Dimensional Out-of-Tolerance: Mould Problem or Machine-Setting Problem?

When T0 parts fail, the most common on-site reaction is "tune the machine and run another shot." But there is a class of defects that machine setting can only alleviate — and the more you tune, the further it drifts — because the root lies in structure, not parameters.

Take a common case: a housing with an appearance surface showed both local sink marks and ejection white marks at T0. On-site they first adjusted holding pressure, then melt temperature, running several shots; the sink looked lighter, but the white marks grew more obvious and the cycle lengthened. Looking back, one issue was locally excessive wall thickness with insufficient packing, the other was draft angle and ejector layout — both structural matters, with parameters only pushing the result toward the boundary. Those machine hours could have been saved by one short-shot test plus one no-load motion check. (Scenario is a generic illustration, not referencing any specific project.)

Public technical sources give a very practical starting point: first distinguish structural defects from parameter defects. Short shot, gas trap, scoring, and ejection white marks typically fall on the structure / mould side, where tuning only alleviates; while dimensional variation, sink, and unstable cycle time relate more to the process window (Source: Weiada Technology and other previously verified public sources; this is empirical classification, subject to the trial-mould report and bilateral technical agreement).

Sequence matters: structure first, then parameters.

The logic is simple: structure defines the feasible domain of the task; parameters only search for the optimum within that domain. If the domain itself is flawed, tuning just pushes the result toward the boundary — it looks improved short-term, but after a machine change, a batch change, or continuous running it returns. Worse, tuning parameters before excluding structural causes uses a new set of variables to mask the old problem, so in the next round no one can tell which step actually worked.

Trial-Mould Defect Attribution Tree

(Classification is general experience, subject to the trial-mould report, material-supplier data, and bilateral technical agreement; the table contains no parameter values.)

The short-shot test deserves a separate note: withdraw holding pressure and incrementally increase shot size to see where the melt stops last. In one run it answers three questions — is the gate count enough, is the location reasonable, where is the venting end — far more accurately than staring at finished parts, and saving more machine hours than "run another shot."

4 Pre-Checks Before Tuning

(General project practice, subject to bilateral technical agreement. Boundary vs. Copy 8 "T0 Report 6 Acceptance Items": Copy 8 audits the report document; here we check the on-site preconditions.)

3 Actions for R&D

• Change only one variable per trial-mould round: record the baseline before and the result after, forming a single-variable comparison. Changing a bundle at once means you can't tell which change helped, and can't roll back when it fails.

• Replace "run another shot to see" with a short-shot test: turn guesswork into an observable flow front, instead of inferring backward from appearance.

• Write each round into an issue list: symptom, suspicion, verification method, conclusion, owner, and closure status — all six fields complete before it counts as closed, then proceed to the next round. This is the same logic as "T0→T1→T2 issue-list closure per round," only here it locates the root cause rather than judging pass/fail.

When to Stop Tuning and Go Back to Structure

• After several tuning rounds the defect keeps recurring, only shifting in position or degree;

• One defect improves yet a new one emerges (white marks, flash, longer cycle);

• The parameter window narrows with each tune and fails immediately after a machine or batch change.

If any of these appears, stop adding variables to parameters — return to short-shot, no-load motion, and wall-thickness analysis.

How Attribution Lands on the Mould-Modification Order

After verification, conclude into three categories — the actions and accountability differ completely:

(Classification is general practice; responsibility and cost are subject to the bilateral contract. The design-change process is covered in Copy 7 and not expanded here.) Misclassifying is costly in practice: treating a structural issue as a chargeable design change, or a design change as the mould maker's responsibility, both turn into disputes in the next round.

A guiding principle: what parameters can solve, don't touch the steel; what must touch steel, don't drag it out with parameters; when unsure, do the short-shot and no-load motion first — don't rush to add parameters.

Jingmu outputs an issue list and improvement records during trial moulding in the order pre-check → structure exclusion → parameter optimization, and lands each round's verification conclusion on either the mould-modification order or the process card.