AVL Research Project: Intelligent Fault Tree Construction for Automotive Diagnostics

Project Overview (January 2023 – December 2023)

The AVL Research Project was a collaborative effort between Graz University of Technology and AVL List GmbH, focusing on developing intelligent methods to support fault tree construction for automotive diagnostics. The project’s success led directly to securing additional funding through the FFG proposal “HybridAIR,” extending research activities for an additional three years.

Research Objectives

• Address the increasing complexity in automotive systems by automating the traditionally manual fault tree analysis (FTA) process.
• Enhance diagnostic accuracy and efficiency using sequential recommendation algorithms.
• Leverage contextual insights by incorporating text embeddings derived from pre-trained language models.

Key Contributions

• Reformulated fault tree creation as a sequential recommendation task, significantly improving fault diagnosis accuracy.
• Developed an innovative framework integrating sequential recommendation and contextual embeddings, preserving expert oversight while reducing manual effort.
• Conducted comprehensive experiments validating model effectiveness, leading to a publication submitted to the high-impact journal Reliability Engineering & System Safety.

Outcomes and Impact

• Demonstrated significant improvements in fault tree accuracy and construction efficiency, validated on AVL’s extensive automotive diagnostic dataset.
• Provided a scalable and flexible methodology applicable to other safety-critical engineering domains beyond automotive diagnostics.
• Played a key role in securing long-term research funding (FFG HybridAIR) to continue advancing diagnostic capabilities and machine learning integration.

Personal Role

• Acted as the main contributor and senior researcher on the project, leading both the methodological development and experimental evaluation.
• Collaborated closely with AVL experts and university researchers to align research outputs with practical automotive industry requirements.
• Authored and submitted the key research publication outlining the developed methods, experimental validation, and future research directions.