A multi-cavity mould produces several pump housings per cycle. Random samples pass, but assembly repeatedly finds a group of covers that will not seat properly. One missing piece of information may be cavity identity. A shared mould does not guarantee identical filling, packing, cooling or local wear in every cavity.
Sampling should preserve cavity and time information before deciding what to measure and how many parts to inspect.
A batch average can hide a local difference
One cavity may differ around a sealing groove, screw boss or register while the others remain satisfactory. A combined average can conceal that pattern. Samples taken from one container or a few consecutive cycles may also fail to represent every cavity, even when the total sample count looks substantial.
Assembly difficulty does not itself prove a cavity defect. Material condition, process settings, distortion during removal and the mating component may contribute. Stratification helps separate these influences rather than presuming one cavity is necessarily bad.
Connect identity with the process history
Where practicable, retain cavity identity, production period, material batch and process revision. Any physical identification must avoid functional sealing and structural areas. If parts have no readable cavity mark, the factory needs an appropriate controlled collection method. Inspectors should not guess cavity identity from appearance.
In a hypothetical four-cavity example, random samples happen to contain only cavities 1 and 2, leaving a groove-depth difference in cavity 3 unseen. Sampling each cavity and covering specified later process periods gives a more informative view. This example does not prescribe a universal number per cavity. Quantity and frequency depend on risk, process capability and the agreed quality plan.
| Sampling dimension | Purpose | Retain |
|---|---|---|
| Cavity | Identify local differences | Identity and critical dimensions |
| Time window | Observe process drift | Period, restarts and process state |
| Material condition | Separate batch and conditioning effects | Material identity and conditioning |
| Mating combination | Verify interface compatibility | Housing and cover revisions and cavities |
Use functional dimensions and consistent datums
Groove depth and width, register diameter, face flatness and boss position can all affect assembly. Select dimensions connected to function rather than expanding the inspection list indiscriminately. Align the datum, fixture and measurement timing.
A flexible plastic housing clamped too tightly can produce a reading that represents neither its free condition nor the assembled state. Temperature and moisture can also change some polymer dimensions. Measurements made under different conditions cannot automatically be interpreted as mould wear.
The measurement system needs suitable resolution and repeatability. Otherwise apparent cavity differences may reflect the inspection method. Where the difference approaches measurement capability, resolve that issue before making a tooling decision.
Verify the assembled interface as well as the loose part
Use an approved plan to examine relevant combinations of cavity and revision. Record fastening condition, seal compression and required functional results. Selecting the pair that assembles most easily does not represent ordinary production.
If controlled matching is necessary, define it as a manufacturing requirement with traceability. Undocumented hand selection at the assembly station hides the extent of the variation and makes spare compatibility difficult later.
When one cavity is abnormal, identify associated stock and production periods, preserve rework records and assess whether mating parts are affected. After tooling or process correction, reconfirm the relevant dimensions and function. Good appearance alone does not close a sealing-interface issue.
Present results by cavity before giving the overall conclusion. Purchasing can then distinguish a local tooling action from a wider material or process change and follow whether the corrective action remains effective over time.
A recurring cavity difference should lead back to tooling and process action. Increasing random final inspection may add workload without correcting the source. Verify the correction in that cavity and over time. Buyers should check actual cavity coverage rather than assuming that an agreed total sample count covers every structural risk.
Frequently asked questions
Does one measured part from each cavity guarantee production quality?
No. It is a limited snapshot. Time variation, process capability and risk still determine the sampling plan.
Must the cavity number appear on the visible exterior?
Not necessarily. An appropriate location or controlled process record can work, provided identity is reliable and function is unaffected.
Can a satisfactory batch average outweigh one cavity’s critical deviation?
No. Apply the individual and interface requirements, then assess the affected scope rather than concealing it in an average.