Explosion is a significant threat to chemical process safety, which often occurs suddenly with a wide range of impact, resulting in catastrophic disasters. Predictions of explosion hazards are required for regional safety assessment and disaster resilience enhancement of chemical industrial parks (CIPs). This paper, therefore, aims to develop a conceptual framework for systematically analyzing explosion evolutions and quantifying the potential hazards by combining system-theoretic process analysis (STPA) method with numerical simulation. The proposed framework consists of 5 steps: (i) establishment of hierarchical safe control structures (SCSs) of important chemical processing zones in the CIP, (ii) computational fluid dynamics (CFD) modeling for potential explosion evolutions in each zone by changing the examined parameters randomly, (iii) development of a convolutional neural network (CNN) prediction model through constant self-learning of CFD pressure field data, (iv) comprehensive assessment of blast damage by incorporating the outputs of the above numerical models into existing evaluation methods, (v) identification of unsafety control actions and causes, and safety constraints for the improvement of SCSs. Provided with monitoring data, the developed analysis architecture can predict explosion process hazards and recommend appropriate safety strategies in real time. This would service the multi-level requirements for explosion prevention and protection, supporting better-informed decision-making. The paper describes the concepts and implementation process of the method as a first step.

爆炸是对化工过程安全的重大威胁，往往是突然发生，影响范围广泛，造成灾难性的灾难。化工园区（CIP）的区域安全评估和抗灾能力增强需要对爆炸危险进行预测。因此，本文旨在通过将系统理论过程分析 (STPA) 方法与数值模拟相结合，开发一个系统分析爆炸演化和量化潜在危险的概念框架。所提出的框架由 5 个步骤组成：（i）在 CIP 中建立重要化学加工区的分层安全控制结构（SCS），（ii）通过改变检查对象，为每个区域的潜在爆炸演化建立计算流体动力学（CFD）模型。参数随机，(CNN) 通过不断自学习 CFD 压力场数据的预测模型，(iv) 通过将上述数值模型的输出整合到现有评估方法中来综合评估爆炸破坏，(v) 识别不安全控制行为和原因，以及改进 SCS 的安全约束。提供监控数据，开发的分析架构可以预测爆炸过程危害并实时推荐适当的安全策略。这将服务于防爆和保护的多层次要求，支持更明智的决策。本文首先描述了该方法的概念和实现过程。