The sampling plan must state the lot size, the sample size, the acceptance and rejection numbers and the classification of non-conformities. Casting non-conformities are usually classified in three levels: critical defects affect safety or function, major defects affect assembly or service life, and minor defects affect only appearance. Using the same acceptance number for all three levels amounts to treating critical defects as appearance issues. The classification of non-conformities should be listed item by item in an annex to the contract, and the judgement criteria confirmed by both parties at the first-article stage.
Acceptance should also specify the re-testing rules for disputes: where re-test samples are taken, which body performs the re-test, and who bears the cost. If none of the three is written down, re-testing turns into a second round of negotiation.
- The threshold for transfer to mass production
Before transfer to mass production, the following should be completed: dimensional distribution data over several consecutive batches, defect-rate trends, packaging and logistics validation, and a complete document package (drawing version, technical agreement, first-article report, material certificate, test reports, heat-treatment records). The threshold should be set on data; the distribution and trend over several consecutive batches say more than the single first article.
The document package should be submitted with the first delivery, without waiting for the customer to ask. Its value becomes apparent only when a dispute arises: if the heat batch numbers, heat-treatment curves and test reports match each other, there is a basis for allocating responsibility; if they do not, the two parties can only fall back on sectioning the physical part, which is costly and may not yield a single unambiguous conclusion.
- Concession acceptance
When a first article or a small batch is non-conforming, the supplier often proposes concession acceptance. Concession acceptance must be in writing, state the deviating item, the amount of deviation, the batches covered and the period of validity, and be signed by the buyer's technical manager. A concession is valid only for the batches named; it cannot be carried over to subsequent batches. Treating a single concession as a long-term standard deprives later batch disputes of any criterion.
- Allocation of costs
Sample costs, trial-production costs and mould costs should be written separately in the contract. Mould cost is usually priced separately, with ownership transferring according to the payment terms; sample cost is charged per piece and trial-production cost per batch. Where cost allocation is unclear, renegotiation in the middle of trial production is not uncommon. Payment for the mould should be tied to two milestones, mould acceptance and first-article approval, and should not be paid in a single lump sum.
Chapter 31 Forms of Common Quality Disputes
Casting quality disputes take a fixed set of forms. Each has its own causes, its own method of proof and its own contractual preventive measure; discussing them all together yields only useless conclusions of the "strengthen communication" kind.
- Dimensional deviation
There are three causes: the drawing datum and the casting datum do not coincide; the tolerance grade is not agreed, so the supplier works to the coarsest grade; and the standard has changed edition, GB/T 6414-2017 having been withdrawn on 1 December 2025 and replaced by GB/T 42124.3-2025 (standard status, new standard, see Chapter 29 for details).
Proof relies on CMM reports, measurement point drawings and the sealed first-article sample. Without a sealed sample and a measurement point drawing, the two parties cannot even agree on which feature each has measured.
The preventive wording is: the technical agreement states the tolerance standard, grade, datum and measurement method, and agrees which report governs in the event of a dispute.
Dimensional disputes also often arise over machining allowance. Casting tolerance and machining allowance are two separate requirements; the drawing marks only finished dimensions, and the rough-casting dimensions are derived by the supplier from the allowance. If the allowance is insufficient, unmachined as-cast surface remains after machining; if it is too large, machining time rises. The technical agreement should attach a rough-casting drawing with the allowance marked separately, rather than leaving the supplier to derive it.
- Material not conforming
The most common cause is that the contract says only "cast iron" or "ductile iron" without stating a grade. The supplier delivers the lowest grade that satisfies the composition: the composition is within specification but the performance is insufficient. Heat treatment and cooling conditions stand between composition and properties, and a composition that alone is within specification does not imply that the mechanical properties are acceptable.
The second cause is specimen location. Specimens taken from the runner, from separately cast test blocks or from integrally cast test blocks have cooling conditions different from those of the casting body, and the figures in the report cannot represent the casting.
Proof relies on re-testing of specimens cut from the casting body and on third-party testing. The preventive wording is: the contract states the full grade, the governing standard, the specimen location and the heat-treatment condition. When grades do not match, compare the standard edition first, then the grade designation, and only then the performance values; reversing the order will misjudge a difference of edition as material falsification.
- Internal defects
Shrinkage cavities, gas porosity and slag inclusions are invisible on the raw casting and appear only after rough machining. The dispute usually centres on whether the defects fall within the permitted range; there is seldom disagreement about whether they exist.
There are three causes: the testing method and level are not agreed; the testing timing is not agreed, and whether testing is done on the raw casting or after rough machining makes a large difference; and the rules for judging defects within the machining allowance are missing.
Proof relies on defect atlases, sectioning and re-testing. The preventive wording is: the contract states the method, standard number, level, proportion and timing, together with the rules for handling defects exposed after machining.
Whether a defect is permitted also depends on where the casting is used and the direction of loading. The same shrinkage cavity may be acceptable on a non-load-bearing surface and unacceptable on a load-bearing surface. If the contract says only that shrinkage cavities are not permitted, the judgement cannot be applied; grading the judgement by location is what makes it enforceable.
- Weld repair
Weld repair is a routine rework method in the foundry industry, and the dispute lies in disclosure. A weld repair is ground flush so that its appearance matches the parent metal; if it is not noted on the quality certificate, the buyer has no way of knowing.
The preventive wording is: the contract specifies the locations where weld repair is permitted, the maximum number, the approval process, the method of marking and the recording requirements for the certificate. For critical load-bearing parts, weld repair may be prohibited by agreement. Weld repair records must identify the specific part number; a note that a batch contains several repairs gives no way of telling which parts were repaired.
- Weight deviation
When pricing is by weight, weight deviation translates directly into money; when pricing is by piece, weight deviation affects machining allowance and tool life. The direction of the dispute is opposite under the two pricing methods, so the contract must first state the pricing method clearly.
The preventive wording is: agree the weight tolerance band and the machining allowance, and confirm them by measurement at the first article. The method of measuring weight tolerance must also be agreed: weighing individual pieces and averaging a batch may lead to different conclusions, so the number of pieces sampled and the judgement rules should be stated together.
- Mechanical properties and heat treatment
There are three causes: heat-treatment batches are mixed; the test specimens heat-treated with the load differ in condition from the casting body; and the number and location of specimens are insufficient to represent the whole batch.
