Industrial Plant Assessment — worked examples

Worked examples on public and anonymised data · CTRL Designer LLC · 11 September 2026

Objective

Take a plant's own historian export and answer three questions: is the data good enough to trust, when did the plant leave normal operation, and which of those departures are worth acting on. Nothing comes from a model of how a plant should behave — everything is measured against a stretch of the plant's own history chosen as normal.

Four kinds of study

Continuous plants run through the five-step A to E pipeline. A batch process needs a golden batch instead, because every batch is a different length and nothing about it is steady. Control loops get their own analysis, because the question there is not when the plant left normal but whether a cycle is being made inside a loop or arriving from outside it.

Continuous · A to E

Tennessee Eastman — continuous chemical plant

The chemical industry's standard benchmark, with faults injected at known times so the analysis can be marked against an answer key. Run three ways to show what each signal set contributes.

TEP-Plant-10days-AIAO ctrldesigner.com/ipa/tep/TEP-Plant-10days-AIAO/	all 52 tags — measurements and outputs	80 MB
TEP-Plant-10days-AI ctrldesigner.com/ipa/tep/TEP-Plant-10days-AI/	41 measurements only	76 MB
TEP-Plant-10days-AO ctrldesigner.com/ipa/tep/TEP-Plant-10days-AO/	11 controller outputs only	35 MB

Continuous · A to E

Synthetic chemical reactors — six units in parallel

Six reactors under two operating regimes with a slow seasonal drift, gradual faults and 6% missing values by design. It tests whether the method survives a multi-train plant.

Chemical-Reactors-21d-AIAO ctrldesigner.com/ipa/cr/Chemical-Reactors-21d-AIAO/	all 78 usable tags	388 MB
Chemical-Reactors-21d-AI ctrldesigner.com/ipa/cr/Chemical-Reactors-21d-AI/	measurements only — the same 78, since every output is a constant setpoint	388 MB

Batch · golden batch and multiway PCA

IndPenSim — 100 penicillin fermentation batches

A 100,000 litre fermentation, 100 batches run three different ways plus ten carrying faults. Normal is a set of good batches rather than a window of time.

IndPenSim-Batch
ctrldesigner.com/ipa/ips/IndPenSim-Batch/

four reports, one per golden batch

25 MB

Control loops · CLA

CDL T123 — 70 control loops

Real controller data from a plant. For every cycling loop it asks whether the cause is inside the loop, meaning tuning or a sticking valve, or outside it, meaning something upstream. Rated with the CDL Control Index and ordered by what each fix buys.

T123-Loops
ctrldesigner.com/ipa/loops/T123-Loops/

70 loops, 48 cycling

57 MB

Questions

These are worked examples. If you want to know how the method would read your plant's data, that is the conversation worth having.

Email info@ctrlDesigner.com

Yahya Nazer · CTRL Designer LLC