Independent investment casting knowledge for global buyersB2B Network
Investment Casting Hub
Choosing a Foundry

How to Audit a Foundry for Dimensional Tolerance Consistency

Published 13 min read

Engineer using a CMM machine to measure metal casting part
Quick answer

To audit a foundry for dimensional tolerance consistency, request historical CMM data, inspect core shell preparation records, and validate their quality control procedures. Verify material thermal expansion coefficients and confirm the foundry uses statistical process control to maintain tight dimensional specifications across production runs.

Key takeaways
  • Request historical CMM data to prove the foundry can hold tolerance consistently over time.
  • Inspect core shell preparation and assembly records to verify process control.
  • Validate the foundry's statistical process control methods and quality control documentation.
  • Confirm material thermal expansion coefficients are accounted for in their measurement procedures.
  • Run a final verification batch to confirm audit findings hold in production.

Why Tolerance Consistency Matters in Investment Casting

Tight dimensional tolerances in investment casting depend on multiple variables working together. The foundry must control core shell thickness, mold preparation, pouring temperature, and cooling rates. A single variable drifting outside its control limits can shift the final part geometry. The core acts as the skeleton of the part, and the shell determines the surface finish and the dimensional accuracy of the internal cavity. If the shell thickness varies across a batch, the metal fills the cavity to different depths. The resulting part may show a uniform size on average, but individual parts will drift toward the high or low side of the specification.

Engineers often see the first sample part pass inspection. That does not mean the foundry can hold the tolerance across a production run. Dimensional tolerance consistency requires verification of the entire process chain, not just the final measurement. A foundry can produce a perfect sample by adjusting the mold or core for that single piece. Production, however, involves hundreds of cycles. Each cycle involves mixing the investment, applying it to the core, baking it, pouring the metal, and cooling the mold. Small variations accumulate. A foundry that treats tolerance as a final inspection target rather than a process control target will eventually fail.

This audit approach targets the data points that predict whether a foundry will deliver parts within your specifications repeatedly. It focuses on the variables that drive dimensional accuracy and the controls that monitor them over time.

Prerequisites: What to Request Before the Audit

Before contacting the foundry, gather your drawing package and define the audit scope. You need the print with tolerance callouts, the material specification, and the quantity. Include any critical feature dimensions and the acceptable process capability index you expect. For example, if you require a flatness tolerance of 0.05 mm over a 100 mm face, specify that. If you require a bore diameter of 50.00 mm with a tolerance of ±0.02 mm, specify that. The foundry needs to know exactly what you are measuring and what level of capability you require.

Request the following documents from the foundry:

  • Historical CMM reports for similar parts from the last production run
  • Core shell preparation and assembly records for a representative batch
  • Quality control plan or process control plan for the specific casting operation
  • Material thermal expansion coefficient documentation for the alloy used
  • Statistical process control charts for critical dimensions from recent production

These documents form the baseline for the audit. Without them, you are auditing only the present, not the capability. A foundry may have excellent equipment but poor record keeping. A foundry may have good records but poor process control. The documents reveal the intersection of both. If the foundry cannot provide historical CMM reports, ask why. Lack of data is a red flag. If they cannot provide core shell records, ask how they control shell thickness. In either case, the gap in documentation suggests a gap in control.

Step 1: Verify the Foundry’s Measurement Infrastructure

Ask the foundry where and how they measure parts. A foundry that relies on hand gauges for critical dimensions will struggle with tight tolerance work. Coordinate measuring machine (CMM) capability is the standard for dimensional tolerance verification in investment casting. CMMs use a stylus or laser probe to touch multiple points on the part and calculate the geometry mathematically. This method is reproducible and can measure complex surfaces that hand gauges cannot reach.

Confirm the CMM software version and the calibration schedule. Ask how often the probe balls and reference artifacts are checked. A foundry that calibrates monthly is more reliable than one that calibrates quarterly for tight tolerance work. Calibration ensures that the machine’s scale is accurate. If the CMM is not calibrated, every measurement is off by the same amount. That bias may not trigger a spec violation, but it will skew the process capability analysis. If the foundry uses a CMM that has not been calibrated in six months, the data is suspect.

