In high-risk industrial settings, the proliferation of sensor data provides crucial support for fundamental research on safety events (SEs) and for precise risk analysis. However, existing data-driven methods struggle to reveal the nonlinear couplings among multiple factors and lack a systematic framework to explain how these factors jointly contribute to different types of SEs. To address these limitations, this study proposes a theory-data integrated model for multitype SE risk identification and assessment. From a cross-scale emergence perspective, the study elucidates that SEs arise from a chain-evolution process driven by the nonlinear coupling of diverse environmental factors and constructs a theoretical framework describing hierarchical factor associations and cross-scale interactions. Building upon this theoretical foundation, a data-driven risk identification and assessment model (RIAM) is established. This model quantifies the contributions of measurable environmental factors (MEFs) that significantly influence SEs through online learning. Experimental results demonstrate that the proposed model effectively captures the cumulative risk effects arising from multi-factor coupling, thereby enhancing both event identification accuracy and model interpretability. This study provides a novel perspective and methodological foundation for SE prediction and integrated prevention in complex industrial environments.
Liu et al. (Mon,) studied this question.