Aller directement au contenu principal
Demander un devis
,

Comment fabriquer un moule de coulée pour le métal : guide de fonderie étape par étape

📅 23 octobre 2026 💬 Par Jeff

Making a casting mold for metal is a sequence of decisions, not a single act of packing sand. The pattern has to be oversized by the right amount, the sand has to be mixed and compacted to a measurable strength, the parting line has to let the pattern come out without tearing the cavity, and the gating system has to deliver metal fast enough to fill the part and quietly enough that it does not entrain air. Get those four things right and the pour becomes routine. Get any one of them wrong and the defect shows up three steps later, looking like a metal problem when it is a mold problem.

This guide is the mold-making operation itself: what to build, in what order, with the numbers a molder actually writes on the shop card. It covers the traditional route (pattern, flask, rammed sand) and the point at which a printed sand mold removes the pattern from the process entirely.

Two halves of a sand mold for an oval port housing, cope and drag showing the cavity and gating

The Direct Answer

To make a casting mold for metal, you build a cavity in bonded sand that reproduces the part, then feed it with metal through a gating system the mold itself contains. In practice that means: make a pattern oversized for shrinkage and drafted so it withdraws cleanly; split the mold at a parting line into cope and drag; mix and ram sand to a controlled strength and permeability; withdraw the pattern and repair the cavity; cut sprue, runner, ingates, risers and vents; set any cores; close, clamp and weight the mold; then pour, let it solidify, and break the sand away.

The mold has to do three jobs at once, and every design choice serves one of them:

  • Hold the shape. The cavity must survive the weight and velocity of the incoming metal without deforming, eroding or collapsing.
  • Let the gas out. Sand and binder release steam and decomposition gas on contact with liquid metal, and the cavity holds air that must escape ahead of the fill.
  • Come apart again. The mold is expendable, so it must break down cleanly at shakeout without damaging the casting or welding itself to the surface.

Everything below is in service of those three jobs. If you only remember one thing: mold quality is decided before the metal arrives. By the time the ladle is over the sprue, the outcome is already largely fixed.

What a Metal Casting Mold Is Made Of

The vocabulary is worth fixing before the steps, because a lot of mold problems are described in the wrong words and therefore diagnosed in the wrong place.

  • Pattern — the physical replica of the part, oversized for contraction and carrying draft and machining allowances. It is withdrawn before pouring and reused.
  • Flask — the frame that contains the sand. The lower half is the drag, the upper half is the cope. A three-part mold adds a cheek.
  • Parting line — the plane where cope and drag meet, chosen so the pattern can be withdrawn without undercuts.
  • Cavity — the impression left by the pattern; it becomes the outside of the casting.
  • Core — a separate sand shape set into the cavity to form internal passages, bores or undercuts the pattern cannot produce.
  • Gating system — pouring basin, sprue, runner and ingates: the channels that deliver metal into the cavity.
  • Risers (feeders) and chills — the reservoir and the heat sinks that control where the metal freezes last, so shrinkage ends up in the riser instead of in the part.
  • Vents — the escape routes for air and binder gas.

In a permanent mold or die casting tool these elements live in steel and are reused thousands of times. In sand casting the mold is used once, which is exactly why the tooling cost is low and the geometry can be changed cheaply. The trade is that every part consumes a mold, and the mold is where the process variation enters.

Step 1: Choose the Molding Route Before You Design Anything

The route decides the pattern, the sand, the tooling cost and the tolerance you can hold. Choosing it after the pattern is made is the most expensive mistake in the sequence.

ItinéraireBinder and cureIdéal pourHolds roughlyWatch out for
Green sandBentonite clay and water, compacted moistVolume production; iron and aluminum; medium complexityCoarser tolerances, rougher finishMoisture-driven gas defects; operator technique matters
Resin no-bake (furan, phenolic)Liquid resin plus acid or ester catalyst, cures at room temperatureMedium and large castings; better finish and strength than green sandBetter than green sandBench life limits working time; sand reclamation is more involved
Shell moldingResin-coated sand cured against a heated metal patternThin, detailed, higher-volume parts; good surface finishGood dimensional consistencyMetal pattern required; size limited by the shell machine
3D printed sand mold and coreSand jetting with a furan or phenolic binder, printed directly from CADPrototypes, one-offs, complex internal geometry, redesign-heavy programsTight for sand, and repeatableBuild volume and machine time; sand is consumed per mold

Three questions settle it: how many parts, how tight is the tolerance, and how likely is the design to change? Tooling is amortised over the batch, so low volume punishes any route that requires a pattern. Where the geometry has internal passages a pattern cannot form, or the program is still iterating, printing the mold removes both constraints — we return to that in patternless sand casting and its cost profile in Coût de l'impression 3D au sable.

Step 2: Build the Pattern, or Decide to Skip It

A pattern is the part plus allowances. Every allowance is a number the molder has to be able to defend, because it removes material from the final casting or makes the pattern unusable if it is wrong.

The allowances a pattern carries

AllowanceValeur typiqueWhy it exists
Shrinkage (contraction)See the alloy table belowMetal contracts as it cools from pouring to room temperature
Draft1–2° on vertical faces; 3–5° on deep pockets and hand-rammed workLets the pattern withdraw without dragging sand out of the cavity wall
Usinage1.5–3 mm on machined faces; more on large steel castingsRemoves the decarburised or sand-affected skin and gives the machinist clean stock
DistortionPart-specific; judged from experience or simulationThin, long or U-shaped parts warp as they cool and restrain each other
Shakeout and handlingAdded to the pattern where the casting needs a lifting or clamping faceGives the cleaning shop something to grip without damaging a finished surface

Shrinkage is the allowance people get wrong most often, because it is not one number for metal in general but one number per alloy, and it is not uniform in a part with mixed section thicknesses.

