Across 360 hours of hourly process logs — covering paste preparation, two esterifiers, two prepolymerisers, the disc ring finisher, transfer line, lab QC, breakages and MEG recovery — this is what January 1–15, 2026 looked like at your CP train. The plant ran a deliberate production ramp-down from 362.3 → 334.5 TPD while holding product IV, colour and composition on spec. What follows is where that discipline held, where it slipped, and which upstream signals foretold every major spin-pack breakage.
The diagram below is a faithful layout of your CP train — from the paste tank through two esterifiers, two prepolymerisers, the DRR finisher, transfer line and on to the spinning section. Each vessel shows the instrument readings for the selected hour. Residence-time estimates (grey text under each vessel) determine when material currently visible here reaches the spin pack.
Production fell from 362 TPD to 334 — a clean, almost linear ramp. Intrinsic viscosity (the plant's single most important quality number) held at 0.640 for all 360 hours. Breakages drifted mildly higher through the first ten days. Vacuums degraded subtly in the finisher and PP2. Everything else was routine.
Spin-pack breakages don't happen because of what the plant is doing right now — they happen because of what the material now at the spin pack saw in the reactors hours ago. To find genuine leading indicators, we tested every candidate process variable against the breakage rate across lags of 0 to 14 hours, detrended to remove month-long drifts, and picked the lag where each variable's correlation with breaks peaked. The heatmap shows every variable; the column position of its strongest cell is the empirical residence-time signature.
Across 17 setpoint/process-variable pairs, tracking errors are typically a fraction of a percent. The one outlier is ES1 DPT level, which ran with mean absolute error of 1.65% against a 63% setpoint — which also happens to be one of your top breakage predictors (5-hour lag, r ≈ +0.25). Every other loop is clean.
| stage | loop | sv mean | mean |err| | max |err| | std(PV) | quality |
|---|
The PTA blend shifts dramatically from day to day — Reliance Tanker jumps from 22% on Jan 4 to 67% on Jan 14, Ineos runs down from 32% to 9%, and GCM climbs to 25% by month-end. On the recovery side, hotwell 13V02 water content creeps up from 7.55 to 8.85 (+17%), and 17V01 takes a one-hour excursion to 5.5 on Jan 9. Worth a recipe audit and a hotwell inspection.
Each card below states the observation, the evidence, and a concrete next step. The high-priority items are the ones most worth checking before the next rate ramp.
0.141 mbar to 0.162 mbar (+15%). The lagged-correlation analysis ranks it #1 among all variables tested — r = +0.35 at a 4-hour lag (p < 0.001, n = 354). In every major breakage window, DRR vacuum readings were elevated 3–5 hours earlier.
1.65% against a 63% setpoint (max 3.3%). Of 17 control loops this is the only non-excellent performer. It also shows r = +0.25 at 5 h lag with breakages — matching the expected ES1-to-spin residence time almost exactly.
7.55 → 8.85 through the month (+17%), with 15V01 showing a parallel creep. Combined with the DRR / PP2 vacuum drift, this points toward accumulating load on the MEG recovery system — possibly tray fouling or a water make-up imbalance.
22% → 67% → 38%; Ineos dropped 29% → 10%; GCM rose to 25%. These are availability-driven shifts, not recipe-driven. Lab ES1 end-group values range 769–802 meq/kg, with the high end clustering in weeks where Reliance-Tanker dominated.
0.638–0.642 across all 360 hours (σ = 0.0008). L-value 81.9–82.6. B-value 3.66–3.89. Paste density locked at 71.30 ± 0.01%. Catalyst trim, temperature step-down, and pressure response were all executed cleanly — operators earned this result.
334 TPD) is still well above typical shutdown rates, the most likely scenario is either (a) a planned shutdown later in January or (b) a demand-driven slowdown that continues into early February. When the second half of January and the 2025 history arrive, the two highest-value analyses are: does the DRR-vacuum trend continue or reverse post-ramp, and what does the normal breakage baseline look like at steady 350+ TPD.