Receiving Inspection for Numbers: Auditing Supplier Material Cards Before They Reach the Solver

The card arrives with more authority than evidence

Most US engineering organizations have a receiving inspection process for purchased hardware. A bracket arrives, and someone checks the certificate of conformance, the revision, the finish, and the dimensions that matter. Data rarely gets the same treatment. A supplier emails a PDF and a spreadsheet, an engineer maps the elastic modulus, yield strength, thermal expansion coefficient, and fatigue curve into a solver material card, and the model proceeds. The click is fast. The consequences are not.

A material card is not a neutral description of a substance. It is a set of test conditions, specimen geometries, statistical bases, and process assumptions compressed into a handful of fields. Once those fields enter the solver, they become the physics of the program. The model does not know whether the modulus came from a production lot, a handbook, a qualification coupon, or a supplier’s optimistic summary. It simply computes.

The practical answer is to treat the material card as a contract: a controlled technical document that the supplier stands behind and the engineering team accepts on defined terms. That means an intake audit before the card reaches the solver, not after the hardware disagrees with the model.

Provenance questions that precede any mapping

The first gate is provenance. Before anyone maps a number into a material card, the supplier should answer a short, unambiguous questionnaire. The goal is not to interrogate the supplier’s integrity. It is to expose the difference between a minimum guaranteed property and a typical value, and between a production material and a laboratory curiosity.

A workable checklist includes:

• Grade and specification. Which material specification and revision govern the card? Is it an ASTM, ISO, SAE, or internal supplier specification? Has the specification changed since the data was generated?
• Source of the data. Was the property measured on a production lot, a qualification lot, a development coupon, or copied from a public database? Who ran the test, and when?
• Test standard and specimen. Which test method was used? What was the specimen orientation relative to the manufacturing process? What were the gauge length, width, and strain rate?
• Condition and environment. At what temperature, humidity, and conditioning state were the properties measured? For thermal expansion, what was the reference temperature and the tested range?
• Definition of the reported value. Is the modulus secant or tangent? Is yield reported at 0.2 percent offset or at a different strain? Is the fatigue curve based on strain-controlled or stress-controlled testing, and what R ratio, surface finish, and notch condition apply?
• Statistical basis. Is the number a mean, a minus-three-sigma value, a minimum guaranteed value, or a single coupon result? How many specimens were tested?

If the supplier cannot answer these questions, the card should not be treated as validated. It may still be usable for early-stage design, but it should carry an explicit status flag in the model and in the report. The danger is not using an uncertain number. The danger is forgetting that it was uncertain.

Validation coupons: small tests that catch large errors

Provenance review catches documentation gaps. It does not catch a card that is well documented and still wrong for the application. That is why a second gate is useful: a small set of validation coupons that compare the supplier’s card against physical behavior under conditions close to the intended use.

These coupons are not calibration specimens. They are acceptance tests. The purpose is to decide whether the card is fit to enter the model, not to tune the model until it matches a single batch of material.

Three coupon types cover most of the risk:

1. Tension coupon at service temperature. Measure modulus and yield strength at room temperature and at the upper service temperature. A modulus deviation of more than about five percent, or a yield deviation of more than ten percent, is a prompt to ask questions. The shape of the stress-strain curve also matters; a card that reproduces the yield point but not the hardening slope may still mislead a nonlinear analysis.
2. Thermal expansion bar. Measure expansion over the intended temperature range and compare it with the card. This is where reference-temperature errors hide. A card referenced to 20 degrees Celsius used in a model referenced to 0 degrees Celsius can produce a thermal growth error that looks like a boundary-condition problem.
3. Modal plate or beam. A free-free modal test gives a quick check on dynamic modulus and damping. It is especially useful for polymers, composites, and castings where the supplier’s static data may not represent the delivered part. Compare the first few modes and treat a consistent offset as a reason to revisit the card, not as a reason to adjust the solver.

For critical applications, fatigue data deserves its own review. Fatigue curves are expensive to generate and easy to misapply. A curve that does not state the R ratio, surface condition, and statistical basis is not a curve. It is a sketch.

Contract language that gives the numbers an owner

The third gate is contractual. A material card that arrives without an owner is a rumor with units. The purchase order or supplier quality agreement should define the card as a deliverable with a revision, a provenance record, and an acceptance criterion.

Useful terms include a warranty that the data represents production material within stated lot-to-lot tolerances, a notification requirement for process or formulation changes, and an obligation to investigate when validation coupons fall outside agreed limits. The agreement should also specify the data format: machine-readable values, explicit units, test-standard references, and a revision identifier that can be linked to the part and the model.

Legal review is essential, and this is not legal advice. The engineering point is simpler: the supplier should know that the numbers are being used for design decisions and that a silent change to the material card is a change to the product definition. A click-through portal that updates a modulus without notification is not a convenience. It is a configuration-control failure.

Configuration control after acceptance

Once a card passes intake, give it an identity. A card identifier, revision, and hash or checksum should travel with the CAD material assignment and the CAE include file. If the supplier issues a new card, it gets a new revision and a new validation status. The old model does not silently inherit the new numbers.

This discipline is unglamorous, but it is cheaper than the alternative. A missed thermal expansion reference temperature, an unstated heat treatment, or a fatigue curve without a surface condition can survive months of modeling and appear only when a prototype is on a shaker table or in a thermal chamber. At that point, the cost is not a data review. It is a redesign, a schedule slip, and a meeting in which someone asks why the solver believed a number that no one owned.

Receiving inspection for material data is not bureaucracy. It is the point where simulation stops treating supplier numbers as defaults and starts treating them as engineering commitments.