Ask about the measurement environment. CMMs are sensitive to temperature changes. The foundry should maintain a stable temperature in the metrology lab, typically around 20 degrees Celsius. If the lab is near a loading dock or an open window, temperature fluctuations will affect the measurements. Ask if they use a thermal expansion correction in their software. Most modern CMM software allows you to input the part material and measurement temperature. This adjusts the reading to the reference temperature of 20 degrees Celsius.

Reason: Measurement infrastructure determines the accuracy of every tolerance check. If the measurement system is not calibrated or properly maintained, the foundry cannot prove dimensional tolerance consistency even if the casting process is stable. A foundry with a high-precision CMM but poor calibration practices is no better than a foundry with a low-precision gauge. The audit must verify the integrity of the measurement system itself.

Step 2: Analyze Historical CMM Data for Statistical Capability

Request CMM reports for at least three production lots of a similar part. Look for the mean and standard deviation of each critical dimension. Calculate the process capability index (Cpk) yourself if the foundry does not provide it. Cpk measures the distance between the process mean and the nearest specification limit, normalized by the process standard deviation. A Cpk of 1.33 or higher indicates the process can hold the tolerance with room to spare. A Cpk below 1.0 suggests the process is not capable of holding the specification consistently.

Do not just look at the pass/fail count. A foundry may report that 98% of parts passed inspection. That looks good. But if the mean is drifting toward the upper spec limit, the next 100 parts may fail. Cpk accounts for the spread of the data. It tells you how much margin the process has. If a foundry produces parts with a Cpk of 0.8, they are running a high risk of scrap. The process is capable of producing out-of-spec parts. If the foundry has a Cpk of 1.67, they have a wide safety margin.

Ask for the raw data if possible. CMM reports often summarize the data. Raw data allows you to check for outliers or shifts that the summary may hide. For example, if a foundry reports a standard deviation of 0.01 mm, but the raw data shows a cluster of points near the spec limit, the average standard deviation is misleading. The process may be unstable. Look for the trend over time. If the mean shifts from lot to lot, the process is not stable.

Reason: Historical data reveals whether the foundry has actually held the tolerance over time, not just in a sample. This is the strongest single indicator of dimensional tolerance consistency in an investment casting operation. A foundry that can produce a sample part within tolerance is not necessarily a foundry that can produce a lot within tolerance. Historical data bridges that gap.

Step 3: Inspect Core Shell Preparation Records

Investment casting dimensional tolerance is heavily influenced by the core shell. Request records showing core shell thickness, green density, and cure temperature. Look for variation between the first and last shell in a batch. The core shell is the barrier between the core and the metal. It must be thick enough to withstand the pressure of the metal during pouring, but thin enough to allow the metal to fill the cavity completely.

Check whether the foundry uses automated or semi-automated core shell application. Manual application introduces more variation in shell thickness. If an operator applies the investment by hand, the thickness depends on the operator’s technique, the viscosity of the investment, and the angle of application. Automated systems use spray guns or dipping baths to apply a uniform layer. This reduces variation.

Ask about the cure process. The investment must be baked to remove moisture and strengthen the shell. If the cure temperature is too low, the shell may be weak and deform during pouring. If it is too high, the investment may crack or shrink, changing the cavity dimensions. The records should show the temperature profile of the cure oven. For example, if the oven is set to 200 degrees Celsius for four hours, the record should show the actual temperature at the core location. If the oven runs hot or cold, the shell properties change.

Reason: Core shell thickness variation is a primary driver of dimensional shift in investment castings. If the foundry cannot demonstrate control over shell thickness, dimensional tolerance consistency is at risk. The core shell is the most direct determinant of the internal cavity size. If it varies, the part varies.

