Although multi-objective optimization of integrated distillation processes can substantially improve process design, the nonlinearity and complexity of the process results in high computational expense for optimization. Here, an approach incorporating surrogate modeling into multi-objective optimization is proposed, in which surrogate models for function evaluation are constructed by using the RBF neural network. Central composite design was adopted as a sampling strategy and surrogate models were individually constructed for different optimization objectives to improve prediction accuracy. Multi-objective bat algorithm was set as an optimizer to obtain the Pareto front. This surrogate modeling-based multi-objective optimization approach was applied to the design of dividing wall column and side-reactor column configuration, and the satisfied design options realizing the trade-offs between capital and operating costs were successfully obtained thereafter.

尽管集成蒸馏过程的多目标优化可以大大改善过程设计，但是过程的非线性和复杂性导致优化需要大量的计算费用。在此，提出了一种将代理模型集成到多目标优化中的方法，其中使用RBF神经网络构造用于功能评估的代理模型。采用中央复合设计作为抽样策略，并针对不同的优化目标分别构建了替代模型，以提高预测准确性。多目标蝙蝠算法被设置为优化器以获得帕累托锋。将该基于替代模型的多目标优化方法应用于分隔壁塔和侧反应器塔构型的设计，