ML Based Misfire Prognostics using Connected Vehicle Data
2022-01-0490
03/29/2022
- Event
- Content
- For Spark-Ignition engine, misfire reduces gas mileage, increase emissions and even damages the engine. Thus, misfire monitor is required by California Air Resources Board (CARB)’s On-board diagnostics (OBD) II regulations. Misfire rates are determined based on the misfire counts, and the monitored event count. If misfire rates are above pre-set threshold under certain conditions, a misfire Diagnostic Trouble Code (DTC) will be set and remind the customer for repairing which increase OEM ‘s engine misfire warranty and cause customer satisfaction issue. The objective is to leverage connected vehicle (CV) data to detect potential misfire issue and fix it through calibration software update, thus reduce warranty and improve quality. An LSTM network enables us to input sequence data into a network, and make predictions based on the individual time steps of the sequence data. Moreover, it is capable of learning long-term dependencies which is a perfect fit for misfire feature. Thus, the work presented in this paper is a LSTM based misfire prognostics algorithm. It was tested on a production vehicle line. The result shows great potential to identify future misfire DTCs on VIN level with only limited misfire data from OBD test report. The result is promising while more improvement and validation are needed.
- Citation
- Wang, Z., "ML Based Misfire Prognostics using Connected Vehicle Data," SAE Technical Paper 2022-01-0490, 2022, .