AlliageLinear shrinkage allowance on the patternNotes
Grey cast iron0.8–1.0%Graphite expansion offsets some contraction
Fonte ductile0.8–1.0%Higher shrinkage tendency than grey iron; feeding design matters more than the pattern allowance
Carbon and low-alloy steel1.5–2.0%Use the high end for large, restrained sections
Acier inoxydable2.0–2.5%High contraction and low thermal conductivity
Aluminium alloys1.0–1.3%The default for general engineering castings
Copper alloys (brass, bronze)1.3–1.6%Tin bronzes tend toward the lower end
Magnesium alloys1.2–1.4%Also needs inhibited sand to control reaction

Treat these as the starting point for the foundry’s own shrink rule, not as a specification. Two foundries casting the same alloy from the same drawing will use different numbers, because the effective contraction depends on how much the mold and cores restrain the casting as it cools.

Pattern material

MatériauRuns it survivesOù cela s'applique-t-il ?
WoodDozensOne-offs and first articles; easy to modify when the design moves
Urethane board, epoxy, resinHundredsThe working default for small and medium batches; stable, easy to machine, resists moisture better than wood
AluminiumThousandsProduction patterns and matchplates; light enough to handle on a molding machine
FonteDes dizaines de milliersHigh-volume work where dimensional stability over years matters
Printed pattern (SLS or resin)One to a fewComplex geometry and design iteration, where making a pattern at all is the bottleneck

Add a pattern plate or matchplate when the flask is handled on a machine: cope and drag patterns mounted on a single board keep the two halves registered to each other and remove an entire class of mismatch defects. Hand molding with loose patterns can match that accuracy, but only with a locating pin system and discipline about using it.

When the pattern disappears from the process

A binder jetting sand printer builds the mold and core directly from the CAD model: no pattern, no draft requirement, no parting line forced by withdrawal, and internal passages that a core would normally form can be printed into the mold as a single piece. The economic case appears when the program is one or a few parts, when the geometry would need several cores, or when the design is still changing — cases where tooling cost and tooling lead time dominate. Beyond roughly a few hundred identical parts with stable geometry, a pattern and a molding line usually win on piece cost.

Sand 3D printer building a foundry mold directly from a CAD model without a pattern

Step 3: Set the Parting Line and Design the Flask

The parting line is the single most consequential decision in mold design, because it fixes the draft direction, the core layout, the gating entry points and where any mismatch can appear on the casting.

A workable parting line satisfies four conditions:

  • Every vertical surface has draft in the direction it is withdrawn — no undercuts against the draw.
  • At least one half of the mold can be formed by a single pattern or plate without loose pieces.
  • Internal passages that cannot be molded directly are reachable by a core that can be located and supported.
  • The split lands on a surface where a small mismatch is acceptable, not on a finished sealing face or a visible datum.

Cope height matters for more than clearance. The metal head above the highest point of the cavity sets the filling pressure, and that pressure is what drives both complete fill and mold penetration. A shallow cope starves thin sections and invites misruns; an unnecessarily tall cope raises the metallostatic force the clamping system has to resist. On most work, keeping 50–100 mm of metal above the highest cavity point is enough to fill the part, with more where walls are thin or the alloy is sluggish.

Flask size should be set from the casting, not from what is on the shelf. Leave enough sand around the cavity to resist the metal pressure and to give the rammer room to work — a common guide is a wall of sand at least equal to the casting’s section thickness, and never less than about 40–50 mm. Sand that is too thin at the wall is where swells, run-outs and mold shifts begin.

Step 4: Prepare the Molding Sand to a Measurable Target

Sand is the mold. Its job is to be strong enough to hold a cavity, permeable enough to pass gas, refractory enough to survive contact with liquid metal, and weak enough to collapse at shakeout. Those four requirements pull against each other, which is why a sand system is controlled by test rather than by eye.

For a green sand system, the mixture is essentially silica sand, clay, water and a small amount of additive. Typical working ranges look like this:

Component or propertyTypical rangeWhat it controls
Silica sandBalance of the mix, roughly 85–92%Refractoriness and the base grain structure
Bentonite clay6–10%Green strength and plasticity; more clay means more strength and less permeability
Water2.5–4.5%Activates the clay; too little gives brittle molds, too much gives gas defects
Seacoal or carbonaceous additive2–6% where usedSurface finish and peel on iron castings
Compactability35–45%The practical handle on water content on the shop floor
Green compressive strengthRoughly 60–100 kPaWhether the cavity wall survives handling and the pour
PermeabilityTuned to the alloy and sectionHow easily trapped gas escapes through the sand
AFS grain fineness40–70 for most foundry workSurface finish against permeability: finer grain, smoother surface, tighter gas path
Mold hardness (B scale)85–95 on machine-molded green sandCompaction uniformity — and uniformity matters more than the peak number

Read those as ranges to tune, not a recipe. The right targets depend on the alloy, the section thickness, the flask depth and the sand system you already have, and they only mean anything if the tests are run the same way every shift. The tests worth having are moisture, compactability, green compressive strength, permeability, loss on ignition and a sieve analysis of the returned sand.

For resin no-bake sand, the logic changes: strength comes from the binder, so the controls are binder addition (commonly around 1.0–1.8% of sand weight), catalyst addition as a percentage of the resin, sand temperature, and the clock. Bench life is the time you have to fill and compact the flask; strip time is when the mold can be lifted off the pattern. Both shorten as sand or shop temperature rises, which is why summer work needs a slower catalyst. Our sand and binder guide goes deeper on grain selection and coatings.

