The Pipeline
From Data to Diagnosis
From a tag reading to a named fault with a price on it.
Six stages, one direction. Every step is inspectable: open any diagnosis and you can see the indicators that made the call, the ones that corroborated it, and the ones that would have ruled it out.
The pipeline, end to end.
- 1
Predict the healthy plant
The reference model takes the boundary conditions — feed, solvent, steam, cooling — and returns what every monitored tag should read, plus the soft sensors and health parameters. Calibration is baked, so a single point runs in seconds.
- 2
Take the residual
Actual minus expected. A fault shows up as a gap, not as a fixed alarm limit that has to be re-tuned every time the plant changes load.
- 3
Gate the operating mode
Load envelope, baseline deviation and steadiness are checked before anything is judged. A plant that is ramping is not a plant that is faulty.
- 4
Grade 54 indicators
Level, trend and oscillation flavours over a six-hour window. Each is given a membership in fixed engineering-unit states rather than standard deviations from a moving average — so a plant that is slowly getting worse cannot quietly redefine normal.
- 5
Match faults, attribute causes
Twenty-two fault signatures combine primary evidence, corroboration and exoneration into a confidence and a tier. Nineteen root causes are then attributed across 247 mapped causal edges.
- 6
Price the consequence
Six cost buckets — off-spec gas, amine losses, excess steam, reliability, throughput — each with a nominal figure per hour, so the shift can rank what to fix first.
Where prognostics fit
- Forecasting runs on the same output. The slow-wear channels are fitted over a three-day look-back and projected six hours ahead — and deliberately no further. The back-test showed the fit has no skill beyond about half a day, so no days-to-failure number is produced. Most tools will give you one anyway.
More on how it works
The method is easiest to judge from the output.
The platform runs on a real amine train with thirteen days of one-minute history behind it. Open a diagnosis and read the evidence for yourself.
Open the platform