SC075A

Submit Paper » Description The emergence of large language models (LLMs) and multimodal foundation models has revolutionized traditional approaches to risk assessment and predictive maintenance in engineering systems. These AI systems demonstrate unprecedented capabilities in processing heterogeneous data streams - from textual maintenance logs and equipment manuals to time-series sensor data and visual inspection reports - enabling comprehensive fault diagnosis across civil infrastructure (e.g., bridges, dams, power grids) and mechanical systems (e.g., rotating machinery, HVAC systems, industrial robots). However, the deployment of LLMs in safety-critical engineering applications introduces profound challenges that demand urgent research attention. First, the probabilistic nature of LLMs leads to inherent epistemic uncertainty in fault diagnosis, compounded by issues of model hallucination when interpreting sparse or noisy field data. Second, the black-box decision-making process of current models creates significant barriers to engineering validation, particularly in regulated industries where traceable risk assessment is mandatory. Third, the dynamic operating conditions of engineering systems (e.g., seasonal load variations in civil infrastructure or wear progression in mechanical components) require continuous model adaptation while maintaining operational safety margins. This special issue seeks to address these challenges through cutting-edge research at the intersection of AI reliability and engineering risk management. ...