Step 5: Ram the Mold to Uniform Density

Compaction is where most mold defects are born. The goal is not maximum hardness, it is even hardness: a mold that is hard at the pattern face and soft at the flask wall will move when the metal arrives, and a mold rammed unevenly around a deep pocket will tear the cavity wall on withdrawal.

Whatever method is used, the mechanics are the same: sand is added in layers, each layer is compacted, and the layers are kept thin enough that the energy reaches the bottom of the flask. Sand is placed around the pattern first and worked into pockets and corners by hand before the bulk of the flask is filled — the places a machine cannot reach are exactly the places that fail.

MéthodeHow it compactsOù cela s'applique-t-il ?Characteristic fault
Hand rammingManual rammer and mallet, layer by layerOne-offs, large flasks, awkward geometryNon-uniform density between molders
Jolt-squeezeTable jolts sand into place, then a squeeze head compacts the topSmall and medium flasks in job shopsSoft zone in the middle of deep flasks
Air-impact / pneumaticCompressed air drives sand against the patternMedium production, deep pocketsOver-compaction at the pattern face if poorly set
High-pressure moldingSqueeze pressures high enough for near-uniform densityVolume production linesRequires matched patterns and sand control to be worth it

Two checks catch most ramming problems before the pour. A mold hardness test across the face shows the spread from pattern to wall; and venting the cope over deep pockets, blind bores and core prints gives the gas somewhere to go that is not through the metal.

Deep Dive: Sand Control — What to Test and How Often

A target value with no test method and no frequency is not process control. The numbers below are typical shop practice; the frequencies matter more than the targets, because a single spot test tells you almost nothing and a trend tells you everything.

TestWhat it tells youTypical targetTypical frequency
MoistureThe water actually in the system, which drives both strength and gas2.5–4.5%Every shift, or continuously on a high-volume line
CompactabilityThe shop-floor handle on water content; the fastest read on whether the muller is doing its job35–45%Every shift
Green compressive strengthWhether the cavity wall survives handling, closing and the pourRoughly 60–100 kPaDaily to weekly, depending on volume
PermeabilityHow easily gas escapes through the sand rather than into the metalTuned to alloy and sectionWeekly, and after any sand-source change
Loss on ignitionBinder and carbonaceous build-up; whether the system is accumulating dead materialSet from your own baselineWeekly to monthly
Methylene blue clay contentActive clay versus total clay — the difference between strength you have and strength you think you haveTracked against additionsMonthly, or after any system change
Sieve analysis / AFS grain finenessWhether the grain distribution has drifted toward fines, which trades permeability for finish40–70 GFN for most workMonthly, and on every new sand delivery
Mould hardness across the faceCompaction uniformity, which matters more than the peak value85–95 (B scale) on machine-moulded green sandEvery pattern change, then spot checks per shift
Sand temperatureWhy the same recipe behaves differently in August; hot sand shortens bench life and ruins compactability controlCool enough to control moistureDaily in warm weather

Two habits separate a controlled sand system from a lucky one. First, sample the same way every time — a test result is only comparable to another result taken the same way. Second, plot the trend: a compactability reading drifting from 38% to 44% over a week is a warning about either the muller, the water addition or the return sand, and it is visible long before it turns into porosity in a casting.

Step 6: Withdraw the Pattern and Repair the Cavity

Withdrawal is a controlled operation, not a pull. The pattern is rapped or vibrated just enough to break contact with the sand, then drawn straight along the draft direction. Drawing at an angle tears the cavity wall on one side and closes it on the other.

  • Break the bond first. A light rap or a vibrator loosens the pattern without distorting the cavity.
  • Draw in the draft direction only. Any side load is a repair job.
  • Use a release agent where the sand wants to stick, and keep it out of the cavity surface where excess will show on the casting.
  • Repair immediately. Fins, loose sand and broken edges are fixed with the same sand and a slick, while the mold is still open and visible. Once the cope goes on, a torn edge becomes a sand inclusion.

After repair, clean the parting surface and check the cavity against the drawing once more. It is far cheaper to catch a misplaced core print or a missing boss here than at inspection.

Step 7: Cut the Gating System

The gating system decides whether the cavity fills completely, whether it fills quietly, and whether the metal arrives with the gas and oxide content it left the furnace with. It has four jobs: deliver the metal, control the fill rate, trap slag and oxide, and establish a thermal gradient that feeds shrinkage toward the risers.

ElementFunctionWhat goes wrong
Pouring basinReceives the stream and keeps the sprue fullToo small or too shallow, so the stream breaks and entrains air
SprueVertical delivery channel from basin to runnerStraight instead of tapered, so the falling stream aspirates air and pulls gas into the metal
RunnerHorizontal distribution to the ingatesUndersized, so the metal arrives cold and slow; or excessively long, so it loses temperature before the last ingate
IngatesEntry points into the cavityToo few or badly placed, giving hot spots, erosion at the entry, or a cold shut where two fronts meet
Risers (feeders)Reservoir of liquid metal that feeds shrinkage as the casting freezesUndersized or badly placed, so shrinkage appears as internal porosity in a heavy section
ChillsLocal heat sinks that accelerate freezing in a heavy sectionNot used where they are needed, so the thermal gradient feeds the wrong way
VentsEscape route for air and binder gasMissing over deep pockets and core prints, producing blows and gas porosity

Three rules do most of the work in practice. Keep the sprue full — a tapered sprue, or a basin deep enough that the metal never draws a vortex, prevents the aeration that shows up later as oxide inclusions. Size the runner and ingates from the fill time you want, not from convenience: too fast erodes sand and entrains air, too slow produces misruns and cold shuts, and the window narrows as section thickness falls. Feed the heavy sections — a riser has to freeze after the section it feeds, which in practice means sizing it from the casting modulus and placing it where the last liquid will be, with chills used to push the gradient in the right direction.

