How to Reverse Engineer a Casting for an Obsolete Part

To reverse engineer a casting for an obsolete part, rebuild its full definition from a sample, then prove the first new casting in the machine. That definition covers geometry, alloy, machining stock and tolerances. A scan supplies only the geometry, and it also copies the wear of one used part.

These steps assume a part you own that runs in your own equipment. The legal points are U.S. federal law, and selling copies to other owners is a separate question. The alloy and allowance figures cover the iron and steel we pour.

Step 1: Confirm the Part Can Be Copied

Reverse engineer the part only if it passes two checks: no patent blocks the copy, and whoever approves the part will accept one. Run both before you pay for any measuring.

Whether It Is Legal to Copy the Part

Copying an obsolete part from a sample you lawfully hold is legal in the U.S. when no patent is in force on it. The U.S. Supreme Court has called reproducing a functional attribute of an unpatented item legitimate competitive activity.

Check the patent first. A live patent blocks the copy, and the patent statute has no exception for your own use.

A utility patent runs 20 years from its filing date, and term adjustments can extend that. A design patent runs 15 years from grant, or 14 years if it was filed before May 13, 2015. The USPTO does not calculate expiration dates, so search by the OEM’s name and work out the date from the patent itself.

Replacing a worn unpatented part in a machine you own counts as repair, even when the machine is patented. That right ends if you effectively rebuild the patented machine.

Two more limits apply to what you copy from:

  • OEM drawings: Make your own drawing from the part. Copyright does not reach a part’s functional shape, but technical drawings are protected. Using a drawing you hold under a duty of secrecy is trade secret misappropriation.
  • Logos and names: Leave the OEM’s registered mark off the pattern, because a cast-in mark on a copy can infringe the trademark.

A live or doubtful patent needs a patent attorney’s opinion on that specific part.

When a Copy Will Not Be Accepted

A copy alone will not be accepted for castings in regulated service. Each case below can be replaced only through its own approval route. Find that route before you measure anything.

  • Pressure-retaining parts: Under National Board interpretations, the R stamp holder verifies a replacement part against the original code of construction or an alternative examination or test. This route covers equipment repaired under the National Board Inspection Code, so check the nameplate and which code built the unit.
  • Plants under OSHA process safety management: A copy escapes management of change only as a replacement in kind, meaning it satisfies the design specification.
  • Drinking water parts: Most states require water treatment and distribution components to comply with or be certified to NSF/ANSI/CAN 61, so the copy needs that compliance too.
  • Aircraft and nuclear safety-related parts: Both have their own federal routes, under 14 CFR 21.9 and, for nuclear, the Appendix B quality program or commercial grade dedication in 10 CFR 21.3.

Step 2: Gather the Sample and Its Mating Parts

Collect everything that shows how the part worked before anyone measures it.

  • The best sample available: A new, unused spare beats a part from service.
  • More than one sample: The Defense Logistics Agency’s source approval guide recommends measuring at least three parts.
  • Mating parts: These set the tolerances and confirm the clearances.
  • Configuration: Record the assembly the part goes into, and confirm the sample is the right version for your machine.
  • Service history: Note load, speed, direction of rotation and how the part failed.
  • Surviving paperwork: Keep old drawings, parts manuals and repair records.
  • Photographs: Take them before disassembly to record the part’s position in the machine.

What to Do With a Broken Sample or No Sample

A broken sample still works if you keep every piece. Scan the pieces separately and align them in the model. Do not weld the part back together first.

Have a metallurgist read the fracture before you copy anything. A copy of a part that failed from a weak section or the wrong alloy will fail the same way, so correct that cause in the definition.

With no sample, you are designing a new part to fit an old machine. Define it from the mating parts: bolt patterns, bores, shaft sizes and the space it has to fit. Add whatever drawings, manuals and photos survive, or borrow a sample from another owner of the same machine.

Step 3: Measure the Casting

Capture a casting in two passes: scan the as-cast surfaces for the overall shape, then measure the machined features with a CMM or hand gauges.

Which Features to Scan and Which to Measure by Hand

Scan what the mold formed, and gauge what the machine shop cut.

  • Scan: Outer as-cast faces, ribs, bosses, fillets, blends and the overall envelope.
  • CMM or hand gauges: Bores, bearing and shaft fits, bolt patterns, sealing faces and threads.

How to Capture Internal Passages a Scanner Cannot See

Use industrial CT if X-rays can get through the casting. If they cannot, section a spare sample and check walls with an ultrasonic gauge.

Ask the CT lab for a measurement uncertainty.

Without CT, cut a spare sample open and accept that it is destroyed. If you have only one sample, finish every outside measurement before you cut it. A core has to be removed through openings in the casting, so the visible openings show you where each passage starts and ends.

An ultrasonic thickness gauge needs two wall surfaces that are roughly parallel or concentric, and rough sand-cast faces impair the reading. Calibrate it on a reference of known thickness and similar metallurgy.

Measure several points around any cored wall. Core shift can leave a wall too thin on one side and too thick on the other.

Step 4: Identify the Alloy

Identify the alloy in stages, and expect a family and a likely specification at the end, not a proven grade. Choose the grade from the loads and service the part has to survive, within that family, agree it with the foundry and write it on the new drawing. Have the new casting qualify to it.