Proof relies on heat-treatment curves, the correspondence of furnace batch numbers and specimen numbers. The preventive wording is: agree the number and location of test specimens heat-treated with the load, and require heat-treatment records to accompany the goods.
- Batch consistency
The first article passes, and production drifts. The triggers are shift changes, material changes, mould repair and subcontracting changes.
Proof requires the measured values of the first batch and the delivered batch to be compared side by side; a certificate for a single batch cannot demonstrate drift. The preventive wording is: write the obligation to notify process changes, and re-verification clauses, into the contract. If this clause is well enforced, the six preceding categories of dispute become much less frequent. Batch-consistency disputes often appear after a working relationship has become stable, and the passing records of earlier batches may actually weaken the buyer's evidence.
- Delivery, moulds and documents
If mould ownership is unclear, the mould cannot be retrieved when production moves; if the quality certificate does not match the heat batch numbers on the goods, the problem is discovered only at export declaration or customer audit; if subcontracted parts are not re-inspected on return, responsibility falls on the last operation.
The preventive wording is: the contract states mould ownership, number, storage location and transfer conditions; documents accompany the goods; and the subcontracting list is filed.
- Allocation of responsibility for surface treatment and coating
Before a casting enters the coating line, sand inclusions, gas porosity and slag inclusions may already be present on its surface. If blistering or detachment occurs after coating, the responsibility may lie with the foundry or with pre-treatment. The order of proof is: first carry out a coating adhesion test, then section the blistered area to see whether the defect originates in the substrate or at the coating interface. The contract must state clearly the acceptance standard for surface treatment and the responsibility for pre-treatment; for quality problems that cross operations, allocating responsibility matters more than the testing itself.
- Drawing versions and design changes
After a drawing revision, the mould, gauges, CMM programs and inspection cards must be updated at the same time. The breakpoint commonly seen in actual production is this: the drawing reaches purchasing, while the mould is still being repaired to the old edition and the inspection card is still the old edition. When delivered dimensions do not match, each party holds a different drawing edition. The preventive wording is: the technical agreement specifies the drawing version number and the change notification process, each revision is signed off by both parties, and the old drawing and the mould are sealed and retained together.
The forms of dispute can be listed; the costs are hard to list. What can be written clearly in the contract should not be left to negotiation after delivery.
Chapter 32 The Position of Castings in the Supply Chain: Information Asymmetry and Verification Cost
A casting is a typical intermediate component. It is formed to an upstream drawing and handed to downstream machining and assembly, and it rarely appears at the end of the chain under its own brand. The buyer can see the drawing and the incoming goods, but not the molten iron, the moulding sand or the heat-treatment curve. This information structure has long allowed castings to be treated as a black box: give a grade and a drawing, receive a part, and the supplier is left to take responsibility for the process in between.
- Knowledge does not travel with the product
The value of a casting lies in its shape; the knowledge lies in the process. The same part made in different plants with different gating systems, different sand and different cooling conditions can look identical while differing in internal quality and service life. Process knowledge does not travel with the product, and it is not written on the drawing either. What the buyer receives is a geometric body, and the only bases for judgement left are dimensions, appearance and certificates.
The fact that process knowledge is not made explicit also explains why the same drawing made in different plants shows marked price differences: part of the difference lies in the process route and part in testing and process control, and neither item is usually visible on the quotation.
- Invisibility of defects
Internal defects can be found only by ultrasonic testing, radiography or destructive sectioning; surface defects can be concealed by grinding, weld repair and coating. Testing itself is by sampling, and its coverage is limited. Incoming inspection can stop dimensional and appearance problems, but it cannot stop shrinkage cavities and slag inclusions.
The configuration of testing equipment also varies widely: a direct-reading spectrometer is used for composition analysis during melting and before tapping, a coordinate measuring machine for comparison with the digital model, and X-ray or industrial CT for detecting internal shrinkage cavities, gas porosity and cracks (notes on testing configuration). The presence of equipment does not mean the data are trustworthy, but without equipment there are certainly no data.
A non-destructive testing report covers only the portion that was tested, and the untested area remains unknown. The representativeness of sampling depends on the proportion and the locations; if the proportion and locations are not written into the contract, even a complete report shows only that the sample is acceptable.
- Quality costs are deferred
Casting defects often come to light only at machining, assembly or even at the customer's end. By then the chain of responsibility has passed through four stages — casting, machining, surface treatment and assembly — and attributing the defect requires sectioning and testing. Because quality costs are deferred, the buyer tends to release the goods on arrival and transfer the risk to the next operation.
- Industry structure
On the basis used by the Ministry of Ecology and Environment in 2020, there were about 23,000 casting enterprises nationwide (Ministry of Ecology and Environment); the same material states that the top 4,500 enterprises accounted for more than 70% of total casting output. With many enterprises and small individual plants, the cost for a buyer to build process knowledge plant by plant is high. Long-tail enterprises lack public information, and a plant visit shows only that day, so the black box hardens. Sparse public information about long-tail enterprises does not mean weak manufacturing capability, and the process records of leading enterprises are not necessarily complete either. Scale is only the starting point for screening; it cannot replace verification.
- The gap in institutional provision
The 2013 Foundry Industry Access Conditions were abolished on 3 June 2019 (abolition notice). The current basis for industry self-regulation is the China Foundry Association group standard T/CFA 0310021, Specification conditions for foundry enterprises, and group standards are not mandatory (2023 MIIT document).
The binding constraints fall on environmental protection and safety. GB 39726-2020, the Emission Standard of Air Pollutants for the Foundry Industry, sets emission limits for eight operations: metal melting, moulding, core making, pouring, shakeout and cleaning, sand preparation, casting heat treatment and surface coating (standard information, standard interpretation); and the heavy-pollution weather performance grading directly determines which operations are suspended during an alert. These thresholds constrain emissions and safety, not the intrinsic quality of the castings. An outsider can use a pollutant discharge licence to judge whether an enterprise is compliant, but cannot use the same licence to judge whether the castings it makes are good.
- Breakpoints in the procurement chain
Castings often reach the complete-machine plant through traders or intermediaries, and the foundry does not appear on the contract. With each additional layer, the transfer of heat batch numbers, test reports and the responsible entity gains one more breakpoint. The breakpoints themselves do not create quality problems, but the cost of tracing rises with their number.
