EVE: Extending life of Vehicles within Electromobility era

Important results from the project

The EVE project pioneered electromobility insights via operational data. The project was aimed at leveraging machine learning for predictive maintenance and providing services related to electromobility. In particular, the project developed general lifetime models for vital components in the electrical drivetrain of buses. Utilizing transfer learning improved prediction accuracy compensating for data diversity. The project also explored energy efficiency and charging patterns applications.

Expected long term effects

Project outcomes were shared through reports with engineering and management, guiding longer-lasting product and service development. Electric component status models has been integrated into Volvo Monitoring Systems to improve detection of wear. High-quality publications in domain adaptation and evolutionary methods for health estimation of electric components emerged. This led to two upcoming Ph.D. defenses, a disseminated licentiate thesis, and presentations at national and international conferences.

Approach and implementation

EVE monitored the behavior of the vehicle using on-board signals. The analyses are as follows: Survival analysis to analyze battery replacement strategy. Classification and regression methods to model the end of life and remaining useful life of the components. Evolutionary algorithms to select domain invariant features. Domain Adversarial Neural Networks for estimating the health status of the batteries in the presence of multiple different domains. Clustering to find energy consumption patterns.