In this study, a multi-factor reaction thermal risk model and a safety optimization method based on partial least squares (PLS) method were proposed. First, orthogonal experiments of key influencing factors were designed to obtain the correlation data between influencing factors and characteristic indicators of thermal risk. Then, the PLS method was employed to establish the reaction thermal risk model between them. Finally, this model was utilized to predict the worst-case scenario and optimize the permissible scope of influencing factors. The proposed method was applied to an arylamine diazo reaction as a representative example. The influencing factors included sulfuric acid concentration and diazonium salt concentration, while the characteristic indicator of thermal risk was the initiation temperature of adiabatic induction period of x hours (T_Dx). Setting T_D24 > 30 °C, T_D8 > 35 °C, T_D4 > 40 °C, and T_D1 > 45 °C as constraints, the optimal result was sulfuric acid concentration ≥ 72.8% and diazonium salt concentration ≤ 41.2%. Moreover, if 5% industrial deviation was tolerated, the optimal result was sulfuric acid concentration ≥ 76.2% and diazonium salt concentration ≤ 38.2%.

本研究提出了基于偏最小二乘法（PLS）的多因素反应热风险模型和安全优化方法。首先，设计关键影响因素正交实验，得到影响因素与热风险特征指标的相关性数据。然后，采用PLS方法建立了它们之间的反应热风险模型。最后，利用该模型预测最坏情况并优化影响因素的允许范围。所提出的方法以芳胺重氮反应为例进行了应用。影响因素包括硫酸浓度和重氮盐浓度，而热风险的特征指标是x小时绝热诱导期的起始温度（T _{D x}）。_以T D24 > 30 °C、T _D8 > 35 °C、T _D4 > 40 °C和T _D1 > 45 °C为约束条件，最佳结果为硫酸浓度≥72.8%，重氮盐浓度≤41.2%。此外，如果允许 5% 的工业偏差，最佳结果是硫酸浓度≥76.2%，重氮盐浓度≤38.2%。