Intermediaries also serve a purpose: providing working capital, consolidating orders, aggregating small-batch orders, and taking on logistics and inventory are all functions the complete-machine plant does not want to perform itself. The problem lies in the transfer of quality information, not in the existence of intermediaries as such. A workable approach is to require the intermediary to disclose the manufacturing plant in the contract and to pass on heat batch numbers and test reports with the goods.
- The inertia of price anchoring
Castings are usually quoted by the tonne or by the piece; comparing prices is easy, comparing processes is hard. When purchasing performance is assessed mainly on unit price, verification effort is compressed to a minimum. Castings usually account for a small share of the cost in a complete machine's bill of materials, yet the consequences of failure can be large, so purchasing attention and risk do not match. The black box persists because the cost of opening it falls on the buyer, while the benefit is spread across the service life of the complete machine.
What is easily overlooked in price comparison is verification cost. Of two plants quoting the same price, one issues a complete test report for every batch while the other's report covers only the first article, and their actual risk exposures differ. Only when verification requirements are written into the request for quotation, so that all quotes are compared under the same set of requirements, does price comparison become meaningful.
- Turning the black box into a grey box
There are four workable actions. First, heat batch number traceability, binding every finished part to the melting, pouring and heat-treatment records. Second, process records accompanying the goods, covering at least furnace-front composition, pouring temperature and heat-treatment curves. Third, third-party testing brought into sampling, with the sampling performed by the buyer or a third party. Fourth, tiered acceptance, dividing castings into critical, major and general parts according to the consequences of failure and applying a different testing intensity to each.
When none of the four actions is properly carried out, plant audits and certificates become substitutes. They provide an impression of one day and proof of a system, not the data of this batch. Tiered acceptance is divided according to the consequences of failure, not unit price; safety parts, load-bearing parts and appearance parts should have separate testing intensities, and allocating the testing budget evenly amounts to moving intensity away from where it is most needed.
Castings are treated as a black box because of how information is distributed along the supply chain; reducing the degree of black-boxing costs money, and that cost ultimately enters the unit price. What the buyer has to decide is: for which castings, and to what degree of visibility, to pay.
Part Six: Outlook and Boundaries
Chapter 33 The Trajectory of Capacity Concentration and Regional Division of Labour
Any discussion of concentration in casting capacity must first establish where the total stands. On the China Foundry Association's basis, national casting output was 51.95 million tonnes (5,195 万吨) in 2020 14th Five-Year Plan for the Foundry Industry, 50.75 million tonnes in 2024 China Foundry Association, down 2.2% year on year association data paraphrased; the figures appear in a chart published by the association, and the text page does not list them, and 52.30 million tonnes in 2025, up 3.1% year on year and the second-highest level on record 2025 industry data. The net increase across the five years is 350,000 tonnes, equivalent to 0.67%. The total is on a plateau: with the total no longer expanding, concentration can only take the form of a redistribution of the existing stock among enterprises; the natural concentration that incremental growth delivers will not occur.
Redistribution of the existing stock has verifiable samples. Botou in Hebei had 322 foundries in August 2025, of which 79 were above designated size; cluster enterprises had an annual output value of RMB 26 billion (260 亿元) and produced 2 million tonnes of castings a year, equivalent to 3.94% of national output calculated on the 2024 figure Hebei Daily, 25 August 2025. The cumulative change reported in the same article is that 288 low-end enterprises have converted to other production since 2017, the number of foundries has fallen by nearly half, and production efficiency has more than doubled; the share of mid- and high-end products in the cluster has exceeded 50%, and cumulative investment in upgrading and transformation reached RMB 6.806 billion (68.06 亿元). Adding the 322 existing enterprises to the 288 that have exited gives a base of about 610 before the exits, leaving 52.8%, which is consistent with the statement that the number "fell by nearly half".
These figures show that a halving of the number of enterprises can coexist with output and output value. What exited was low-end capacity; the cluster still retains 2 million tonnes of casting output and RMB 26 billion of output value. What declined is the number of enterprises; the scale of output remains.
A contrary sample exists within the same province. Laizhou in Shandong had 245 foundries in 2022, of which 30 were above designated size and 215 below; planned annual capacity was 3 million tonnes, actual annual casting output was close to 2.2 million tonnes, and annual operating revenue was more than RMB 9 billion (90 余亿元) Laizhou Municipal Bureau of Industry and Information Technology, 23 May 2022. Enterprises above designated size account for 12.2%, and more than 80% are below designated size. Laizhou's output comes to 73.3% of planned capacity, while Botou's comes to 50% of the 4 million tonnes of designed capacity given on the industry platform basis. The two ratios show only that a gap exists between designed capacity and actual output; they cannot be used as operating rates, for the reasons given in Chapter 35.
Placed together, the two samples show that concentration has not appeared as a uniform national process: Botou is exiting, while Laizhou still retains 215 enterprises below designated size. For the four verifiable clusters, the share of enterprises above designated size in the total number of enterprises is 24.5% in Botou, about 57% in Ningguo Ningguo Municipal Bureau of Economy and Information Technology, about 40% in Changzhou Changzhou Municipal Bureau of Industry and Information Technology, and 12.2% in Laizhou; the highest is more than four times the lowest, and local differences in enterprise size structure are marked. This report does not extrapolate to the national level on this basis, because the four figures differ in year and in statistical object; see Chapter 35.
The first axis of regional division is material. In 2025, grey iron accounted for 40.2%, ductile iron for 27.2%, cast steel for 12.3% and aluminium (magnesium) alloy for 17.3%, the four together 97.0%, with copper alloy, magnesium alloy and others making up the rest 2025 industry data. Sand casting is predominantly iron and die casting predominantly aluminium, so the material distribution corresponds to two different industrial geographies: the former sits close to machine tools, pumps and valves, pipe fittings and heavy machinery, the latter close to complete vehicles and consumer electronics.
The second axis is the weight of the casting and the organisational form. Botou's large-casting mix has a public statement: large machine-tool castings account for 25% of the national total, pump and valve castings 35%, castings for automotive body-panel dies 16%, tool and measuring-instrument platform castings 60% of the national total, and full-mould castings more than 50% of the national total industry platform basis. This set of ratios was published by an industry platform and no original government text has been seen; this report cites it as a paraphrase. Zhuzhen in Liyang, Jiangsu, is a park-based sample: the park gathers 12 of the town's foundries, with annual casting output of more than 300,000 tonnes, accounting for more than 60% of local casting output Xinhua Daily, 8 June 2023; in 2023 the town produced 300,000 tonnes of castings with sales revenue of RMB 3.8 billion (38 亿元) Yangzhou Network reprinting Xinhua Daily, 28 July 2024.