Because gating is where the mold and the metal interact most violently, it is also where the most common defects are decided. Casting defects covers the full picture; the mold-side causes are repeated in the troubleshooting table below.

Deep Dive: Gating Design You Can Calculate

Most gating advice stops at “keep the sprue full”. Four numbers decide whether a system works, and all four can be estimated before anything is cut into sand.

QuantitéExpressionWhat it decides
Fill time tChosen from casting weight and the thinnest sectionEverything downstream: flow rate, choke area, velocity
Flow rate QQ = Vcasting / tHow much metal must pass per second
Choke area AA = Q / (Cd √(2gh))The smallest cross-section, and therefore the fill rate
Velocity at the choke vv = Q / ASand erosion and air entrainment in the gating itself
Velocity at the ingatevingate = Q / AingatesWhether the metal enters the cavity quietly or as a jet
Sprue taperAtop / Abottom = √(hbottom / htop)Whether the sprue runs full or aspirates air
Reynolds number ReRe = ρ v D / μHow turbulent the stream is — useful for diagnosis, not a design target

Fill time is the first decision because it sets the flow rate. Heavy castings and thick sections tolerate a slow fill; thin sections need a fast one, because the first metal in must still be liquid when the last metal arrives. The windows below are starting points for medium-size castings — the point is to fix a number, write it down, and correct it from the first pour rather than to fit someone else’s formula.

Thinnest sectionStarting fill window (5–50 kg casting)What you are avoiding
Under 6 mmFast — seconds, not tens of secondsMisrun and cold shut; the metal freezes before the cavity is full
6–25 mmThe broad middleBoth failure modes are live; velocity is the usual mistake
Over 25 mmSlow and quietSand erosion, penetration and burn-on from a fast, heavy stream

Worked example: choke area for a 25 kg aluminium casting

Take a 25 kg aluminium casting. At a liquid density of about 2,700 kg/m³ the casting volume is roughly 0.0093 m³. Choose a fill time of 10 s, which gives a flow rate of Q = 0.0093 / 10 ≈ 9.3 × 10⁻⁴ m³/s.

The effective head h at the choke is the metal height in the pouring basin above that point — say 0.20 m. With a discharge coefficient Cd of about 0.7 for a sand gating system, the velocity through the choke is v = Cd √(2gh) = 0.7 × √(2 × 9.81 × 0.20) ≈ 1.4 m/s.

The choke area follows: A = Q / v = 9.3 × 10⁻⁴ / 1.4 ≈ 6.7 × 10⁻⁴ m², or about 6.7 cm². As a round sprue that is a diameter of roughly 29 mm.

Now the part most gating advice skips. At 1.4 m/s the stream in the choke is fast — and fast metal entrains air and surface oxide. That is precisely why the system is expanded after the choke: with a non-pressurized ratio of sprue : runner : total ingate area = 1 : 4 : 4, the runner and the ingates carry 27 cm² each, and the velocity entering the cavity falls to roughly 0.35 m/s — below the entrainment threshold commonly quoted at about 0.5 m/s for aluminium alloys. A fast choke feeding a quiet cavity is the whole point of the arrangement; a sprue that chokes at 6.7 cm² feeding 6.7 cm² of ingate would deliver the same metal at 1.4 m/s straight at the cavity wall.

Why the sprue has to taper

Metal accelerates as it falls, so a parallel sprue cannot stay full: the stream narrows, breaks contact with the sand, and the resulting low-pressure region sucks air into the metal. The law of continuity gives the fix. With Atopvtop = Abottomvbottom and velocity from the head h at each level, the areas must satisfy Atop/Abottom = √(hbottom/htop).

For a 150 mm sprue with 50 mm of metal standing above its entry, htop = 50 mm and hbottom = 200 mm, so the ratio is √(200/50) = 2. The sprue must halve in area from top to bottom: a 29 mm top becomes about 20 mm at the base. Cast or print the sprue with that taper and it runs full; leave it parallel and every pour pulls air into the metal.

Reynolds number: useful, but not the target

Re = ρvD/μ tells you the flow is turbulent, and in liquid-metal gating it almost always is, because metal viscosity is low. For the aluminium example above, Re at the choke runs into the tens of thousands — far above the 4,000 that marks the end of pipe-flow laminar behaviour. That is worth knowing, because it explains why gating is designed around surface turbulence and air entrainment rather than around achieving laminar flow: you cannot make the stream laminar, so you keep the sprue full, avoid free-falling streams, expand the area after the choke, and keep the ingate velocity low. Reynolds number is a diagnostic that tells you why a system that looked reasonable on paper still oxidized the metal.

Pressurized or non-pressurized

SystemTypical ratio (sprue : runner : ingate)Where the choke isCharacter
Non-pressurized1 : 4 : 4Sprue baseMetal slows after the choke; quieter entry, less erosion, larger system volume. The default for aluminium and most general work.
Pressurized1 : 2 : 1 (or 1 : 3 : 3 depending on the source)IngateSystem stays full and stiff, better for keeping iron hot in thin sections; the cost is higher velocity where the metal meets sand.

Ratios are quoted inconsistently across foundry literature, so always state which element is the choke when you write the design down. The physics is the same either way: the choke sets the flow rate, and the areas downstream of it set how violently the metal arrives.