Start by separating iron from steel, then use these tests:

  • Polished section: Shows the graphite form under a microscope, which separates gray iron from ductile iron.
  • XRF: Reads alloying elements quickly but does not read carbon.
  • Spark OES: Reads carbon in steel. It needs a clean, flat surface and leaves a small burn mark. On gray and ductile iron it can read carbon low.
  • Combustion analysis: Measures carbon from 250 mg to 1000 mg of drillings, so the sample survives.
  • Hardness: Means different things in gray and ductile iron.

Gray Iron

Flake graphite in a polished section identifies gray iron. Chemistry will not give you the class, because ASTM A48 subordinates composition to test-bar tensile strength.

Hardness is a good indicator within one class but not between classes. It also reads lower in thick sections, so compare readings taken at similar sections. A48 groups test bars by the controlling wall section, so give the foundry that wall thickness.

Ductile Iron

Spheroidal graphite identifies ductile iron. The common grades differ mainly in matrix structure, and hardness correlates well with tensile properties. Matrix and hardness together point to a likely grade.

ASTM A536 grades 65-45-12 and 80-55-06 are as-cast, while 60-40-18, 100-70-03 and 120-90-02 are heat treated. If the sample points to one of those three, the definition must state the heat treatment.

Carbon Steel

Carbon steel needs a carbon reading, so use spark OES or combustion, not XRF alone. ASTM A27 grades set limits on carbon, manganese, silicon, sulfur and phosphorus.

Every A27 grade except N-1 and N-2 must also meet tensile properties. Chemistry on the sample leaves those unproven, so set them on the new order.

Stainless Steel

XRF identifies the stainless steel family, but confirming a grade such as CF8M takes a carbon reading. In corrosion-resistant cast grades, the number after the two letters is maximum carbon: CF8M allows 0.08% carbon maximum. Heat-resistant H grades use that number differently, so run OES or combustion on those too.

Step 5: Separate Wear From Design Intent

Treat every dimension on a surface that rubbed, sealed or took load as wear until the mating part and the service history say otherwise. Dimensions on untouched surfaces can stand as measured.

Inspect the sample before you trust the numbers. Look for wear, damage, corrosion and old repairs such as weld or sleeves. Remove each of these from the model.

Recover the nominal size of each worn feature from what it has to fit:

  • Mating part: A bore that takes a bearing or shaft gets its size from that part and the fit it needs.
  • Standard sizes: A worn dimension often sits close to a round inch or metric size, or to a catalog bearing, seal or thread size.
  • Unworn areas: The same bore or face often survives outside the contact path.
  • Other samples: The least worn sample shows which way each dimension moved.

If the worn mating part stays in service, match its worn surface, because the new casting has to run against that part.

A surviving drawing is not automatically right. Casting dimensions are often changed in tooling without the drawing being updated. Where the two disagree on an unworn feature, trust the hardware.

Step 6: Turn the Model Into a Casting Definition

A scan-based model becomes a casting definition when each surface is marked as-cast or machined, stock is added to the machined faces, and tolerances go only where they matter. Shrinkage stays out of the model.

Sort the sample’s surfaces first. As-cast surfaces show draft, fillets, rounds and parting line or sprue marks, while machined surfaces show sharp corners and tool marks.

A sample’s as-cast measurements are already casting sizes, so the definition keeps them. The foundry adds the pattern allowance on the pattern and coreboxes. If you use the sample itself as a pattern, the casting comes out undersize by about that allowance.

Model machined surfaces at finished size, then add machining stock. A diameter, or a feature machined on both sides, takes that stock twice.

On steel castings, leave enough stock that the first cut stays at least 1.5 mm (1/16 in) below the cast surface. Avoid less than 6 mm (0.25 in) on the cope surfaces of a large casting. On cast iron, stock typically runs from 0.12 in on parts up to 12 in to 0.25 in on parts of 20 in to 40 in.

Stock goes on first, then the foundry applies shrinkage to the result, then draft goes on top. Here is the sequence for a 500 mm carbon steel part with 6 mm of stock per end and a 2.08% allowance:

Casting length = 500 mm + 6 mm + 6 mm = 512 mm

Pattern length = 512 mm × 1.0208 = 522.6 mm

The 2.08% carbon steel figure is only a rule of thumb, so leave the shrinkage to the foundry that pours the part.

As-cast faces in a scan already carry their draft. New draft matters on faces the model rebuilds, and wherever the foundry moves the parting line. Typical draft on steel castings is 1.0° to 1.5° on most features.

Put tight as-cast tolerances only on the few dimensions that need them. Agree the rest with the foundry against its process. The drawing should also state the alloy specification, any heat treatment, the machined surfaces and the datums.

Step 7: Prove the First Casting Against the Part

Prove the first casting against the agreed drawing and in the machine, because a match to the scan is not acceptance on its own.

Have the first casting made with the identical foundry practice planned for production. Then check it in this order:

  1. Stock layout: Lay out the casting from the agreed datums and confirm stock on every machined surface. A simple shape may need only a rule, while a complicated one may need target points and scribed lines.
  2. Surface and internal soundness: Examine the surfaces, then run the NDT named on the order, with its standard and acceptance level.
  3. Internal passages: Section one casting if there is no other way to check them.
  4. Fit: Machine the casting and assemble it with the real mating parts to confirm clearances.
  5. Function: Run it in the machine or test it against the part it replaces. Failure-limit tests are the most convincing proof of equivalence, though a simple, low-risk part may not need one.

If the first casting fails, revise the foundry practice or the design, then resample. When it passes, put every change into the definition, so the next order repeats the approved casting.