The third axis is proximity to downstream users. In 2025 demand for castings from the automotive industry was 15.8 million tonnes, 30.2% of the total, an increase of 5% 2025 industry data. The 5% growth in automotive castings is higher than the 3.1% growth in national casting output; in 2024 the automotive share was 29.6% association data. Calculated from the shares and totals of the two years, automotive demand for castings increased by about 780,000 tonnes in a year while national output increased by 1.55 million tonnes, so automotive accounts for about half of the increment. The remaining demand fields are cast pipe and fittings, mining and metallurgical heavy machinery, construction machinery, internal combustion engines and agricultural machinery, machine tools, and power generation equipment and electricity [ibid.]. Die castings are mostly small and medium-sized parts, while sand-casting clusters take on large parts such as machine tools, pumps and valves and pipe fittings; the latter are more sensitive to transport, so their service radius is shorter.
The change in material mix is more pronounced than the change in the total. Aluminium (magnesium) alloy castings rose from 6.8 million tonnes in 2020 to 9.05 million tonnes in 2025, growth of 33.1% over five years, a compound annual growth rate of 5.88%, with the share rising from 16.4% in 2023 to 17.3% in 2025 2025 industry data. Calculated against the 6.8 million tonnes of 2020 and the 51.95 million tonnes of total output that year, the 2020 share was 13.1%. Subtracting the two sets of figures, aluminium-magnesium castings increased by 2.25 million tonnes over five years while all castings together increased by only 350,000 tonnes, so other materials combined fell by about 1.9 million tonnes. This calculation is an arithmetic result from two sets of association figures; the association has not published a corresponding table of increases and decreases by material.
Export geography is shifting. In January–November 2024, exports under 20 tariff lines totalled 5.57 million tonnes, up 12.5% year on year, with a value of USD 15.6 billion (156 亿美元), up 5.9% year on year compiled by the China Foundry Association, 16 January 2025. In January–September 2025, exports on the same basis totalled 4.763 million tonnes, up 4.4% year on year, with a value of USD 13.23 billion, up 3.7% year on year; exports to the United States were 1.034 million tonnes, down 11.9% year on year, with a value of USD 3.16 billion, down 5.4% year on year China Foundry Association data reprinted. On these figures the United States accounts for 21.7% of export volume and 23.9% of export value. The total is still growing while exports to the United States are falling, so other markets are carrying the increment; this report has no full table by country and cannot identify where the increment went.
The two sets of data cannot be joined directly into a single curve. January–November 2024 and January–September 2025 are different statistical periods, and full-year data for 2025 have not been obtained. The 20 customs tariff lines for "cast products" cover only part of China's castings; automotive parts, castings exported with host machines and non-ferrous castings are largely outside them, and the association attaches this note to every monthly report [ibid.]. Exports can be read only for direction on a consistent basis; they cannot be taken as the total volume of China's casting exports.
The institutional conditions for concentration were set out in Part Three; only the direction is summarised here. The 2013 Foundry Industry Access Conditions (《铸造行业准入条件》) were repealed on 3 June 2019 Ministry of Industry and Information Technology Announcement No. 19 of 2019; the current basis for industry self-regulation is the China Foundry Association group standard T/CFA 0310021, Specification Conditions for Foundry Enterprises (2019 edition), which is a non-mandatory standard Ministry of Industry and Information Technology 2023 guidance. With the scale threshold withdrawn, the effective constraints come from emission standards and performance grading; the objects assessed are processes and equipment, and the cost of compliance includes fixed expenditure. Fixed expenditure is diluted by output, so smaller enterprises bear a heavier burden. The specific mechanism is set out in Chapters 16 to 20.
National concentration cannot be answered with data. Only one historical basis can be verified: of about 23,000 foundries, the top 4,500 enterprises account for more than 70% of casting output Ministry of Ecology and Environment, 29 December 2020. On these two figures, about one-fifth of enterprises carry 70% of output. That basis dates from 2019–2020; the total number of enterprises in 2025 and the stock in the two foundry sub-classes C3391 and C3392 have not been obtained, so whether concentration has risen has no denominator at national level and must remain at the level of cluster case studies.
Within the plateau in the total, concentration takes the form of redistribution of the existing stock; regional division is drawn by material, casting weight and service radius, and a single cost advantage cannot explain the differences between clusters.
Chapter 34 Alternative Processes: Forging, Powder Metallurgy and Additive Manufacturing
Any discussion of substitution for casting must first delimit the question. Moving a given part to another process usually requires changing the structure, the machining allowance and the acceptance method; the drawing does not transfer unchanged. Substitution therefore happens at the design stage, and procurement can only choose among processes already fixed by design. Whether a process can be applied to a given part is judged on four capabilities: whether it can form the geometry, whether it can reach the performance, whether it can bear the batch size, and whether it can be accepted by the existing inspection system. Forging, powder metallurgy and additive manufacturing are compared on these four below; the capability envelope of casting is given in Chapter 5.
In forging, solid metal is plastically deformed in a die, and the grains elongate along the direction of material flow to form a flow-line structure. Its capabilities lie in mechanical properties: the structure is dense, free of the shrinkage cavities and shrinkage porosity produced by solidification shrinkage, and load-bearing and fatigue resistance are good. Its boundaries lie in geometry: a forging cannot form complex internal cavities or deep holes, it requires a parting line, draft angles and generous fillets, and flash and machining allowance make material utilisation lower than that of near-net-shape processes; forging dies are costly and subsequent machining is extensive. Typical parts are crankshafts, connecting rods, gear blanks, flanges and shafts. For load-bearing parts where fatigue life is the primary metric, a forging is usually the preferred option; for parts whose main demands are complex internal cavities, thin walls and part integration, forging cannot deliver.
