The increasing complexity of high-value, safety-critical engineered systems—such as gas turbines, jet engines, nuclear reactors, and aircraft—demands exceptional performance and reliability. Despite rigorous testing of individual components, early-life failures stemming from manufacturing processes continue to pose significant challenges, often going undetected until the integration phase. These failures not only reduce system reliability but also lead to substantial energy consumption and increased operational costs due to the need for repeated test-bed trials. 

This research tackles the critical issue of diagnosing early-life failures by developing a novel hybrid approach that combines probabilistic graph theory with artificial intelligence. Traditional physics-based failure models often struggle with limited data and the complex interdependencies between system components and manufacturing processes. To address these limitations, our work integrates fault tree analysis with Bayesian methods, allowing for a structured, probabilistic representation of failure pathways and enhancing predictive accuracy.

By systematically mapping the relationships between system components and their manufacturing origins, this approach establishes a robust framework for diagnosing failures and making reliable inferences. This research advances the field of intelligent reliability engineering by providing valuable insights into manufacturing-induced failures in complex, high-stakes engineered systems.

PhD Student

Hanu Indiran

Supervisor

Ajith Parlikad

Industrial Sponsor

Siemens