How Should Dimensional and Leak-Test Results Be Correlated for Cast Housings?
Dimensional and leak-test results should be linked by the same part identity, seal-interface zone, assembly condition and process history, then analyzed as complementary evidence. Dimensions show whether face, groove, bore and bolt pattern meet the drawing; a leak test shows whether the tested component or assembly meets a defined integrity limit. Neither result automatically overrides the other.
Correlation needs enough variation to learn which dimensions or surface conditions influence leakage, not only a few ideal passing parts.
What the Leak Test Must Define
State medium, pressure or vacuum differential, fill/stabilization, test time, temperature, internal volume, allowable rate, equipment resolution, calibration, fixture seals and test-port locations. Pressure decay is sensitive to temperature and volume. Bubble and tracer methods have different sensitivity and localization capabilities.
Test Input | Why It Matters | Control |
|---|---|---|
Medium | Gas and liquid flow differently | Name approved test/service relationship |
Pressure differential | Changes path flow and part loading | Regulated measured condition |
Temperature | Changes gas pressure and component size | Stabilization and compensation rule |
Fixture seal | Can create or hide external leakage | Blank/master and maintenance checks |
Test duration | Affects sensitivity and throughput | Validated sequence |
Which Dimensional Data Should Be Linked?
Link free-state flatness or profile, groove depth/width, land width, bolt-hole or thread position, surface roughness and lay, local indication map and mating-part geometry where controlled. Record whether the part was measured before or after coating, impregnation, cleaning or assembly.
Use the drawing datum alignment. A best-fit scan can make a flange look better while losing the functional relationship to a bore or port. Preserve raw maps instead of only pass/fail labels.
How to Investigate the Four Result Combinations
Dimensions | Leak | Action |
|---|---|---|
Pass | Pass | Release only if all remaining requirements pass; trend normally |
Pass | Fail | Check porosity path, seal damage, cleanliness and test fixture |
Fail | Pass | Hold for drawing disposition; test success does not waive geometry |
Fail | Fail | Contain population and separate form, assembly and integrity mechanisms |
When a leak fails, preserve the assembled seal, torque/load record and as-found surface. Locate the path before reworking where possible. A repeated test after tightening more or adding sealant changes the condition and cannot replace the original result.
How to Build a Correlation Study
Select production-intent parts spanning approved flatness, waviness, texture and casting indication ranges. Include multiple cavities and fixture positions. Assemble with controlled mating parts, seals and load. Run the leak method at defined conditions, then analyze continuous raw values rather than only pass/fail.
Correlation does not prove causation. If leak rate rises with flatness, confirm that gasket, temperature and fixture leakage did not change simultaneously. Destructive section or seal-imprint evidence can strengthen the mechanism.
How Test-Fixture Leakage Is Separated From Part Leakage
Use blanks, calibrated leaks or masters as appropriate, maintain fixture seals and record their life. Design test seals so they do not cover the product path being evaluated. A fixture can compress outside a visible pore and create a false pass, or leak at a connector and create a false part failure.
Repeatability studies should include loading and unloading the same part. If variation is near the acceptance limit, improve the test method before using it for production disposition.
What Records Support Corrective Action?
Retain serialized/lot identity, casting cavity, machining fixture/program/tool, dimensional raw maps, surface-indication map, cleaning/coating/repair status, seal and mating part, assembly load and raw leak curve or rate. The post-machining process and casting process can then determine whether failures follow face form, machining depth, one cavity or an internal path.
Correlated evidence prevents two common errors: accepting an out-of-drawing part because it passed one leak test, and changing machining when the test fixture or seal assembly caused the failure.
How to Qualify the Leak-Test Measurement System
Run repeated loads of a stable master, known calibrated leak or qualified reference to quantify repeatability, warm-up and drift. Challenge fixture seals, connector torque, part seating, operator sequence and temperature. The study should span the acceptance region and normal test cycle, not only a zero-leak blank.
For pressure decay, record raw pressure and temperature over time. Automated compensation should be verified against physical references. If repeated results on one stable part cross the acceptance limit, improve the method before sorting production.
MSA Input | Question | Reaction |
|---|---|---|
Repeated same-part loads | How much does loading/test repeat? | Improve fixture seals or sequence |
Known leak standards | Does the system detect the relevant range? | Calibrate or change method |
Temperature challenge | How sensitive is result to thermal drift? | Stabilize or validate compensation |
Multiple operators/fixtures | Are results transferable? | Standardize and retrain |
How to Use Data Without Creating False Correlation
Do not correlate pass/fail labels only. Use raw flatness, groove, roughness, indication and leak values with cavity, fixture, seal lot and temperature. Stratify by process source before fitting trends. A relationship that appears across mixed cavities can disappear when one cavity or test fixture is isolated.
Confirm any suspected driver with a controlled trial. Changing face milling, gasket and test fixture together may create passing parts but cannot show which change solved the failure or whether the new process is stable.