In powder metallurgy, metal powder is compacted in a die and then sintered in a protective atmosphere, forming metallurgical bonds between particles. Its capabilities are near-net shaping, high material utilisation and controllable porosity; oil-impregnated bearings and filter materials make use of that porosity. There are three boundaries: shape is constrained by the pressing direction, so cross holes, undercuts and complex internal cavities require subsequent machining or assembly from separate parts; density forms a gradient along the pressing direction, so properties are uneven; and compaction dies are costly, so only large batches dilute the cost. Typical parts are gears, sprockets, connecting rods, bearing sleeves and small structural parts in automotive engines and transmissions. Small parts, large batches and pressability along the axis are the application range of powder metallurgy; large parts and complex internal cavities remain with casting.
In additive manufacturing, material is deposited layer by layer and bonded; for metal parts, powder-bed fusion and directed energy deposition are common, while sand moulds use binder jetting. Its capability is geometric freedom: conformal cooling channels, lattice structures and internal channels can be formed in one operation without tooling, and lead times for single parts and small batches are short. Its boundaries are cost and post-processing: the cost per part is high and deposition rate limits batch size; build size is limited by the machine chamber; surface roughness and dimensional accuracy usually require subsequent machining; residual stress and anisotropy must be removed by heat treatment or hot isostatic pressing; and powder management, support removal and non-destructive testing form a new post-processing chain. Typical applications are complex aerospace parts, medical implants, conformal cooling inserts for injection moulds, and prototypes.
The relationship between additive manufacturing and casting has two layers, which need to be kept apart. One is substitution for the casting itself: metal additive manufacturing forms the part directly, and the part no longer passes through melting and pouring. The other is substitution for the moulding step of casting: sand 3D printing prints the sand mould and cores directly, dispensing with wooden and metal patterns, while casting remains casting. The 2023 guidance from three ministries lists "integrated die casting, mould-free casting and sand 3D printing" together as advanced process directions for the foundry industry Ministry of Industry and Information Technology 2023 guidance. In the policy's terms, sand 3D printing belongs to casting, not to the replacement of casting. Confusing the two layers mistakes progress in the moulding step for the replacement of castings.
The qualitative comparison of the four capabilities is given in the table.
| Dimension | Sand casting | Forging | Powder metallurgy | Additive manufacturing |
|---|---|---|---|---|
| Material state during forming | Liquid filling followed by solidification | Solid-state plastic deformation | Powder compaction followed by sintering | Layer-by-layer deposition and bonding |
| Geometric freedom | Medium; sand cores can form internal cavities | Low; limited by the parting line and draft | Medium-low; limited by the pressing direction | High; internal channels and lattices can be formed in one operation |
| Tooling | Wooden patterns, metal patterns, core boxes | Forging dies, trimming dies | Compaction dies | No tooling |
| Batch range | Single parts to large batches | Medium to large batches | Large batches | Single parts to small batches |
| Subsequent machining | Medium | Large | Small | Depends on the part |
| Typical defects | Shrinkage cavity, shrinkage porosity, gas porosity, scab | Laps, cut flow lines, incomplete filling | Porosity, density gradient, sintering distortion | Residual stress, anisotropy, lack of fusion |
The table is a qualitative comparison on process-capability dimensions; it contains no data on output, cost or market size; specific parameters for the casting side are given in Chapter 5.
Casting has a mature delivery acceptance standard, a capability often overlooked in substitution comparisons. GB/T 40805-2021, General Technical Conditions for Delivery Acceptance of Steel Castings, specifies ordering requirements, technical requirements, inspection and test methods, inspection rules, marking, quality certificates, packaging and transport, and is a modified adoption of ISO 4990:2015 standard information; ultrasonic testing of steel castings for general purposes is covered by GB/T 7233.1-2023 standard information. An acceptance method that can be written into a contract lowers the buyer's cost of judging quality, and it also creates friction in process switching: change the process and the acceptance method must change with it.
Three decision rules can be drawn from the comparison above.
The first: the failure mode decides the process family. If a part fails through fatigue or overall load bearing, the dense structure and flow lines of forging are an advantage; if the key requirement is sealing, internal channels or thin-wall structure, the internal-cavity forming capability of casting is an advantage; if the part is small, the batch is large and the shape can be pressed along one direction, the near-net shaping of powder metallurgy is an advantage; if the geometry exceeds what a die can be parted into, additive manufacturing enters the candidates. When a part requires both fatigue resistance and complex internal cavities, process selection becomes a trade-off; no single process satisfies both requirements at once.
The second: batch size decides how tooling is amortised. Tooling cost is spread over parts, so the larger the batch, the lower the unit cost of a tooled process; the smaller the batch, the more evident the relative advantage of a tool-free process. The same part may use different processes at the prototype stage and in volume production, and the switch point is decided by batch size.
The third: inspection capability decides whether substitution can be implemented. To accept a part made by a new process, the buyer needs the corresponding standards, equipment and personnel. Casting's inspection system is complete, whereas this report has not obtained evidence on the acceptance methods of the alternative processes, so their acceptability in specific orders cannot be judged. This is a transaction cost and has nothing to do with technical capability.
This report gives neither a substitution rate nor a market size. The evidence gathering did not obtain output figures on a consistent basis for forgings, powder-metallurgy parts and additive-manufacturing parts, did not obtain the installed base and capacity of sand 3D printing, and did not obtain substitution volumes. All that can be confirmed is that the policy basis lists mould-free casting and sand 3D printing as development directions; a direction is not a scale. Any claim as to how many tonnes were substituted in a given year cannot be verified on the publicly available data.
The boundaries of the four process families are drawn jointly by geometry, batch size and failure mode; where they do not overlap on these dimensions, they do not stand in a substitution relationship to one another.
Chapter 35 The Data Boundaries of This Report
The data in this report fall into three categories: keyword-match results from the Tianxia Gongchang enterprise database, public documents from government and statistical agencies, and disclosures by industry associations and listed companies. The three sources differ in statistical object; this report's approach is to label each separately and not to combine them.
The total volume and geographic distribution of our own data are given in Chapter 8; only three figures relevant to the boundaries are listed here: 56,563 entities, 2.7108 million (271.08 万人) insured persons in total, and 35,283 entities with insurance records. The third figure accounts for 62.4%; for the remaining 37.6%, insured persons are zero or missing. The top eight provinces together account for 40,711 entities, or 72.0%; the top eight districts and counties together account for 4,878 entities, or 8.6%.
