Abstract:

Safety evaluation is a prerequisite for the normal operation of chemical processes. Although the traditional quantitative risk assessment method can reduce the frequency of accidents, it relies heavily on the experience of experts and is difficult to assess the potential risk of accidents caused by dynamic chemical conditions. To address this problem, an intelligent quantitative risk assessment method based on dynamic simulation (DQRA-BiLSTM) was proposed in this paper. The process under study was first simulated with process simulation software to obtain a dynamic data set under abnormal conditions. Then, the potential relationships between variables were deeply learned using bi-directional long short-term memory (Bi-LSTM) to characterize the complex mechanistic relationships between predictor variables that had a direct impact on the severity of accident hazards. A reliable control scheme was proposed to ensure the safe operation of the process. The method was applied to the system of carbon di-hydrogenation and deethanization in the ethylene separation process, and the experimental results showed that the model had good performance for the dynamic risk assessment of the ethylene separation process, which had some practical application value.

Key words: chemical processes, simulation, dynamic quantitative risk assessment, safety, risk prediction

摘要：

安全评价是化工过程正常运行的前提，传统的定量风险评估方法虽然可以减少事故发生的频率，但严重依赖专家的经验，难以评估动态化学条件引起的潜在事故风险。针对这一问题，本文提出了一种基于动态模拟的智能定量风险评估方法（DQRA-BiLSTM）。首先，用流程模拟软件对所研究的过程进行模拟，得到异常条件下的动态数据集。然后，利用双向长短期记忆（Bi-LSTM）深度学习变量之间的潜在关系，表征对事故危害严重程度有直接影响的预测变量之间复杂的机制关系，提出了可靠的控制方案，保证了工艺的安全运行。将该方法应用于乙烯分离流程中的碳二加氢和脱乙烷系统。实验结果表明，该模型对乙烯分离流程的动态风险评估具有良好的性能，具有一定的实际应用价值。

关键词: 化学过程, 模拟, 动态定量风险评估, 安全, 风险预测