Deep Dive: Sizing Risers with the Modulus Method

A riser is not a guess, and it is not “a bit bigger than the boss”. Its size follows from one number: the modulus, M = volume ÷ cooling surface area. Two castings with the same modulus freeze in the same time regardless of shape, which is what makes the method so useful on real parts.

Chvorinov’s rule links modulus to solidification time: t = B × M², where B is a mould constant that depends on the sand, the metal and the pouring temperature. The practical consequence is that solidification time scales with the square of the modulus, so small changes in section size have a large effect on freezing order.

The design rule is short: the riser modulus must be larger than the casting modulus, and the usual factor is Mriser ≥ 1.2 × Mcasting. The riser must be the last thing to freeze, or it stops feeding while the casting is still liquid and the shrinkage appears as internal porosity.

Worked example: riser for a 200 × 200 × 40 mm plate

For the plate, volume = 0.2 × 0.2 × 0.04 = 0.0016 m³ and cooling surface = 2(0.2 × 0.2) + 4(0.2 × 0.04) = 0.112 m², so Mc = 0.0016 / 0.112 = 0.0143 m ≈ 1.43 cm.

The riser therefore needs Mr ≥ 1.2 × 1.43 = 1.71 cm. A cylindrical riser with height equal to diameter has M = D/6 (its volume is πD³/4 and its cooling surface is 1.5πD²), so the diameter must be at least 6 × 0.0171 = 0.103 m, or 103 mm. Round up to 110 mm.

The commercial consequence is worth seeing. The plate in iron weighs roughly 0.0016 × 7,200 ≈ 11.5 kg; the 110 mm riser adds about 7.5 kg and the gating perhaps 1.5 kg. The casting yield — casting weight divided by poured weight — is therefore around 55%. Riser volume is where melt cost lives: an exothermic sleeve lets the same feeder do its job at a smaller size, because the sleeve cuts the heat loss the modulus calculation assumes, and that saving repeats on every part.

Where the riser goes matters as much as its size. A riser only feeds a limited distance along a thin section — a few section thicknesses, and the exact multiple depends on the alloy and geometry — so long thin sections need either an additional riser or chills to accelerate freezing locally and push the gradient back toward the feeder. Printed moulds open another option here, since a chill or a conformal feeder neck can be designed into the mould rather than assembled into it.

Step 8: Make and Set the Cores

Cores form what the pattern cannot: internal passages, undercuts, and cavities that would otherwise require loose pieces. They are made from sand with a stronger binder system than the mold, because they are surrounded by liquid metal on most of their surface and have to survive it longer.

  • Locate them positively. Core prints should position the core without relying on the molder’s judgment, with clearance that allows assembly but not movement.
  • Support them. A core that floats or sags produces a wall thickness that varies, or a breakout. Chaplets and core supports are sized for the buoyancy of the core in liquid metal, with margin.
  • Vent them. Core gas has to escape through the core print or a vent channel, or it goes into the metal.
  • Coat them. A refractory coating limits metal penetration and burn-on and improves the surface of the internal passage.
  • Cure them fully. A partially cured core behaves like a sealed container of gas at pouring temperature.

Complex cores are exactly where printed sand has changed the calculation. A passage that would need three cores glued and wired into position can be printed into the mold itself, removing the assembly step and the core-shift risk with it. How sand 3D printing works walks through that workflow.

A-frame structural sand core for a metal casting, printed in bonded sand

Step 9: Close, Clamp and Weight the Mold

A closed mold is a pressure vessel. Liquid metal exerts a force that tries to lift the cope, and the only thing holding it down is the clamping system and whatever weight has been stacked on top.

The force is straightforward to estimate: F = ρ × g × h × A, where ρ is the metal density, h is the metal head above the cavity, and A is the plan area the metal presses on. For a cavity of 0.2 m² with 0.3 m of head in iron, that is roughly 0.2 × 0.3 × 7,000 × 9.81 ≈ 4.1 kN, or about 420 kg of lift. Weighting practice commonly applies several times the calculated force to cover the transient surge as the mold fills, and the number rises quickly with cavity area — which is why large castings are clamped, bolted or weighted with purpose-built frames rather than a few ingots.

Four things decide whether the closed mold holds:

  • Alignment. Pins, or a mark carried across the joint, keep cope and drag registered. Misalignment becomes a step on the casting that no amount of cleaning removes.
  • Sealing at the joint. A clay wash, a gasket or a well-bedded parting surface stops metal escaping along the split. Thin metal fins at the parting line are the first symptom of an inadequate seal.
  • Clamping or weighting. Sized from the calculation above, with margin, and applied so the load is distributed rather than concentrated at the flask corners.
  • Backing the sand. In flaskless or thin-walled molds, the sand body itself has to resist the pressure, which puts a premium on uniform compaction.

Step 10: Pour the Metal

Pouring is the moment the mold is tested, and the variables are temperature, rate and stream quality. All three are visible from across the shop floor, which is why experienced molders watch the pour rather than the clock.

