A real case
A ductile-iron housing, 180 a year in green sand, and the pattern shop wants the go-ahead Friday. Here is what a foundry puts the model through first: casting findings mapped onto the part with the rule that caught each one, parting line and flask fit for the sand job, and a pour run inside the quote.
The situation
A STEP file lands on a Tuesday afternoon. Ductile-iron pump housing, EN-GJS-500-7, 180 a year, green sand on your own flask line. The customer wants a number by Friday, and the pattern shop wants the go-ahead the same day to hold a slot next month. The volute wall runs 8 mm and then meets the discharge flange in a junction heavy enough that half the shop would have an opinion about it — and normally nobody is asked, because there is no time to ask. The first proper conversation about that junction happens at first article, with a pattern already cut and a delivery date already promised. This is the three days before that, and what you can actually get through in them.
Step by step
Drop the model in and manufacturability findings come back on the upload itself, before a process has been chosen. You are not committing to casting, to green sand, or to a quantity to get the first look — you are finding out what is in the geometry while the pattern order is still an email nobody has sent.
Thin walls, hot spots, undercuts and missing draft are flagged and mapped directly onto the 3D model, each carrying the rule that caught it. The junction behind the discharge flange either comes back as a hot spot with a named rule or it does not, and either way you now have something to point at on a screen rather than a recollection of a similar housing from 2023.
X, Y and Z pull axes are scored for undercut volume, draft rework and projected area, and the best parting line is picked automatically. On the sand path, parting line, core geometry and flask fit compute for you, and the part is checked against the flask sizes, platen and handling limits of the machine classes you actually run — with a named reason when it exceeds one, rather than a shrug at the moulding line.
Gating and risers are generated off the geometry, then GPU lattice-Boltzmann fill returns fill order, the last regions to fill and trapped-air risk, and transient solidification returns freeze order and which risers actually feed. Both run as part of the casting job, not as a specialist's booked week. Casting-yield ranking for orientation and riser placement is scored against your own recorded jobs, and refuses to be trusted automatically if it is not beating a simple baseline. Every panel mounts against the same job as the upload, so none of this is a second upload or a second explanation of the part.
Two answers exist and they are different answers. A feeding problem you can rig around is a gating change you price and quote. A junction that will never feed as drawn is a geometry conversation with the customer, and the finding, its rule and the freeze order are what you send with it. Every result names its confidence basis — a published prior, a calibration off your own actuals, a deterministic rule, or a flag that it needs a human — so you know how hard to lean on it before the pattern order goes out.
What you get
By Thursday you are holding the same model the customer sent, marked with the findings that survive a second look, a parting line and flask fit you did not work out by hand, a fill and freeze order for the rigging you would actually have used, and a yield figure scored against your own poured jobs instead of the last housing anyone remembers. Friday's email is either a price with a gating note attached or a specific question about one junction — and both of those are cheaper than the version of this week where the pattern goes ahead and the question waits for first article. Partimation does not redraw the part for you and does not promise the pour: it is decision support, and the point is that the argument now happens over a rule and a freeze order rather than over whose memory of a similar job is better.
Already shipped
Built into MFGDFM.
Built into MFGDFM.
Built into MoldDFM.
Built into MFGDFM.
Built into CostDFM.
Built into SimDFM.
Built into SimDFM.
Built into SimDFM.
Built into CostDFM.
FAQ
Thin walls, hot spots, undercuts and missing draft come back on the upload itself, mapped onto the 3D model with the rule that caught each one, before a process has been chosen. On the same job you get the pull-axis study and parting line, core geometry and flask fit for a sand part, gating and risers generated off the geometry, and a GPU fill and solidification run. None of that is a specialist's booked week, which is the only reason it fits between a Tuesday file and a Friday answer.
Missing draft is flagged as a finding mapped onto the 3D part, carrying the rule that caught it, and the pull axis it is measured against is the one the parting-line study picked after scoring X, Y and Z for undercut volume, draft rework and projected area. A customer arguing about draft is usually arguing about pull direction, and both halves are on the screen rather than in an estimator's head.
Transient solidification returns freeze order and which risers actually feed, and the GPU lattice-Boltzmann fill run returns fill order, the last regions to fill and where air gets trapped — both on the rigging generated from the part's own geometry, on the same job as the upload. A junction that freezes off before its riser shows up as a freeze order you can read, months before first article would have told you the same thing at the cost of a pattern.
Parting line, core geometry and flask fit compute automatically on the sand path, and the part is checked against the flask sizes, platen and handling limits of the machine classes your foundry actually runs, with a named reason when it exceeds one. Where a cut pattern is not the answer at all, printed sand molds and cores have their own should-cost path rather than being priced as a pattern job with a discount on it.
Where this fits
You pour metal, and the expensive surprises land late: a section that won't feed, a pattern price someone recalled, a yield nobody scored. Partimation screens the casting before tooling, runs fill and solidification physics on every casting job, prices the pattern into the piece, and quotes on your foundry's rates.
Undercuts, thin walls, hot spots and missing draft usually surface at first article, when the pattern or the die already exists. Partimation flags them on the uploaded part with the rule that caught each one, and runs GPU fill and solidification physics on every casting job — before anyone commits steel.
Get started
Send a model or a drawing and book a walkthrough — or start free and run one through yourself.
Book a demo
Tell us about your shop and we'll set up a walkthrough on a real part.