Predictive Maintenance as Vehicles Become More Software Defined

The automotive industry is witnessing its next phase of transformation. Vehicles with electric drivetrains and automated features are becoming advanced and sophisticated with continuous over-the-air software updates. For these complex software-defined vehicles, prognostics and predictive maintenance become ever more critical than before. This presentation proposes a machine learning based framework created and deployed through MATLAB^®. It utilizes minimally labeled vehicle data and identifies and flags anomalous behavior that went undetected or got introduced with the aging of components. This framework can be adopted for large real-time or time-series data for early identification of failures and can be deployed on the cloud or vehicle edge.

Highlights include:

• Handling and preprocessing the big data.
• Developing the AI models using the MATLAB toolchain to detect anomalies in the powertrain subsystem and proactively flag them to customers or OEMs.
• Deploying the predictive maintenance solution on the cloud, ensuring scalability and accessibility from anywhere.
• Designing app-based dashboards in MATLAB with intuitive visualizations, empowering users to make informed decisions about maintenance actions.