Predicting Materials Failure Before It Happens: A Critical Review of AI for Multimodal Degradation Forecasting Under Coupled Environmental Stressors

Infrastructure, energy, transportation and industrial materials are increasingly exposed to coupled thermal, moisture, chemical and mechanical stressors that can interact to produce nonlinear and history-dependent degradation. Conventional periodic inspection and single-stressor testing often provide limited information about the progression of damage and may fail to capture interactions that influence long-term material performance and failure. Artificial intelligence (AI) provides an opportunity to address these limitations by integrating heterogeneous information from sensors, imaging, experimental measurements, environmental monitoring and physics-based simulations. However, identifying existing damage is fundamentally different from forecasting its future evolution. This critical review examines the transition from AI-based damage detection toward temporal degradation forecasting, remaining useful life estimation and failure probability prediction under coupled environmental and mechanical conditions. Machine learning, deep learning, multimodal learning, physics-informed AI, digital twins, uncertainty quantification and explainability are critically evaluated in relation to their potential and limitations for materials degradation forecasting. Major translational challenges include limited longitudinal datasets, inconsistent experimental protocols, multimodal data alignment, limited cross-material and cross-environment generalization, uncertainty representation, interpretability and the laboratory-to-field gap. In response, an integrated framework is proposed that links environmental exposure and material state evolution with multimodal sensing, AI-based forecasting, physical constraints and uncertainty estimation. This framework provides a pathway toward proactive intervention by connecting predicted degradation trajectories and failure probability with risk-informed inspection and maintenance decisions.