Causality-enhanced system reliability and safety analysis: An overview

Ensuring system reliability and safety is paramount for all industries, particularly critical ones like aerospace, nuclear, chemical where failures can lead to catastrophic consequences. Traditional causality-based methods, such as Fault Tree Analysis (FTA) and causal Bayesian networks, have long been employed but face limitations in scalability and adaptability. Recently, data-driven approaches have emerged as a powerful way of mapping complex and nonlinear relations, but these approaches often lack interpretability and struggle with accuracy under data distribution shifts occurring during system life. This paper systematically explores the integration of causality theory into system reliability and safety for bridging the gap between conventional techniques and modern advancements in causal inference and reasoning. We analyze three key approaches: causal discovery algorithms, causality-guided models and causal graph-based methods, covering their theoretical foundations and research applications. The review of the literatures highlights that causality-enhanced frameworks facilitate gaining deeper insights into system behavior, enabling improvements on model robustness, generalizability and transparency while mitigating spurious correlations. Furthermore, challenges like cross-scale and evolving causal dynamics motivate further dedication into causality-enhanced research. By embedding causality into system reliability and safety, this work underscores its potential in advancing system predictive maintenance, risk assessment and resilient design, paving the way towards trustworthy reliability and safety analysis.