The top eight districts and counties together account for 4,878 entities, only 8.6%, which indicates that the number of entities is dispersed at county level and that no single point carries a very high share; the full county-level distribution has not been obtained, so this report does not use this to judge whether the whole distribution is long-tailed. Dispersion at county level is related to the mechanism of keyword matching: hit density depends on the frequency with which process terms appear in names and business scope, not on capacity. Traders, supporting plants and downstream workshops in industrial clusters use process terms extensively, so hit density is correspondingly high.
The basis must be stated explicitly: the industry field in the database goes only to the broad category (C33, manufacture of metal products), with no code marker for C339, C3391 or C3392, and the 56,563 entities were obtained by matching process keywords, without filtering by industry code. Keyword matching brings both missed matches and spurious matches.
Missed matches bias the result downward. Enterprises whose name and business scope do not contain terms such as "casting", "castings", "die casting", "investment casting", "precision casting", "sand mould", "ductile iron" or "grey iron" are not counted. These include the captive foundry shops of downstream equipment manufacturers, specialist foundries named after "machinery", "hardware" or "metal products", small plants that only do job-work processing, and entities that have ceased production but remain in the database. Enterprises "containing a casting process" make up a considerable proportion, and missed matches affect this group most. This report cannot estimate the proportion of missed matches because there is no independent enterprise directory to compare against.
Spurious matches bias the result upward, and this can be checked. Two examples. First, Botou returns 692 hits in this database, while the official figure for Botou City in 2025 is 322 foundries Hebei Daily; the former is 2.15 times the latter. Second, Wujin District alone returns 639 hits, whereas the Changzhou Municipal Bureau of Industry and Information Technology's 2024 figure is more than 600 foundries for the whole city reply from the Changzhou Municipal Bureau of Industry and Information Technology; the hit count for a single district exceeds the official enterprise count for the entire city. The two comparisons support only one conclusion: keyword hits cover a wider set of objects than the administrative definition of a foundry and include traders, equipment and material suppliers, mould shops, testing bodies and downstream captive workshops. The figure can be taken neither as the total number of foundries nor as the number of specialist foundries.
Insured-person data follow the reporting basis. The insured-person counts in enterprise annual reports are affected by reporting quality and cannot be compared directly with persons employed on the statistical basis. Dividing 2.7108 million by all 56,563 entities gives 47.9 persons per enterprise; this average puts the 37.6% of entities with zero insured persons into the denominator and has no interpretative meaning. Counting only the 35,283 entities with insurance records, the average is 76.8 persons. Neither figure can be used to calculate output value per person.
Public statistics go only to the broad category. Table 3-2 of the Third Bulletin of the Fifth National Economic Census lists 455,000 legal entities and 7.651 million persons employed in the manufacture of metal products (C33), and Table 3-3 lists operating revenue of RMB 6,236.51 billion (62,365.1 亿元) for the same broad category National Bureau of Statistics, Fifth National Economic Census Bulletin (No. 3). Casting falls under the two sub-classes C3391 and C3392, both within the medium class C339 under C33; the bulletin does not list medium classes, so the number of enterprises, persons employed and operating revenue for the medium class have not been obtained. The 455,000 legal entities cannot be taken as the number of foundries, nor apportioned proportionally: C33 also contains a large number of metal-products enterprises that do not cast.
Two bulletins from the same census show one small discrepancy: the Second Bulletin gives 4.049 million legal entities in manufacturing National Bureau of Statistics, Fifth National Economic Census Bulletin (No. 2), while Table 3-2 of the Third Bulletin gives 4.048 million. This report cites the latter.
Cluster data are incomplete, and both year and definition are inconsistent. The table lists the cluster bases and gaps that this report can confirm.
| Cluster | Verifiable basis | Data period | Source | Gap |
|---|---|---|---|---|
| Botou, Hebei | 322 foundries, 79 above designated size, output of 2 million tonnes, output value of RMB 26 billion | August 2025 | Hebei Daily | Designed capacity of 4 million tonnes and 28,000 persons employed come from an industry platform; no original government text has been seen |
| Ningguo, Anhui | More than 170 wear-resistant casting enterprises, 97 above designated size, output value of RMB 12 billion | 2022 | Ningguo Municipal Bureau of Economy and Information Technology | By the end of 2025 there were more than 170 upstream and downstream enterprises with output value exceeding RMB 12 billion China Quality Daily, and the two figures of 170 do not mean the same thing |
| Changzhou, Jiangsu | More than 600 foundries, about 240 above designated size | May 2024 | Changzhou Municipal Bureau of Industry and Information Technology | City-wide output and output value not published |
| Zhuzhen, Liyang, Jiangsu | 12 enterprises in the park, output of more than 300,000 tonnes; 300,000 tonnes and sales revenue of RMB 3.8 billion for the whole town | 2023 | Xinhua Daily, Yangzhou Network | The relationship between the two figures of 300,000 tonnes, one for the park and one for the town, is not explained |
| Ningbo, Zhejiang | Die-casting capacity about 10% of the national total and 42% of the provincial total; 800 enterprises to be upgraded to standard by 2025 | 2021 | Ningbo Municipal Bureau of Economy and Information Technology | The total number of die casters does not appear in the bureau's documents |
| Laizhou, Shandong | 245 foundries, 30 above designated size, output of close to 2.2 million tonnes, revenue of more than RMB 9 billion | 2022 | Laizhou Municipal Bureau of Industry and Information Technology | No updated year |
| Jiangsu province | More than 3,000 foundries | 2024 | Jiangsu Economic Daily | Operating revenue is given only as "among the highest in the country", with no specific figure |
| Linqing, Shandong; Foshan, Guangdong; Wuxi, Jiangsu | Not obtained | — | — | Public material from the Linqing municipal government covers only the bearing industry; statistics for Foshan do not list casting separately; Wuxi has had no municipal-level official basis since 2023 |
The table is compiled from the results of a check of public material, with a verification date of 18 October 2026.
This table cannot be used for ranking clusters or for aggregation. The years of the rows range from 2021 to 2025, and the statistical object switches among "foundries", "upstream and downstream enterprises", "upgrading targets" and "enterprises gathered in a park". Adding figures with different definitions does not give a national cluster total. When citing cluster data, this report labels the year and definition for each item and makes no comparison across clusters.
