Abstract

Safety-critical engineering systems, such as autonomous vehicles, industrial automation, and aerospace control, demand highly reliable decision-making under uncertain conditions. While Artificial Intelligence (AI) and machine learning models provide powerful predictive and control capabilities, their “black-box” nature often limits trust, accountability, and regulatory compliance. This paper explores the role of Explainable AI (XAI) in enhancing transparency, interpretability, and safety in engineering applications where failure can lead to catastrophic consequences. We present a framework combining model-agnostic explanation methods, such as SHAP and LIME, with domain-specific knowledge to provide actionable insights into AI-driven decisions. Case studies on autonomous vehicle navigation, industrial robotic arms, and aircraft fault detection demonstrate that XAI improves operator understanding, supports risk assessment, and facilitates compliance with safety standards. The integration of explainability with real-time monitoring enables identification of potential hazards, anomaly detection, and post-incident analysis. Results indicate that XAI not only enhances user trust and system robustness but also accelerates adoption of AI in regulatory-sensitive domains. This research highlights the importance of interpretable AI as a core component of future safety-critical engineering systems.