AlliageTypical pouring temperatureNotes
Aluminium alloys700–780 °CNarrow window; oxide entrainment is the dominant risk
Magnesium alloys700–780 °CInhibited sand and strict housekeeping required
Copper alloys (brass, bronze)1,050–1,200 °CGas pickup and drossing both increase with superheat
Grey cast iron1,350–1,450 °CFluidity is good, so the window is comparatively forgiving
Fonte ductile1,380–1,450 °CMagnesium treatment adds slag and gas that the gating system must handle
Carbon and low-alloy steel1,550–1,620 °CHigh superheat stresses the sand; coatings matter more
Acier inoxydable1,550–1,650 °CLow thermal conductivity keeps the mold hot longer

Pour hot enough to fill, cold enough not to damage. Every degree of unnecessary superheat buys fluidity you may not need and pays for it in gas pickup, sand penetration, burn-on and a coarser structure. Practical discipline on the floor:

  • Keep the basin full and the sprue full. A partially filled sprue is an air pump. Pour a steady stream into the basin and let the basin feed the sprue, rather than trying to aim metal down a hole.
  • Pour at a rate matched to the cavity. The first metal into the cavity must not freeze before the last metal arrives; too fast and you erode sand and entrain air, too slow and you get misruns and cold shuts.
  • Keep slag and dross out. Skim, and use a basin that gives the slag somewhere to float rather than a direct path into the sprue.
  • Do not interrupt the pour. A pause creates a cold front inside the mold, and the casting will show it.

Step 11: Control Solidification, Then Shake Out

Solidification is where shrinkage is either fed or trapped. The objective is directional solidification: the thinnest sections freeze first, the heavy sections freeze last, and the last liquid to freeze sits in a riser that is scrapped rather than in the part.

The mold is an active participant here, not a passive container. Sand is an insulator, so a heavy section surrounded by sand stays liquid long after the thin walls have set — which is precisely why risers, chills and exothermic sleeves exist. Chills placed in the sand accelerate freezing locally and steepen the gradient; insulating sleeves and exothermic risers keep the feeder liquid longer than the casting around it.

Cooling time in sand is measured in hours, not minutes, especially for heavy sections. A mold retains heat, so a thick casting may need to sit overnight before it can be shaken out safely — knocking out too early risks cracks and distortion in a part that is still hot and weak. The safe rule is to shake out when the casting is cool enough to handle the thermal shock, which for heavier work means measuring rather than guessing.

Shakeout itself is designed, not improvised. Vibration, knockout grids and water jets break the sand away; the casting is separated from gating and risers; cores are removed from internal passages. Residual sand in a passage is a defect that reaches the customer, so complex cores deserve a defined cleaning and verification method rather than a visual check.

Step 12: Clean, Finish and Inspect

The casting leaves the mold with gating attached, sand on the surface and a skin that may be decarburised, burned-on or penetrated. Finishing removes all three, and inspection confirms that what is left meets the drawing.

  • Cut off gates, risers and feeders by abrasive wheel, saw or torch, leaving the specified stock rather than gouging into the part.
  • Remove sand and scale by shot blasting, brushing or tumbling, with internal passages cleared and checked.
  • Grind and blend the areas where gating was attached, then verify with the drawing — over-grinding is a nonconformance like any other.
  • Heat treat where the specification calls for it, before final machining.
  • Inspect to the agreed scope: dimensional report, material certificate, and non-destructive testing where the application demands it.

Inspection scope should be written into the specification before the first mold is rammed. It is much easier to add a radiographic check to a first article than to argue about acceptance criteria after the casting is made. If the part is going into a production program, the tolerance grade the drawing assumes is worth confirming against what the process can actually hold — our accuracy and tolerance guide covers what printed molds hold against conventional tooling.

When the Mold Is Printed Instead of Rammed

Everything above describes a mold built around a pattern. In a printed mold, the same elements — cavity, gating, risers, vents, cores — are produced in one operation from the CAD model, and several constraints disappear with the pattern:

ConstraintTraditional sand mold3D printed sand mold
Motif requisYes, plus a pattern plate for machine moldingNo — the mold is printed from CAD
Angle de tirageRequired on all withdrawal surfacesNot required; no withdrawal
Parting lineChosen to allow withdrawalFree — it can follow function, or be avoided
Internal passagesCores, located, supported, vented, gluedPrinted into the mold as one piece
Délai de fabrication des outillagesDays to weeks before the first moldHours to days of machine time
Modifications de conceptionPattern modification or replacementA new print
Piece cost at high volumeFalls with the amortised patternRoughly linear in mold volume

The decision is therefore not “printed versus traditional” in general, but “at what point does the pattern stop paying for itself”. For a single casting, a redesign, or a geometry that needs multiple cores, printing wins on total time and often on total cost. For a stable part made in the thousands, a pattern and a molding line remain the cheaper route. Traditional versus patternless sand casting works through the crossover with real numbers, and our sand mold and core printing service runs the printed route end to end: mold, core, casting, machining.

Multi-cavity sand mold printed in one piece from CAD, with gating and cores integrated

The Numbers to Write on the Shop Card

Mold making is a trade built on a small set of numbers that get applied again and again. This is the short version — starting points to be tuned to your alloy, sand system and drawing, not a substitute for the foundry’s own standards.

ItemStarting pointNotes
Shrinkage allowance0.8–1.0% iron; 1.5–2.0% carbon steel; 2.0–2.5% stainless; 1.0–1.3% aluminium; 1.3–1.6% copper alloysConfirm against the foundry’s shrink rule
Draft1–2° vertical faces; 3–5° deep pockets and hand moldingIncreases with pocket depth and surface roughness
Machining allowance1.5–3 mm; more on large steel castingsApplied to faces that will be machined
Metal head above the cavity50–100 mm minimumMore for thin walls and sluggish alloys
Sand wall around the cavityAt least the section thickness, never below about 40–50 mmResists metal pressure and gives ramming room
Green sand water2.5–4.5%, judged by compactability 35–45%The single most useful shop-floor control
Green compressive strengthRoughly 60–100 kPaBalanced against permeability
Mold hardness85–95 on the B scaleUniformity across the face matters more than the peak
Cope weight or clamp loadSeveral times the calculated metallostatic forceF = ρghA; calculate, then add margin

Mold-Side Defects and How to Fix Them

Most defects that get blamed on the metal are set in the mold. The useful discipline is to read the defect, find the mold variable that produced it, and fix the variable rather than the symptom.

