Process Control Parameter Optimisation · Brief 01
Line 1 quality incident on 2025-07-31
A running-process excursion, not a stop window
The one line you configure
Which lines and products are running defect rate above their learned band today, and is the cause upstream process or the product itself? Read it across the 19 hourly PLC parameters, quality and conveyor speed.
In a nutshell
This is a running-process excursion, not a stop window. Line 1 is the affected line because line-level defect rate reached 537.50K against 168.91K expected (+218.2%) — the only requested KPI anomaly visible at line level. In the selected Line 1 × SKU-4417 context the PLC read shows wash tank 3 temperature breaking out at 50 against a 39 band by 03:00, peaking twice at 56 vs 39 at 05:00 and 06:00 (+43.6%) and staying above band through 49 at 09:00, with a late secondary move in wash tank 1 at 48 vs 39 at 19:00. Conveyor speed near 47.00K vs 48.00K (−2.1%) says the line kept running while the parameter drifted.
At a glance
PLC parameters by band index · 100 = on band, higher = above band
What moved
Connected signals — what else moved, and what didn't
Conveyor speed stayed inside its running range and the whole curing stage held — the discriminators that rule out a stop window or a curing fault. Quality yield % is elevated, not depressed, so it is context rather than the anomaly. Only the wash stage moved.
Wash tank 3 broke band at 03:00 and stayed above it to 09:00 · Line 1, hourly
Break-out at 03:00, a twin peak of 56 vs 39 at 05:00 and 06:00, then a slow decay that never returns to band before 09:00 — a long high band, not a momentary spike. A daily read would have shown only the aggregate.
Where the excursion sat · PLC parameters × hour (above learned band)
Two parameters move all day and both are wash-stage on the same line-product: wash tank 3 across 03:00–09:00, then wash tank 1 once at 19:00. Rinse, cooling and the whole curing stage never leave band, and conveyor speed never leaves its running range — which is what makes this a process item rather than downtime.
What the connected view adds
Each source looked ordinary alone: the historian showed a warm tank inside its hardware limit, the quality stage picked a line, and OEE showed the line still running. Only reading them together on the same line × product × hour slice isolates a long high wash-tank-3 band on the line quality had already selected — surfaced during the shift, not in the next morning's quality review. It is also why quality yield % printing 87.08 vs 79.85 (+9.1%) is context, not the anomaly: it is elevated in the drill slice and did not select the line.
So what
Line 1 is the affected line because line-level defect rate landed at 537.50K against 168.91K expected (+218.2%) — the only requested KPI anomaly visible at line level. On the selected Line 1 × SKU-4417 slice the hourly PLC read isolates a long high wash-tank-3 band: break-out at 03:00, a twin peak of 56 vs 39 at 05:00 and 06:00 (+43.6%), still above band at 09:00, with a late wash-tank-1 move at 19:00. Conveyor speed near 47.00K vs 48.00K (−2.1%) says the line kept running while the parameter drifted, so this is a running-process excursion and not a stop window.
The points that matter
The line was selected by defect rate, not by quality
Line-level defect rate at 537.50K vs 168.91K is the only requested KPI anomaly visible at line level — that is what picked Line 1.
+218.2%defect rate vs expectedQuality yield % is context, not the anomaly
In the drill slice quality yield % printed 87.08 vs 79.85 — elevated by 9.1%, so it confirms the slice but did not select the line.
+9.1%quality yield · elevatedA long band, not a spike
Wash tank 3 breaks out at 03:00, peaks twice at 56 vs 39, and is still above band at 09:00 — six hours of drift on one circuit.
56 vs 39+43.6% peakThe line kept running
Conveyor speed near 47.00K against 48.00K (−2.1%) stayed in running range, which rules out a downtime artefact and makes this an engineering item.
−2.1%conveyor speed · runningRecommendation
Send maintenance to the wash-tank-3 circuit on Line 1 — the controller setpoint, the sensor and the steam valve, in that order — and alarm the learned band rather than the hardware limit. 56 against a 39 band was legal for the equipment and still ruinous for the product. Watch wash tank 1 on the same line-product next: its 19:00 move is the same circuit family.
Questions it already answers
Is this the process or the product?
The process. Quality picked the line and the hourly PLC read then isolates wash tank 3 on that same line-product across 03:00–09:00; the product context itself prints an elevated quality yield %, so it is not the product failing.
What can the data not tell us here?
Not why the tank ran hot — a controller setpoint, a failing sensor or a steam-valve fault all look identical in the historian. It pinpoints the circuit, the hours and the running state so maintenance checks one thing, not nineteen.
Where else should we look?
Watch wash tank 3 against its learned band hourly on both lines, and alarm the learned band rather than the hardware limit — 56 against a 39 band was legal for the equipment and still ruinous for the product. Wash tank 1's 19:00 move on the same line-product is the one to watch next.