Leveraging AI/ML Algorithms for Analyzing Fleet Vehicles In-Use Performance Ratio Monitoring Data

In recent times, the governments are pushing for stringent emission regulations. These regulations call for reduction of pollutants as well as monitoring of engine components which are critical for emission control. Monitoring these emission critical engine components are to be done in real world driving conditions. The In-Use Performance Ratio Monitoring (IUPRm) framework quantifies how often onboard diagnostic systems check these components within defined boundaries for each vehicle. IUPRm is divided into several monitoring groups like catalyst monitoring, oxygen sensor monitoring, exhaust gas recirculation (EGR) monitoring, gasoline particulate filter monitoring and others. These groups are differentiated based on fuel type, engine technologies and exhaust treatment system configurations. For an Automotive manufacturer analyzing these parameters across large vehicle fleets is a complex and data intensive task. To address this, a user-friendly application was developed in-house, which includes the new method based on Artificial Intelligence and Machine Learning algorithms for automating complex IUPRm Data analysis. This method contains techniques, such as structured decision tree based classification and rule based logic algorithms for automating classification of vehicles into a particular OBD family from a large and mixed fleet data and filtering all anomalies in the data. The K-Means clustering along with the elbow logic, groups the vehicles with similar IUPRm ratios and checks if selected vehicles meets the compliance requirement. This application enables to automate and speed up large scale IUPRm data analysis by reducing manual effort and enhancing overall efficiency. The newly developed method also provides automated reports. This paper explains selection and working principles of different algorithms and techniques used in development of this application for efficient IUPRm monitoring.