Cost data come from scattered sources and cannot be combined into a cost per tonne. Three listed precision-part companies disclose direct material shares of 36.72%, 50.65% and 45.58%, an average of 44.32%; direct labour averages 16.82%; manufacturing overhead (including power) averages 38.87% Nante Technology response to an inquiry. These figures are on a precision-parts basis and include machining; applying them directly to pure casting would overstate the material share and understate labour and depreciation. For raw materials there is only one source: the Mysteel spot index page captured on 4 October 2026, which shows live quotations whose values change with each trading day; this report has not obtained an archivable point-in-time value. For energy, what is available is the electricity consumption bands for coreless induction furnaces (including medium-frequency) in the 2013 access conditions: 630 kWh per tonne of molten metal or less for capacities of 1 tonne and below, and 590 kWh per tonne of molten metal or less for 5 tonnes and above full text of the 2013 access conditions reprinted; these are the band limits of the time, and industry averages have not been obtained. For the respective shares of resin-bonded sand and electricity in the cost per tonne of castings, this report has not obtained a verifiable public basis. The three parts cannot be added, so this report gives neither a cost per tonne of castings nor a percentage breakdown of cost structure.
For items marked "not obtained" in the evidence file, this report leaves the figure blank throughout. Besides the three cluster gaps in the table above, they include: absolute tonnages by material for 2025; the national installed base of moulding lines and medium-frequency furnaces; the shares of resin-bonded sand and electricity in cost per tonne; the specific emission limits for the four grades of the heavy-pollution-weather performance classification; and the total number of foundries nationwide in 2025.
The evidence file grades reliability from A to D, and this report handles them as follows: grade A is cited directly with a link to the original publication page; grade B is marked as "paraphrased" or "industry platform basis"; grade C is used only as background, with no specific figures; grade D is never used. Typical grade B items are the 2024 tonnages by material, Botou's designed capacity of 4 million tonnes and its 28,000 persons employed; grade C is used only for process examples.
Standards and policies have a time boundary, and the verification date is 18 October 2026. The standard for dimensional tolerances of castings changed edition on 1 December 2025: GB/T 6414-2017 was withdrawn on 1 December 2025 standard status, and GB/T 42124.3-2025 took effect the same day, being a modified adoption of ISO 8062-3:2023 information on the new standard. The target year for many items in the 2023 guidance is 2025; this report has not obtained the achieved values and cites them as targets, not as results.
The directional judgements in this report are usable; absolute values are not citable. Conclusions on the three matters of industry totals, concentration and substitution rates are all constrained by the gaps listed above.
Chapter 36 Questions This Report Does Not Resolve
The following are the questions that this report established it cannot answer in the course of writing. For each, it states what data are missing, what type of conclusion cannot stand without them, and what survey is needed to fill the gap. Most of these problems arise from a missing denominator, a few from data that are not public. Figures cited in this chapter have already been sourced in Chapters 33 and 35 and are not linked again here.
First, the industry stock and size distribution. Missing are the number of legal entities, persons employed and operating revenue for the two foundry sub-classes C3391 and C3392, as well as the size distribution of enterprises by revenue or by output. The public tables of the Fifth National Economic Census stop at the broad category C33, and medium-class data require the China Economic Census Yearbook 2023. Without these data, the national average scale per plant and national concentration cannot be calculated, nor can it be judged whether Botou's "half the enterprises, undiminished output" is representative nationwide. This report can only use the 2020 Ministry of Ecology and Environment basis of 23,000 foundries with the top 4,500 accounting for more than 70% of output, a point now five years past. It is likewise impossible to judge whether the 56,563 entities in this database are high or low relative to C339, because the denominator itself has no authoritative value.
Second, the separation of capacity from output. There is no unified national basis for casting capacity; local bases appear only in cluster material: Botou has a designed capacity of 4 million tonnes (industry platform basis) against actual output of 2 million tonnes, and Laizhou has planned capacity of 3 million tonnes against actual output of close to 2.2 million tonnes. Designed capacity is an approval and planning basis while actual output is a statistical basis, and the ratio between them cannot be used as an operating rate. There is no enterprise-level capacity ledger or operating-rate survey, so it cannot be answered how much is surplus, or how much capacity is actually idled by environmental production restrictions. These two questions decide whether concentration is an exit driven by cost or a result of administrative restriction; this report cannot distinguish between them.
Third, the harmonisation of cluster bases. Official casting bases for Linqing, Foshan and Wuxi are missing; the existing cluster data range in year from 2021 to 2025, and the statistical object switches among "foundries", "enterprises above designated size", "upstream and downstream enterprises" and "enterprises in a park". Without a unified survey, clusters cannot be ranked and the scale of capacity migrating from one cluster to another cannot be quantified. The migration discussed in Chapter 14 can rest only on individual company cases and local plans and cannot give a net migration figure. Filling this gap requires an enterprise directory survey with a uniform year, uniform definitions and coverage of the main clusters.
Fourth, cost and energy consumption. Missing are the itemised shares of resin-bonded sand, electricity, labour, tooling and environmental protection in the cost per tonne of castings, as well as line-level measured electricity consumption and yield statistics. The existing 590 to 630 kWh per tonne of molten metal are the electricity consumption limits of the 2013 access conditions, and industry averages have not been obtained; the cost structure of the three precision-part companies includes machining, and the cost structure of pure casting has not been obtained. Without these data, the cost per tonne cannot be calculated, electricity-price sensitivity analysis cannot be performed, and the cost curves of sand casting, die casting and investment casting cannot be compared. Chapter 22 can give only a list of cost items and public prices, not structural proportions.
Fifth, the penetration of alternative processes. Missing are output figures on a consistent basis for forgings, powder-metallurgy parts and additive-manufacturing parts; the installed base and capacity of sand 3D printing; and the volume of substitution by the three process families in uses previously served by castings. The China Foundry Association publishes annual statistics on casting output, but this report has not obtained corresponding statistics for the other three, so the four processes cannot be placed in a single output table. Forgings and powder-metallurgy parts have their own industry statistics, but these use a different basis from castings, and subtracting one from the other would mix in differences in statistical scope. Without these data, Chapter 34 can only assess capability boundaries; it cannot answer how much substitution has already occurred, nor judge whether substitution has changed the material mix. The rise in the aluminium-magnesium casting share to 17.3% in 2025 may be related to automotive lightweighting, or to an expansion in the statistical scope for die castings; this report cannot distinguish between the two.