DefectWhat you seeMold-side causeFix
Sand inclusionSand or slag trapped in the surface or a pocketTorn cavity wall on withdrawal; loose sand left after repair; eroded gatingCorrect draft and draw technique; clean and repair the cavity; reduce fill velocity
Porosité gazeuseSmooth round voids, often just below the surfaceExcess moisture or binder; low permeability; missing vents; under-cured coresCut water to the compactability target; vent deep pockets and core prints; cure cores fully
BlowholesLarger rounded cavities, frequently on the cope surfaceGas trapped against the cope face; hard, impermeable surface layerVent the cope; reduce surface compaction; check sand temperature
Metal penetration and burn-onRough, welded sand on the casting surfaceCoarse sand, low compaction, high pouring temperature, no coatingFiner or better-graded sand; coat the mold; reduce superheat
Erosion and scabbingWashed-out channels, rough laps, expansion defectsIngate aimed at a sand wall; low green strength; expansion of silica sandRedirect the ingate; raise green strength; change sand grading or add cushioning additive
Misrun and cold shutIncomplete fill, or two fronts meeting without fusingSlow fill, undersized runners, low metal or mold temperatureIncrease gating area; raise pouring temperature; shorten the fill path
Porosité de retraitInternal voids in heavy sectionsMissing or undersized riser; gradient feeding the wrong directionSize risers from the casting modulus; add chills; reposition the feeder
Mold shift or mismatchA step at the parting lineWorn pins, pattern not fixed to the plate, cope moved during closingRebuild the locating system; check pin fit; clamp before handling
Crush and swellDistorted wall, bulged face, displaced coreInsufficient compaction, weak sand, inadequate weighting, floating coreRam uniformly; verify strength targets; weight from the calculation; support the core
VeiningThin fin-like veins on internal surfacesThermal expansion of silica sand against a hot coreAdd cushioning additive; use a different base sand or a coating

Two patterns are worth internalising. First, defects that appear where metal entered the cavity usually trace back to gating: velocity, direction and fill time. Second, defects that appear where sand met metal usually trace back to sand: strength, permeability, moisture and coating. Sorting the defect into one of those two families eliminates half the possible causes before any testing starts. For a deeper treatment of causes and inspection methods, see printed mold and core defects alongside the general casting defects guide.

Reading a Defect by Where It Appears

Defect names are less useful than defect locations. Where a flaw sits on the casting tells you which part of the system produced it, which is the difference between debugging and guessing.

Where you see itMost likely familyCheck first
Cope top face, scattered round holesGas and blowholesCope venting over pockets and prints; permeability; whether the surface layer was over-compacted into an impermeable skin
At the ingate entry, rough and washedErosion and sand inclusionIngate velocity and direction; green strength; whether the ingate aims at a sand wall instead of along it
Centre of the heaviest section, internal voidsPorosité de retraitRiser modulus against the casting modulus; riser position relative to the last liquid; chill placement
Thin, long sections — incomplete or lappedMisrun and cold shutFill time; pouring temperature against the alloy liquidus; number and position of ingates
Along the parting line — a step or a thin finMould shift and run-outPin and register wear; clamping load; whether the joint was sealed before closing
On a core-formed surface — veins or rough penetrationCore expansion and penetrationCore coating; cushioning additive in the core sand; core venting and cure
Thin wall next to a heavy boss — a crackHot tearingRestraint from cores and mould; fillet radius at the transition; cooling uniformity
Everywhere — sand welded to the surfaceBurn-on and penetrationSand grading and fines content; compaction; pouring temperature above what the section needs
Bulged or distorted facesSwell and crushCompaction uniformity; sand strength; clamping or weighting against the metallostatic force

The diagnostic shortcut is the location itself. Defects clustered where metal entered the cavity point at gating — velocity, direction, fill time. Defects spread across where sand met metal point at the sand — strength, permeability, moisture, coating. Defects in the last region to freeze point at feeding. Sorting the casting into those three zones before touching a single variable removes most of the search space.

Safety Around Molds and Molten Metal

Mold making and pouring combine three hazards that each deserve respect: molten metal above 700 °C, fine silica dust, and binder chemistry. None of them are managed by experience alone.

  • No water, ever. A damp floor, a wet tool or a cold ladle in contact with liquid metal produces an explosion, not a splash. Preheat tools and ladles, and keep the pour area dry.
  • Dress for the pour. Face shield over safety glasses, aluminised or leather protection, leggings over boots, and no synthetic clothing that melts into skin.
  • Control silica dust. Sand handling, mixing, reclamation and shakeout generate respirable crystalline silica. Use extraction and enclosure at the source, and respiratory protection where dust cannot be controlled.
  • Respect binder chemistry. Furan and phenolic systems involve acids, alkalis and organic vapours. Read the safety data sheets and follow the storage, handling and ventilation requirements; curing and pouring release decomposition products that must be extracted.
  • Keep the floor clear. A pour floor with a trip hazard is a burn injury waiting for the worst possible moment.
  • Train and supervise. Only trained people should run a furnace or lead a pour, and the first pour on a new pattern is not the time to find out who has not been briefed.

Tolerances, Surface Finish and Checking the Mould

The tolerance on a casting drawing is a promise made before the mould exists. Sand casting sits in the coarse half of the ISO 8062 dimensional tolerance range, and the grade depends on the moulding process as much as on the pattern.

