Development of a Prognostic Health Monitoring System to enable Predictive Maintenance in an Electric Two-wheeler Vehicle

Quality management of high voltage electric power train components, such as Lithium-ion Batteries, Chargers, Traction Motors, Motor Control Units, and DC-DC converters, is crucial for the safety of electric two-wheelers. Defects in design, process, or service can lead to safety hazards. To address these concerns, it is imperative to adopt a preventive approach. With the adoption of electric vehicles still at a nascent stage, it is a continuous critical activity to understand electric vehicle sub-system performance parameters and necessary improvements. This technical paper outlines the development and implementation of a Prognostic Health Monitoring System (PHMS) that uses Artificial Intelligence and Machine Learning techniques to monitor vehicle performance. The system detects fault symptoms by validating the actual state against the intended function, thereby indicating non-safe operation in the vehicle. The system can initiate automatic alerts about potential issues to initiate actions on such vehicles before customers experience mobility or safety issues. Vital granular performance data is generated from each vehicle subsystem, captured using an edge telematics device, and transferred to a cloud application for analysis. Predictive analysis, utilizing the significant amount of data acquired from the entire fleet of vehicles, enables data-driven detection of potential future faults in near real-time. The system also investigates the interaction between user and vehicle to establish the impact of individual driving styles on component performance and vehicle health. The key challenges in data acquisition, such as loss, non-uniformity, jump, low frequency, etc., are addressed by the software algorithm, which performs multi-variate relationship analysis between vehicle parameters to generate metrics with high quality, efficiency, and accuracy. The Natural Language Processing (NLP) techniques built with this system enable the machine to provide tailored analysis based on the user’s request. Compared to conventional infrequent manual processes that use limited data sets and time-consuming complex diagnostic analysis, the system automates the process, allowing for continuous improvements to quality, safety, and reliability.