Effective dynamic risk assessment is crucial for identifying process hazards and preventing accidents. The rapid development of modern technology urgently requires the development of new dynamic risk assessment methods to address current societal needs, especially in situations where a large amount of real-time data is available. Consequently, this manuscript presents an improved support vector machine model (SVM), integrated mathematical modeling, to facilitate dynamic risk assessment for hazardous chemical warehouses (HCWs). Firstly, an indicator framework and a quantitative risk factor table are crafted. Then a mathematical model using knowledge graph and DEMATEL-variable weight theory (VWT) is established, which combines accident information with VWT to optimize the weight calculation and provides reliable prior knowledge for the improved SVM model. Ultimately, the improved SVM model is crafted by the establishment of an improved hybrid kernel of polynomial and radial basis function kernel, fine-tuning its hyperparameters with particle swarm optimization, and utilizing sensitivity analysis to simplify the model. The model is formulated to accommodate the intricate non-linear relationships within indicators and dynamic risk, thereby establishing dynamic risk assessment rules. Through a case study, the improved SVM model demonstrates higher precision in dynamic risk prediction compared to original models. The results substantiate that the model can provide innovative insight and methodology for the precise evaluation of dynamic risk in HCWs.

中文翻译：

---

基于改进SVM和数学方法的危化品仓库动态风险评估新方法

有效的动态风险评估对于识别过程危险和预防事故至关重要。现代技术的快速发展迫切需要开发新的动态风险评估方法来满足当前的社会需求，特别是在有大量实时数据的情况下。因此，本文提出了一种改进的支持向量机模型（SVM）和集成数学建模，以促进危险化学品仓库（HCW）的动态风险评估。首先，制定指标框架和定量风险因素表。然后利用知识图和DEMATEL-变权理论(VWT)建立数学模型，将事故信息与VWT相结合来优化权重计算，为改进的SVM模型提供可靠的先验知识。最终，通过建立多项式和径向基函数核的改进混合核，通过粒子群优化对其超参数进行微调，并利用敏感性分析来简化模型，从而构建出改进的SVM模型。该模型的制定是为了适应指标和动态风险之间复杂的非线性关系，从而建立动态风险评估规则。通过案例研究，改进后的SVM模型比原始模型表现出更高的动态风险预测精度。结果证实该模型可以为医护人员动态风险的精确评估提供创​​新的见解和方法。