A novel approach for explainable commercial electric vehicle fault detection

Context

Ensuring the reliable operation of commercial electric vehicles and their systems remains a major challenge, especially in the diagnostics field. In many cases, traditional diagnostic systems fail to identify the true root cause of a failure. This often results in repeated or unresolved issues, as the actual problem is only discovered through detailed post-event analysis. Furthermore, multiple simultaneous faults only reveal the visible symptoms, rather than the underlying source of the failure. To address these challenges, the goal of this thesis is to develop a novel approach for explainable fault detection in commercial electric vehicles. By integrating Machine Learning (ML) and explainable Artificial Intelligence (XAI), the proposed method shall aim to improve diagnostic accuracy while providing interpretable insights into detected faults. This includes identifying likely root causes, the effect of the failure, and the systems affected. Such an approach would enable more transparent decision-making, supporting maintenance and control strategies that can address the primary causes of faults. The expected outcome is to automate a diagnostic framework that does precise fault detection, providing an explanation of why and how they occurred, facilitating more efficient fault resolution and improved system safety.

Goals

• State of the Art of the used ML approaches for fault detection and explainable AI for cause investigation in the literature
• Investigation of suitable ML and explainable AI approaches for the use case of explainable fault detection of commercial EV faults
• Development and implementation of a novel explainable fault detection methodology
• Evaluation and validation of the novel proposed explainable fault detection methodology on real recorded faults

Requirements

• Interest in the development of AI/ML systems
• Basic knowledge and experience in data processing
• Programming skills (Python/Matlab)