Operations Analytics Felix Fu  /  Engine health forecaster Synthetic data · live in your browser Portfolio
ENGN

Engine health forecaster

Twenty engines are wearing out at different rates. From each engine's exhaust gas temperature (EGT) margin, forecast how many flight cycles it has left, with an honest range, so removals can be planned rather than forced.

20 engines, 10 aircraftPer-cycle sensor dataBootstrap intervals

Observed EGT marginFitted wear curveForecast80% range

Supporting signals, 10-cycle averages

Vibration (in/s)

Fuel flow above baseline (%)

Fleet removal forecast

Ranked by the cautious end of the forecast. Select an engine to inspect it. Remove: cautious forecast under 50 cycles; monitor: under 120.

EngineCyclesEGT marginCycles left
80% range
Status

Is the model any good?

EnginePerfect forecast

Method

EGT margin shrinks as an engine wears, slowly at first and then faster. Each engine gets its own curve, margin = a − b(et/τ − 1), fitted by least squares over a grid of τ. The forecast is the cycle where that curve crosses the 10 °C limit.

Sensors are noisy, so a single number would overstate confidence. Residual bootstrap (refitting on resampled noise 80 times) gives the 80% range, and planning uses its cautious end. Expect the range to run slightly tight, because it ignores uncertainty in the shape of the wear curve itself; a production version would add that, for example by blending several curve families.

The replay slider shows the practical point: early in an engine's life the range is wide; it tightens as wear accelerates, which is exactly when the removal decision has to be made.