Two diagnostic modes, trained and validated in this study. Runs entirely in your browser: no server, install, or data upload required.
Two-stage Random Forest trained on EngineFaultDB (Vergara et al., IEEE Access 2023), 55,999 samples from an exhaust-gas-analyzer bench rig. Stage 1 (good/bad) test accuracy: 100%. Stage 2 (fault sub-type) test accuracy: approx. 65%.
Gradient Boosting Classifier trained on the primary dataset assembled for this study (19,535 samples). All six inputs correspond to standard or commonly available OBD-II / engine-ECU parameters, making this the more field-deployable mode for a generic OBD-II scanner. Held-out test performance: 64.4% accuracy, macro-F1 0.632 (decision threshold 0.59, tuned on a validation split, never on the test set).
OBD / Sensor Readings
Model trained on EngineFaultDB (Vergara et al., IEEE Access 2023), 55,999 samples.
Stage 1 (good/bad) test accuracy: 100%. Stage 2 (fault type) test accuracy: approx. 65%,
shown as a confidence score below rather than a single flat answer.
Result
Enter readings and click "Run Diagnosis" to see a prediction.
OBD / Sensor Readings
Model: Gradient Boosting Classifier (150 trees, max depth 4, learning rate 0.06),
trained on the primary dataset assembled for this study (19,535 samples: Engine rpm, Fuel pressure,
Lubricant oil pressure, Lubricant oil temperature, Coolant temperature, Coolant pressure).
Held-out test performance: 64.4% accuracy, macro-F1 0.632. All inputs map to standard or
commonly available OBD-II / engine-ECU parameters (see PID notes on each field).
Result
Enter readings and click "Run Diagnosis" to see a prediction.