Remote vehicle state of health monitoring and its application to vehicle no-start prediction

This paper reports a recent effort at GM to develop a remote vehicle diagnostics service under a previously proposed framework of connected vehicle diagnostics and prognostics. An algorithm development methodology combining the physics-based approach and the data-driven approach is presented to identify, select, and calibrate failure precursors to predict vehicle no-start due to battery failures. Initial results based on real field data are promising. Also presented is a proposed implementation solution that supports the cost and performance optimization of remote vehicle no-start prediction.