Ruta, Andrzej (Stellantis), Chyrowicz, Katarzyna (Stellantis), Gentile, Mattia Gianfranco (Stellantis)

Preventive Vehicle Battery Maintenance Using Recurrent Neural Networks

Scheduled for presentation during the Poster Session "Data Mining and Data Analysis" (FrAT14), Friday, September 27, 2024, 10:30−12:30, Foyer

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on October 14, 2024

Abstract

This work presents insights from a vehicle battery life modeling project launched at Stellantis in pursuit of its strategic Dare Forward 2030 goal which, among others, includes full powertrain electrification. The developed solution aims to predict a broad spectrum of battery failures weeks in advance to reduce the risk of sudden vehicle immobilization, as well as to cut down on the maintenance and warranty costs. The underlying classifier is built upon a recurrent neural network architecture with roots in computer vision. It is fed by temporal sequences of window-aggregated multi-dimensional signals composed of raw measurements that vehicles continuously emit and send over the air. Prediction results obtained using the above setup in two scenarios involving high-voltage traction battery and 12V battery are presented and discussed. Practical tips on deploying and using this type of predictive model, as well as quantify its business value are also given.