Biens immobiliersTypical as-cast rangeWhat moves it
Dimensional toleranceRoughly DCTG 8–13 depending on process, size and dimensionMachine moulding and rigid tooling tighten it; hand moulding, large flasks and long dimensions loosen it. The grade applies per dimension, not uniformly over the part.
Surface roughnessAbout Ra 12.5–50 µm from green sand; 6.3–25 µm with finer sand and a coatingGrain fineness, coating, compaction at the pattern face, and pouring temperature
Minimum wall thicknessA few millimetres for small parts, more as the casting growsFluidity of the alloy and how far the metal has to travel; thin walls far from an ingate are where misruns start
Draft and fillets1–2° vertical faces; internal fillet at least a third of the wallPocket depth, surface roughness, and whether the mould is hand or machine rammed

Two failure modes hide inside that word “tolerance”. The first is a part that is dimensionally correct but out of tolerance where it matters, because the drawing assumed a uniform grade and the critical dimension is the longest one on the casting — where the tolerance band is widest. The second is a part that is out of tolerance through mould damage rather than pattern error: a torn cavity wall, a shifted cope or a crushed boss produces a deviation that no pattern correction will fix. Confirming the intended CT grade per critical dimension, before tooling, is cheaper than discovering the mismatch at first article.

Because the mould is destroyed at shakeout, the only moment you can verify it is while it is open. Four checks belong in the routine, not in a post-mortem:

  • Verify the pattern before it goes into sand. Check the critical dimensions, the draft and the fillets against the drawing once, record it, and re-check after any repair to the pattern.
  • Prove registration on the first mould. Carry a witness mark across the parting line; cope-to-drag alignment is visible in one glance, and mismatch is the defect that survives every later process.
  • Map mould hardness across the face. Read several points from pattern to flask wall on the first mould of a new pattern. The average matters far less than the spread.
  • Inspect the cavity after withdrawal and before closing. Repair torn edges, confirm core prints, and confirm gating geometry. Once the cope is on, none of it can be checked again.

Then close the loop on the casting itself: a first-article dimensional report against the drawing, the material certificate for the heat, and any non-destructive testing the application requires. If the first article passes, the same mould routine becomes the process record for the part.

Foire aux questions

How do you make a casting mold for metal step by step?

Make a pattern oversized for shrinkage and drafted for withdrawal, then split the mold into cope and drag at a parting line. Mix and ram bonded sand to a controlled strength and permeability, withdraw the pattern, repair the cavity, cut sprue, runner, ingates, risers and vents, set any cores, then close, clamp and weight the mold before pouring. Break the sand away after solidification and finish the casting.

What is the difference between a pattern and a mold?

The pattern is the reusable physical replica of the part, carrying shrinkage, draft and machining allowances, and it is withdrawn before pouring. The mold is the cavity formed in bonded sand by that pattern, plus the gating, risers, vents and cores it contains. The pattern is tooling and survives many cycles; the mold is expendable and is destroyed to remove each casting.

What sand is used to make a casting mold for metal?

Silica sand is the base for most molds, bonded either with bentonite clay and water for green sand or with a resin system for no-bake and shell molding. Chromite, zircon and ceramic sands are used where higher refractoriness or lower thermal expansion is needed, for example against heavy steel sections or on critical core surfaces.

How much shrinkage allowance should a pattern carry?

As a starting point: about 0.8 to 1.0 percent for grey and ductile iron, 1.5 to 2.0 percent for carbon steel, 2.0 to 2.5 percent for stainless steel, 1.0 to 1.3 percent for aluminium alloys and 1.3 to 1.6 percent for copper alloys. Confirm the figure against the foundry’s own shrink rule, because restraint from the mold and cores changes the effective contraction.

What draft angle does a sand casting pattern need?

One to two degrees is typical on vertical withdrawal faces, rising to three to five degrees on deep pockets, rough surfaces and hand-rammed molds. Draft must be added in the direction of withdrawal on every surface the pattern leaves behind, and insufficient draft is one of the most common causes of a torn cavity wall and sand inclusions.

Can you make a casting mold for metal without a pattern?

Yes. A binder jetting sand printer builds the mold and cores directly from the CAD model, so no pattern is needed and no draft is required because nothing is withdrawn. It suits single parts, design iteration and geometry that would otherwise need several assembled cores, while high-volume stable parts usually stay cheaper with a pattern and a molding line.

Why do casting molds need vents?

Vents give air and binder gas somewhere to go other than through the liquid metal. As metal enters the cavity it displaces air, and the heat decomposes binders and vaporises moisture, so gas is generated throughout the pour. Without vents over deep pockets, blind bores and core prints, that gas is trapped and appears as blowholes or gas porosity in the casting.

How long does it take to make a sand casting mold?

Ramming and closing a small green sand mold takes minutes on a molding machine and a good deal longer by hand for a large flask, but the mold itself is rarely the bottleneck. The lead time sits in the pattern, which takes days to weeks to make and modify. A printed mold removes that step and is typically ready in hours to a few days of machine time depending on size and quantity.

Articles connexes

If you are weighing a pattern against a printed mold for a specific part, send us the drawing and the quantity. We will tell you which route is cheaper at that volume, and what the mold design would look like either way — demander un devis or see the Gamme d'imprimantes 3D à sable if the mold making is moving in house.

Jeff

Forte de plus de 30 ans d’expérience dans la fabrication additive et disposant de plus de 25 imprimantes 3D à sable en service dans nos propres locaux, l’équipe d’ingénieurs de 3DPTEK s’appuie sur une expérience pratique de la production, et non sur la théorie. Nous utilisons nous-mêmes les produits que nous vendons.