Step 4: Validate the Quality Control Plan

Review the foundry’s quality control plan for the specific casting operation. The plan should define which dimensions are checked, how often, and by whom. Look for control charts and reaction plans when a dimension drifts. A good quality control plan is specific. It does not just say “check dimensions.” It says “check the bore diameter every five parts using the CMM, and if two consecutive measurements exceed the upper control limit, stop the pour.”

Ask how the foundry reacts when a control chart signal occurs. Do they stop the pour? Do they rework the core? Do they scrap the batch? The reaction plan determines whether a tolerance shift becomes a shipment problem. A foundry that waits until the end of the batch to inspect will ship non-conforming parts. A foundry that inspects during the batch can stop the problem before it spreads.

Ask who is authorized to stop the process. In many foundries, the line operator does not have the authority to stop production. They must call a supervisor. This delay can allow dozens of bad parts to be poured. The audit should confirm that the operator can stop the line if a critical dimension drifts.

Reason: A quality control plan without enforcement is documentation. The audit must confirm that the foundry has defined and followed the reaction procedures when dimensional tolerance drifts occur. A plan that exists only in a binder is not a plan. It is a liability. The audit must verify that the plan is used.

Step 5: Confirm Material Thermal Expansion Compensation

Ask the foundry how they account for material thermal expansion during measurement. Investment castings are often measured while still warm. The foundry must compensate for the difference between measurement temperature and reference temperature. Metals expand when heated and contract when cooled. If you measure a steel part at 40 degrees Celsius, it will be larger than it is at 20 degrees Celsius. If the foundry does not compensate, the part will appear out of tolerance when it is actually in tolerance.

Verify the foundry applies the correct thermal expansion coefficient for the specific alloy. For example, a foundry measuring aluminum castings with the expansion coefficient for steel will produce incorrect dimensional readings. Aluminum expands more than steel. If the foundry uses the steel coefficient, the aluminum part will appear smaller than it is. This can cause a false pass or a false fail.

Ask if the foundry measures parts at room temperature. Ideally, parts should cool to 20 degrees Celsius before measurement. If the foundry measures parts immediately after the pour, the data is unreliable. If they measure parts after cooling, but the lab temperature is 25 degrees Celsius, the data is shifted. The CMM software can correct for this, but only if the correct coefficient is entered.

Reason: Thermal expansion compensation is a common source of measurement error. If the foundry does not apply the correct coefficient, the dimensional tolerance data they provide is inaccurate, and you cannot trust their consistency claims. A foundry that ignores thermal expansion may have a capable process, but their data will show it as incapable. Or, more dangerously, they may have an incapable process, but their data will show it as capable.

Step 6: Review Statistical Process Control Implementation

Ask the foundry to show how they use statistical process control (SPC) for critical dimensions. SPC charts track the mean and range of a dimension over time and flag when a process shifts. Common charts include the X-bar and R chart, which plot the average and range of a sample over time. Control limits are set based on the process variation, not the specification limits.

Look for control limits, not just specification limits. A foundry that only checks whether parts are in spec but does not monitor the process trend will react late to tolerance drift. For example, if a part is 0.01 mm above the upper spec limit, the foundry may reject it. But if the next five parts are 0.005 mm above the spec limit, the process is drifting. SPC would flag this drift before the parts fail.

Ask how often they collect data. If they measure one part every hour, they may miss a shift that occurs for 30 minutes. If they measure every 15 minutes, they catch the shift sooner. The frequency depends on the production rate and the criticality of the dimension. For a critical dimension, more frequent measurement is better.

Ask if they use automated data collection. Manual entry into a spreadsheet is prone to error. Automated data collection from the CMM eliminates entry errors and provides a continuous record. If the foundry uses a CMM that exports data directly to their SPC software, that is a sign of a mature quality system.

Reason: SPC enables the foundry to detect dimensional tolerance shifts before parts fall outside specification. This is the difference between reactive quality control and proactive dimensional tolerance management. Reactive quality control finds the bad part after it is made. Proactive quality control stops the bad part from being made.

