Open Access

Driving Behavior Modelling Framework for Intelligent Powertrain Health Management

Journal Article
ISSN: 1946-3936, e-ISSN: 1946-3944
Published August 09, 2022 by SAE International in United States
Driving Behavior Modelling Framework for Intelligent Powertrain
                    Health Management
Citation: Doikin, A., Campean, F., Priest, M., Angiolini, E. et al., "Driving Behavior Modelling Framework for Intelligent Powertrain Health Management," SAE Int. J. Engines 16(4):2023,
Language: English


The implementation of an intelligent powertrain health management relies on robust prognostics modelling. However, prognostic capability is often limited due to unknown future operating conditions, which vary with duty cycles and individual driver behaviors. On the other hand, the growing availability of data pertaining to vehicle usage allows advanced modelling of usage patterns and driver behaviors, bringing optimization opportunities for powertrain operation and health management. This article introduces a methodology for driving behavior modelling, underpinned by Machine Learning (ML) classification algorithms, generating model-based predictive insight for intelligent powertrain health management strategies. Specifically, the aim is to learn the patterns of driving behavior and predict characteristics for the short-term future operating conditions as a basis for enhanced control strategies to optimize energy efficiency and system reliability. A case study of an automotive emissions aftertreatment system is used to comprehensively demonstrate the proposed framework. The case study illustrates the approach for integrating predictive insight from ML deployed on real-world trip behavior data, in conjunction with a reliability-based model of the operational behavior of a particulate filter, to propose an intelligent active regeneration control strategy for improved efficiency and reliability performance. The effectiveness of the proposed strategy was demonstrated on an industry standard model-in-the-loop setup with a representative sample of real-world vehicle driving data.