Sixth, the true scale of exports. The 20 customs tariff lines for cast products cover only part of China's castings; automotive parts, castings exported with host machines and non-ferrous castings are largely outside them; and full-year data for 2025 have not been obtained. Without a full tariff-line match and disclosure of the association's estimation method, the export dependence of the foundry industry cannot be calculated, and the full effect of the 11.9% fall in exports to the United States on total exports cannot be assessed. What can be confirmed is only the direction on a consistent basis: the total is still growing, and exports to the United States are falling.
Seventh, the capacity effect of compliance. Missing are the distribution of enterprises across grades A, B, C and D of the heavy-pollution-weather performance classification; the actual duration of production restrictions at each grade during warnings; the investment in compliance upgrading by enterprises; and a register of the reasons why enterprises exited. The only officially stated exit figure this report could find is the 288 in Botou, but the reporting does not break down the reasons for exit: how much is attributable to environmental non-compliance, to being eliminated by the market and to land clearance cannot be determined. Without these data, the mechanism analysis in Chapters 20 and 21 can go only as far as direction and cannot give a figure for how much capacity was withdrawn under environmental constraints.
Eighth, the geographic distribution of quality and yield. Missing are industry statistics on reject rates, rework rates and batch consistency, and the installed base of inspection equipment, including direct-reading spectrometers, coordinate measuring machines, X-ray and industrial CT. What this report could find is limited to descriptions of inspection methods in company technical articles and enterprise-level configuration cases, with no industry distribution. Without these data, differences in quality capability between clusters cannot be judged, nor can cluster self-descriptions such as "the share of mid- and high-end products has exceeded 50%" be verified. Batch consistency, the thing buyers care about most, is precisely the thinnest layer of public data. Yield is directly related to cost, and the absence of yield data also leaves the cost analysis in Chapter 22 without a key variable.
The eight questions differ in the route to filling them and in priority. Medium-class stock comes from the China Economic Census Yearbook 2023; cluster bases from the annual ledgers of local industry and information technology departments; cost and energy consumption from enterprise surveys and meter-level measurements; substitution and exports from the full-scope statistics of the association and customs; compliance effects from the grading lists and the register of exited enterprises. The first three decide the denominator, the fourth to the sixth decide direction, and the seventh and eighth decide the credibility of judgements on compliance and quality. Until these data are public, discussion of Chinese casting can stay only at three levels: process boundaries, institutional boundaries and cluster case studies. What this report can establish is where the boundaries lie; what it cannot establish is the speed and scale of change. If medium-class stock and cluster ledgers become public, the report's judgement on concentration can rise from individual cases to the national level; if measured cost and yield data become public, the analysis of process choice can move from boundaries into quantification.
Data Sources and References
The data in this report fall into three categories. The first is the supply-side statistics of the Tianxia Gongchang enterprise database, compiled by matching keywords against company names, products and business scope rather than by taking figures by industry code under the Industrial Classification for National Economic Activities; the basis is explained in each chapter, and these data are used only for distribution and structural comparison, not as national totals. The second is government documents, national standards, association statistics and census bulletins. The third is public company disclosures and industry reporting, whose nature is indicated in the text where they are cited. The last verification date for all links is 18 October 2026.
- Tianxia Gongchang industrial platform. The counts of manufacturing entities, insured persons and regional distributions in this report are all taken from the Tianxia Gongchang enterprise database.
The other sources cited in the text are as follows.
- China Foundry Association
- Guidance from the Ministry of Industry and Information Technology and two other ministries
- Sand casting capability page
- Industry encyclopaedia process entry
- Sina reprint
- Investment casting process description
- Die casting technical reference page
- Casting process comparison page
- Die casting capability introduction page
- GB/T 13295-2026 excerpt page
- National standards full-text public system
- National standards full-text public system
- Standard information page
- 2024 foundry industry data
- Paraphrase of 2024 data
- 14th Five-Year Plan for the Foundry Industry
- Paraphrase of 2024 data
- Fifth National Economic Census bulletin
- Ministry of Ecology and Environment
- Hebei Daily
- Industry material
- gov.cn: Xinhua News Agency
- Laizhou Municipal Bureau of Industry and Information Technology
- China Economic Net
- Changzhou Municipal Bureau of Industry and Information Technology
- Xinhua Daily
- Yangzhou Network reprinting Xinhua Daily
- China Jiangsu Network
- Jiangsu Economic Daily
- Industry material
- Ningbo Municipal Bureau of Economy and Information Technology
- Foshan industrial map
- China Foundry Association data
- 2024 export data
- Ningguo Municipal Bureau of Economy and Information Technology
- China Quality Daily
- Ningguo municipal government
- Announcement reprinted on the State Council website
- Ministry of Industry and Information Technology repeal announcement
- Full text of the access conditions reprinted on the Anyang Municipal Bureau of Industry and Information Technology website
- Metadata of the national standards full-text public system
- Mysteel spot index page
- Judgement standard published in the State Council Gazette
- Reply to a proposal published by the Yizheng Municipal Ecology and Environment Bureau
- Heavy-pollution-weather emergency response measures published by a local government
- Nante Technology response to an inquiry on its public offering
- GB/T 40805—2021, General Technical Conditions for Delivery Acceptance of Steel Castings, lists the quality certificate as a mandatory item of delivery acceptance
- Direct-reading spectrometers are used for rapid composition analysis during melting and before tapping, coordinate measuring machines for comparison with the digital model, and X-ray or industrial CT for detecting internal shrinkage cavities, gas porosity and cracks
- GB/T 7233.1—2023
- Not applicable to austenitic steels or austenitic-ferritic duplex stainless steels; martensitic stainless steels may be used for reference
- GB/T 5677—2018, Radiographic Testing of Castings
- GB/T 9444—2019, Magnetic Particle Testing of Steel and Iron Castings
- GB/T 9443—2019, Penetrant Testing of Steel and Iron Castings
- ISO 4990:2023
- Example certificate
- National Bureau of Statistics, Fifth National Economic Census Bulletin (No. 2)
Cover image: Mohsina Faiz / CC BY-SA 4.0 / original file page. Used under the original licence without alteration to the content; only resized proportionally to 2,000 px wide and stripped of metadata.