Making a mold for metal casting means packing bonded sand around a pattern in two flask halves, cutting the passages the metal will run through, setting any cores, and clamping the halves shut. The twelve steps below are the sequence a production sand foundry runs, from a finished pattern to a mold ready to pour.
Each step fixes something you will later measure on the casting: the tolerance it can hold, the machining stock it carries, where the parting line lands, and which defects it is prone to. Melting, pouring, shakeout and cleaning belong to a separate sequence.
1. Choose the Type of Casting Mold
The mold type follows the casting’s weight and which of its features have to hold tolerance. Below about 50 lbs, green sand’s dimensional capability is close to that of the other sand processes; above 200 lbs, no-bake typically produces tighter tolerances than green sand; shell molding is the most dimensionally capable of the three but is limited in casting size.
What a sand mold can hold, in ISO casting tolerance grades:
| Tolerance target | Long production series | Short series |
|---|---|---|
| Within process capability, but not appropriate for every casting type or sand process | CT 8–10 | CT 9–11 |
| Appropriate for most casting types and sand molding processes | CT 10–12 | CT 11–13 |
| All sand processes fully capable; most appropriate for large castings | CT 12–14 | CT 13–15 |
Grades mean nothing until they are converted: at CT 12, a 400 mm feature carries 9 mm of total spread, ±4.5 mm about nominal. Investment casting sits at CT 5–7, which is why a feature holding a few tenths of a millimetre either gets machined or gets a different process.
Read the differences between the three sand processes as smaller than they look. The spread in dimensional capability between molding processes is less than the variation from one foundry to another running the same process, so a supplier’s demonstrated capability on a part like yours outranks the process name on the quote.
2. Make or Obtain a Pattern
The pattern is the finished part plus everything the process adds: draft on the faces perpendicular to the parting line, machining stock where you intend to cut, and a contraction allowance nobody can guarantee until a casting is measured.
| Molding process | Most features | Deep pockets |
|---|---|---|
| Green sand, manual | 1.5° | 2.0° |
| Green sand, automated | 1.0° | 1.5° |
| No-bake and shell | 1.0° | 1.5° |
Meet or exceed those angles on every surface perpendicular to the parting line, and give interior surfaces in green sand more draft than exterior ones. What the angle costs in material is:
DA = L tan θ
So a 200 mm deep pocket drafted at 2° gains about 7 mm of wall at its opening — material the designer either accepts or pays to machine away.
Do not expect the contraction allowance to land where the patternmaker predicted. SFSA’s own position is that there is high risk in assuming the shrink built into pattern equipment will put final dimensions inside tolerance, which is why a first article is poured and measured before production dimensions are agreed.
Pattern construction shows up in the measurements. Separate cope and drag patterns instead of a match plate add 0.03 in. (0.8 mm) of 6σ dimensional variability, and a pattern in poor condition adds 0.04 in. (1 mm) over one in good condition.
3. Prepare the Flask
Pick a flask that leaves enough rammed sand around the pattern to contain the metal and no more, because every inch of cope or drag height adds about 0.003 in. of 6σ dimensional variability. An oversized flask is not free depth; it is paid for in spread.
Then check what locates it. The pins and bushings that register cope to drag are the datum every later dimension inherits, and a worn set arrives on the casting as mismatch at the parting line. Bars and lifters that hold sand up in the cope get inspected at the same time.
4. Place the Pattern in the Drag
Set the pattern in the drag with the surfaces you intend to machine facing down. Cope surfaces need more machining stock than drag surfaces or side walls — 6 mm is the minimum on a cope surface of a large casting, and less is not recommended at all.
That penalty is physical. The cope face is the top of the pour, where sand and slag collect and where the mold face carries the metal’s full head for the whole fill, so it sits lower and less predictably than the drag face.
5. Add and Compact the Molding Sand
Fill the drag in layers and ram to a measured compactability rather than to feel. For synthetic molding sands the working window is 35–45%, and moisture is what gets adjusted to hold it there.
Compaction trades directly against accuracy: each unit increase in the compactability number adds about 0.02 in. (0.5 mm) of 6σ dimensional variability. Softer sand fills and reproduces detail; harder sand resists the metal and holds size. Every point given to fill quality is paid back in dimensional spread.
