Gas detector network is an important layer of protection in process facilities for prevention gas leakage accidents. But traditional standards just provide basic principles for the installation of detectors. In this study, three stochastic programming (SP) models are developed and contrasted, namely minimal detection time P-Median model (MDTP), minimal leakage concentration P-Median model (MLCP), and minimal individual risk P-median model (MIRP). Meanwhile all possible leak scenarios are identified based on the combination of wind field set and leakage sources. And clustering analysis is used to filter similar scenarios and select representative leak scenarios. The leak consequences are predicted by computational fluid dynamics (CFD) method and the results are served as the input data of these SP models. A case study is carried out in a diesel hydrogenation unit.

气体检测仪网络是过程设施中重要的保护层，可防止气体泄漏事故的发生。但是传统标准仅提供了安装探测器的基本原理。在这项研究中，开发并对比了三种随机编程（SP）模型，即最小检测时间P中间值模型（MDTP），最小泄漏浓度P中间值模型（MLCP）和最小个人风险P中间值模型（MIRP） 。同时，根据风场和泄漏源的组合，确定所有可能的泄漏情况。聚类分析用于过滤相似的情况并选择代表性的泄漏情况。通过计算流体力学（CFD）方法预测泄漏后果，并将结果用作这些SP模型的输入数据。