Technical Paper
Application of Machine Learning Algorithm to Engine Air System Failure Prediction
2024-04-09
2024-01-2007
With the probability of avoiding failure in advance, failure prediction is important to not only end users in vehicle industry, but also the service engineers. There are numerous studies focusing on failure prediction in vehicle industry with different approaches. We are proposing an approach to predict the failure of engine air system with Turbo chargers, using Machine Learning algorithms. Specifically, the anomaly detection algorithm is used, with Telematic data to predict the air system failure. We have analyzed the relationship between air system and all the features (about 60) we have, both by physical mechanism and data-wise. The parameters including altitude, air temperature, engine output power and charger pressure are used within the model, with sampling period 1 minute. Then, we have defined the normal state of each vehicle as benchmark, which is then used to determine the Medium surface for the specific vehicle.