6. Turn the Drag Over
Roll the drag 180° so the pattern’s parting face is up, then lift the molding board away. What this produces is the parting surface the entire cope will be built on.
Level and repair that surface before anything else touches it, since sand pulled loose at the pattern’s edge becomes flash or a mismatch line on the casting. Dust it with parting compound so the cope sand releases cleanly, then set the cope half of the pattern and bring the cope flask down on the alignment pins.
7. Form the Sprue and Riser
Set the sprue and riser patterns before the cope is filled: the sprue tapers from the pouring basin down, and each riser is sized by the section it has to feed.
A riser must solidify after the section it feeds, so its modulus — volume divided by cooling surface area — needs to be at least 1.2 times that of the section. Sizing by modulus rather than by volume is what stops a riser freezing off while the casting is still drawing on it.
Riser count comes from feeding distance. In steel plate sections, where width is at least three times thickness, one riser feeds about 4.5 T of length beyond itself, so a 2 in. thick carbon steel plate is fed radiographically sound for roughly 9 in. past the riser. Treat that as the conservative figure: simulation-derived feeding rules give longer distances in most casting situations, which is how a foundry running solidification simulation holds soundness with fewer risers and better yield.
Taper matters as much as size. A sprue that does not narrow toward its base runs partly empty, and the falling stream pulls air in with it — a plunging stream can be at least half air by volume — which lands in the casting as entrained bubbles and folded oxide films.
8. Fill and Ram the Cope
Fill the cope in the same layered way as the drag and ram it to the same compactability. A hard cope over a soft drag moves differently when the metal arrives, and that mismatch shows at the parting line.
Ram evenly around the sprue and riser patterns rather than driving down onto them; sand left loose there erodes when the stream hits it and travels into the cavity. Vent over heavy sections, strike the top off level so clamps or weights bear evenly, and keep the cope no deeper than the job needs.
9. Separate the Mold Halves
Lift the cope straight up off the pins, draw the pattern from each half, and repair the surfaces the pattern damaged on the way out. Rapping to loosen a pattern enlarges the cavity, so a pattern that needs heavy rapping is a draft problem rather than a technique problem.
The parting line itself carries a dimensional penalty, and it depends on the process:
| Molding process | Added 6σ variability across the parting line |
|---|---|
| Green sand | None measured |
| No-bake | 0.040 in. (1 mm) |
| Shell | 0.008 in. (0.2 mm) |
So the process that holds tighter tolerances overall on a large casting is the worst of the three on any feature spanning cope and drag. Keep close-tolerance features inside one half of the mold; where a dimension has to cross, expect it to be the loosest one on the part.
10. Cut the Runner and Gates
Cut the runner and ingates to the cross-section that keeps metal entering the cavity at or below about 0.5 m/s. Above that the stream jets, folds its own oxide skin into the metal, and drags mold gas along with it. Gates are sized by that velocity rather than by a fixed sprue-to-runner-to-gate ratio.
Keep runners slim and rectangular rather than deep, since the slimmer the section the more laminar the flow. Sand inclusions in a finished casting are a reliable sign that the gating ran turbulent: air pushed back and forth through the mold wall oxidises the binder, and the sand comes away.
11. Install Cores if Needed
Set cores only where the pattern halves cannot form the cavity, and check that the prints can hold what you set. A core displaces liquid metal and floats, and the print has to carry that lift — which is why print length is matched to core strength rather than just to what fits on the drawing.
Cores also cost tolerance. A feature formed between a core and the mold inherits the core’s placement and the clearance between core and print, so it carries more variables than a feature formed inside a single piece of tooling. Where a dimension is critical, form both of its surfaces in the same tool.
12. Close the Mold
Close the cope onto the drag on the pins and clamp it. Clamped molds run about 0.02 in. (0.5 mm) tighter in 6σ dimensional variability than weighted ones, which makes clamping the cheapest accuracy improvement available on the molding floor.
What the clamps resist is the metal’s pressure acting on the whole upper mold face. That lift is the casting’s projected area at the parting multiplied by the head above it and the density of the liquid metal, so it does not scale with casting weight — a wide, thin part under a tall sprue lifts far harder than its mass suggests, and every core adds its own buoyancy on top.
Before the halves come together, blow the cavity clean, confirm each core is seated on its prints, and check that nothing has dropped into the downsprue. It is the last time anyone sees the inside of the mold.