“The average cushion is unchanged” can be true while every other cycle looks different. Keep the cycle sequence before deciding that a cushion trend is reassuring—or that a check ring needs replacing.
Cushion variation is a reason to compare a clearly defined machine value with the same cycles' process and part records. It is not a stand-alone wear diagnosis. Start by identifying what the controller measures, then examine the order of the readings and the observations that accompany them.
What does this screen mean by cushion?
In RJG's “Why Cushion?”, cushion is discussed as material remaining ahead of the screw at its furthest forward position during hold. The article distinguishes that condition from the material remaining at transfer, and discusses volume in its eDART context. It does not establish an interchangeable field definition for every machine.
The reading may be expressed as a position or a volume, with a particular reference and sampling event. Preserve that identity. A number copied into a spreadsheet column called “cushion” can lose the detail needed to compare it with last week's export.
- Exact field
- Copy the full controller label and, where available, the export field identifier.
- Event or interval
- Locate the manual's definition: when is the value captured, or over what interval is an extreme selected?
- Units and reference
- Retain millimeters, volume units, sign convention, and the documented reference. Do not supply a guessed conversion.
- Record identity
- Keep press/controller identity, recipe revision, timestamp, and cycle ID with the value.
A field called “minimum” and one called “end” should not be merged merely because both contain the word cushion. One may select an extreme over an interval while another captures an event; the exact definitions must come from that controller. If the manual or export definition is unavailable, mark the comparison unresolved. Renaming the columns does not resolve it.
Similarly, equal position readings on different presses are not automatically equal volumes. Comparing presses needs equipment geometry and measurement definitions. The exercise below deliberately stays with one hypothetical field on one hypothetical press.
Two averages of 5.0 mm, two different records
Constructed teaching data. Series A and B each contain eight consecutive cycle readings of the same hypothetical field, in millimeters. The field is assumed to have a consistent definition and reference. These values are not target settings, acceptance limits, or evidence from a production machine.
| Series | Cycles 1–8 (mm) | Mean | Range |
|---|---|---|---|
| A | 5.0, 5.1, 4.9, 5.0, 5.0, 4.9, 5.1, 5.0 | 5.0 mm | 0.2 mm |
| B | 4.0, 6.0, 4.0, 6.0, 4.0, 6.0, 4.0, 6.0 | 5.0 mm | 2.0 mm |
Each series sums to 40.0 mm; dividing by eight gives the same mean. Series B has a much wider observed range and an alternating pattern. A report containing only the mean conceals both differences.
It would still be wrong to label A “capable” and B “out of specification.” No approved limit, measurement capability, longer baseline, or part result is provided. Eight invented readings illustrate a reasoning problem; they do not establish statistical control limits. Also consider display rounding: repeated values can conceal changes smaller than the displayed increment.
Check the sequence before explaining the pattern
If real data alternated like B, first check whether each row represents one consecutive cycle from the same press and field. A spreadsheet could interleave two streams, duplicate records, or sort text IDs incorrectly. Verify the source export before looking for a physical mechanism that repeats every second shot.
A step change calls for a timestamped comparison around the transition. A gradual drift calls for the longer sequence and its operating history. A single extreme calls for its original record, alarm context, and sample identity. None of these patterns names a failed component by itself.
Which part belongs to which reading?
“Cushion changed and parts got lighter” is not yet a paired comparison if the trend covers the morning and the weights came from an unlabeled afternoon sample. Match the part and cycle IDs. Record every cavity separately when cavity differences matter.
For an initial review, assemble the same field before and during the change, actual fill and hold records available for those cycles, material and recipe identity, interruption history, and the applicable part checks. Mark any fields that are settings rather than measured results. The process-sheet reading guide explains that distinction.
| Evidence available | Reasonable statement | Still needed |
|---|---|---|
| Cushion and part mass vary in the same identified cycles | The observations are associated in this window. | Defined methods, other process evidence, and an approved investigation to distinguish explanations. |
| Cushion varies; selected dimensions meet their documented criteria | Those checks passed for those samples. | Required checks not represented by the selected dimensions; the reason for the trend. |
| Field name or reference changed | The records may not be directly comparable. | The controller definition and change history before interpreting a physical shift. |
| A change appears after maintenance | The timing provides a lead. | Maintenance scope, restart history, and comparable before/after records. |
A check-ring hypothesis needs its own evidence
RJG discusses screw-motion variation and check-ring leakage as related considerations in its cushion article. That makes the non-return valve a possible line of inquiry in an appropriate investigation. It does not mean that every variable cushion proves leakage, or that a stable cushion proves every part is acceptable.
Before sending a maintenance request that names a failed component, separate three things: the observed trend, the proposed explanation, and the verification requested. For example: “This defined field has a wider range than our comparison window. Cycle IDs and corresponding measurements are attached. Please review whether an approved non-return-valve evaluation or another investigation is warranted.”
Whoever owns that evaluation needs the machine identity and relevant history. A trend screenshot without axes, units, event definition, or cycle labels gives them less to work with than a short, traceable export. Preserve the original file as well as any chart made from it.
Leave a usable finding
A concise finding for the teaching data is: “A and B both average 5.0 mm. B ranges from 4.0 to 6.0 mm and alternates across eight cycles; A ranges from 4.9 to 5.1 mm. These observations alone do not establish part acceptance or component condition.” That states the difference without turning it into a repair instruction.
In production, add the applicable escalation rule, current product disposition, responsible person, and next review point. If an authorized trial follows, use the troubleshooting log to record the question and the observations that would support or challenge the explanation. If sink marks are also present, retain their cavity and feature identity; see reviewing sink marks after a hold-pressure increase.