Step 7: Conduct a Final Verification Batch

After the document review and data analysis, order a small verification batch. Run the same part through the foundry’s standard production process and measure the results independently. Do not accept the foundry’s measurements for this step. Use your own CMM or a third-party lab. This eliminates bias and confirms that the foundry’s data is accurate.

Compare your measurements to the foundry’s CMM data. If there is a consistent bias, investigate the cause before releasing the production order. A bias of 0.01 mm may be within tolerance, but it indicates a problem with the foundry’s measurement system or process. If the foundry’s data shows a mean of 50.00 mm, but your data shows 50.01 mm, the foundry’s data is shifted. This could be due to calibration error, thermal expansion error, or a process shift.

Ask the foundry to provide their raw CMM data for the verification batch. Compare the raw data to their summarized report. If the summary is incorrect, the foundry’s reporting is unreliable. This is a final check on the integrity of the foundry’s quality system.

Reason: The final verification batch confirms that the audit findings hold in actual production. Document review and data analysis can identify gaps, but a verification batch proves the foundry can execute the process as described. The verification batch is the final gate before production release.

Common Mistakes in Foundry Audits

  • Trusting only the sample part. A single good part does not prove the process can hold tolerance across a run. Always request historical CMM data. A sample part is a proof of concept. A production run is a demonstration of capability.
  • Ignoring core shell records. Shell thickness variation is a primary driver of dimensional shift. If the foundry cannot document shell control, dimensional tolerance consistency is questionable. The core shell is the most critical variable in investment casting.
  • Skipping the thermal expansion check. If the foundry measures parts warm without proper compensation, all dimensional tolerance data is skewed. Thermal expansion is a physics fact. Ignoring it introduces error into every measurement.
  • Accepting “in spec” as sufficient. Parts that fall within specification but show a drifting mean are not consistent. The audit must evaluate process capability, not just pass/fail results. A process can be “in spec” and still be failing.
  • Not running a verification batch. Document review identifies what the foundry says it does. A verification batch confirms what the foundry actually does in production. The verification batch is the final test.

Final Verification: Releasing the Production Order

Once the audit is complete and the verification batch passes, document the findings. Record the Cpk values, the core shell control methods, and the SPC implementation details. Share the summary with your quality and purchasing teams. This document becomes part of the foundry’s qualification record. It should be reviewed before each major production order.

If the foundry fails on any critical audit point, return to the prerequisites and request the missing data. Do not release the production order until the dimensional tolerance consistency is verified. A foundry that cannot provide the data is not ready to produce the part. A foundry that provides the data but shows poor capability is not ready to produce the part.

The audit is not a one-time event. Revisit the foundry’s performance after each major production lot. Dimensional tolerance consistency is a capability that must be maintained, not just demonstrated. Foundries change. Operators change. Equipment wears out. Processes drift. The audit must be repeated to ensure the foundry continues to meet your requirements.

Frequently asked questions

How many lots of historical CMM data should I request during a foundry audit for dimensional tolerance consistency?

Request at least three production lots of a similar part. This provides enough data to calculate a reliable process capability index and identify any trend in dimensional tolerance drift.

What Cpk value indicates a foundry can hold tight dimensional tolerances consistently in investment casting?

A Cpk of 1.33 or higher indicates the process is capable of holding the specification with margin. Values below 1.0 suggest the process is not capable and will produce out-of-spec parts.

Do I need to run a verification batch even if the foundry's historical data looks good?

Yes. Historical data shows what the foundry did in the past. A verification batch confirms the process is still stable and that the foundry can execute the same control methods for your specific part.

How does thermal expansion affect dimensional tolerance data from an investment casting foundry?

Castings are often measured while warm. If the foundry does not apply the correct thermal expansion coefficient for the alloy, all dimensional tolerance readings will be offset. This makes their consistency data unreliable.

What should I do if the foundry cannot provide core shell preparation records?

Treat this as a significant gap. Core shell thickness variation is a primary driver of dimensional shift. Without documented shell control, you cannot verify that the foundry can hold tight dimensional tolerances consistently.