Now live: a hybrid process digital twin for amine gas sweetening — it names the fault and the root cause, not just the alarm. Explore the platform →

QEM & OPS logo

Fault Prognostics

Prognose

Live in the platform

See the slow faults coming, without inventing a date.

Some faults arrive in minutes. Fouling, scaling, corrosion and filter plugging take days. Prognose fits those channels over a three-day look-back and projects each one against the threshold that would declare the fault — with an uncertainty band, and without a number we cannot defend.

See it in the platform How it works
4Slow-wear faults in the pilot set
3 daysLook-back window
6 hProjection horizon
±2σUncertainty band

What it does

Four things it does that a dashboard does not.

01

Slow-wear channels only

Lean/rich exchanger fouling, reboiler fouling and scaling, general corrosion, and solvent filter plugging. These are the faults where a projection is meaningful. Fast faults belong to Detect, and mixing the two produces confident nonsense.

02

The curve is chosen by evidence

Linear, exponential, power and logistic fits are all tried and compared by information criterion. The winner is reported with its R², so an engineer can see how much of the trace the model actually explains.

03

Margin, rate and destination

Each channel reports where it stands now, the threshold that declares the fault, the margin between them, the rate of change per day, and where the projection lands at the end of the horizon.

04

Five verdicts, in plain words

In-fault, approaching, deteriorating, stable, or insufficient data. An engineer should be able to read the outlook without reading the manual.

How it works

The pipeline, end to end.

  1. 1

    Slice the look-back

    A three-day trailing window ending at the moment you select.

  2. 2

    Keep only steady operation

    Unsteady rows are dropped and the remainder is reduced to hourly medians. A degradation trend fitted through load swings is a fit to the load, not to the wear.

  3. 3

    Fit four candidate curves

    Linear, exponential, power and logistic, selected against each other by information criterion rather than by whichever looks best.

  4. 4

    Project with a band

    Six hours ahead, carrying a ±2σ uncertainty band on the projection.

  5. 5

    Compare and rule

    The projection is set against the fault threshold, and the verdict follows from the margin and the direction of travel.

Under the hood

The engineering, stated plainly.

Every figure on this page is countable in the delivered system. If you want to see where a number comes from, ask us and we will show you the record it is read from.

Pilot fault set
Exchanger fouling · reboiler fouling · corrosion · filter plugging
Conditioning
Hourly medians over steady rows only
Model selection
AIC across linear / exponential / power / logistic
Projection
6 h horizon with ±2σ band
Refitting
None per query — the baked calibration is reused
Reported
Verdict · current · threshold · margin · rate/day · R²

What it does not do

The limits, published on the same page as the promise.

  • It does not give you a days-to-failure number. The back-test showed the fit has no predictive skill beyond about half a day, so the tool refuses to print a figure it cannot support. Most prognostics products will happily give you one.
  • Four pilot faults today, not all twenty-two. The rest are candidates, not coverage.
  • A projection assumes the operating mode holds. Change the load or the solvent and the trend restarts.

Where it applies

Built for amine systems today. Not limited to them by design.

The platform is focused on acid gas removal — gas sweetening and post-combustion carbon capture — because that is where the model is calibrated and where the fault library is real. The method underneath is not amine-specific: a physics reference model, residuals against expected-healthy, fuzzy diagnosis over graded indicators, and costed consequences. Extending it to another unit means new datasheets and a new calibration, not a new platform.

  • Gas sweeteningLive today
  • Post-combustion carbon captureLive today
  • Other acid gas removalMethod applies
  • Wider process unitsFuture

The rest of the platform

One model. Seven ways to use it.

RunThe process digital twin that says what a healthy train should be doing.DetectNames the fault, the root cause, and what it costs per hour.DiscoverThe whole train in KPIs, trends, profiles and correlations.qemIntelAsk the plant a question, and get an answer grounded in real data.OTSIn developmentTrain crews on faults they have never seen. In development.TuneIn developmentFind the valves and loops working against you. In development.

The fastest way to judge this is to open it.

The platform runs on a real amine train with thirteen days of one-minute history behind it. Nothing on these pages is a mock-up.

Open the platform