All services
Fouling or performance prediction
Forecast duty loss on exchangers and similar equipment from existing historian data.
From $2,000 for a typical scope
Discuss this service“Performance drifts over time”
Exchangers lose duty gradually, and cleaning crews need notice. The question is whether your existing data gives useful warning, and whether a model does better than the rule you use today.
A good fit when
- One to three exchangers or similar assets to start with
- Twelve to twenty-four months of history
- A cleaning or maintenance decision with a measurable cost
How the work runs
Timing is a planning estimate, confirmed after the first stage.
| When | Work | You can review |
|---|---|---|
| Week 1 | Tag mapping, unit checks and heat-balance audit | Data-quality report and go/no-go decision |
| Weeks 2–3 | Clean-state model, campaign boundaries and simple baselines | Baseline report and frozen test split |
| Weeks 3–5 | Fit and calibrate models on held-out periods and assets | Errors, misses, false alarms and band widths |
| Weeks 5–6 | Review failed forecasts with process and maintenance engineers | Failure atlas and rules for when not to forecast |
| Weeks 6–8 | If justified, run shadow forecasts from a read-only feed | Time-stamped predictions and review log |
What we need
- Inlet and outlet temperatures on both streams
- Both stream flows, with units and basis
- Exchanger area, arrangement and design clean performance
- Cleaning records and known operating changes
- Pressure drop, where available
What you keep
- Clean-state thermal model and fouling estimate, with assumptions
- Versioned forecasting pipeline
- Baseline comparison and calibration report
- Failure atlas covering missing data, drift and unfamiliar operation
- Decision note: proceed, narrow the scope or stop
Agreed before we start
- Planning lead time and duty floor
- Maximum missed events and false alarms per month
- Useful timing tolerance and maximum useful uncertainty
- Minimum number of held-out campaigns
When we’d stop
Too few cleaning cycles, unmetered bypasses or poor sensor agreement can make a narrower monitoring study the right first step.