The integrated distillation configurations have an enormous number of alternative flowsheets leading to great challenges for process synthesis. The superstructure-based approach can be the countermeasure but it is difficult to solve the complex networks using rigorous models due to the nonconvexity and nonlinearity. Although stochastic algorithms can robustly find high-quality solutions for a single predefined flowsheet, the performance will still degrade when dealing with complex superstructure models. This paper introduces an adaptive strategy to determine the substructure branches of the superstructure, which can reduce the complexity of the superstructure models optimized by stochastic algorithms. The improved method selects the downstream flowsheet based on the calculation results of the upstream modules, instead of predefining the branches before the simulation. Thereby, stochastic algorithms could optimize the simplified superstructure model with fewer sequence variables more robustly and efficiently. A multiple-effect distillation system and a double side-stream distillation system separating aromatics are optimized as case studies. The results show that the improved method, compared with the conventional approach, could find the optimized side-stream design with 17.8% lower cost more robustly, and find the optimized multi-effect distillation design with 2.4% lower cost within the same iterations.

集成蒸馏配置有大量的替代流程，这给工艺合成带来了巨大挑战。基于上层结构的方法可以作为对策，但由于非凸性和非线性，难以使用严格的模型求解复杂网络。尽管随机算法可以为单个预定义的流程稳健地找到高质量的解决方案，但在处理复杂的上层结构模型时，性能仍然会下降。本文引入了一种自适应策略来确定上部结构的子结构分支，该策略可以降低通过随机算法优化的上部结构模型的复杂度。改进方法根据上游模块的计算结果选择下游流程，而不是在模拟之前预先定义分支。因此，随机算法可以更稳健、更有效地优化具有更少序列变量的简化上层结构模型。作为案例研究优化了多效蒸馏系统和双侧流蒸馏系统分离芳烃。结果表明，与传统方法相比，改进方法可以更稳健地找到成本降低 17.8% 的优化侧流设计，并在相同的迭代中找到成本降低 2.4% 的优化多效精馏设计。作为案例研究优化了多效蒸馏系统和双侧流蒸馏系统分离芳烃。结果表明，与传统方法相比，改进方法可以更稳健地找到成本降低 17.8% 的优化侧流设计，并在相同的迭代中找到成本降低 2.4% 的优化多效精馏设计。作为案例研究优化了多效蒸馏系统和双侧流蒸馏系统分离芳烃。结果表明，与传统方法相比，改进方法可以更稳健地找到成本降低 17.8% 的优化侧流设计，并在相同的迭代中找到成本降低 2.4% 的优化多效精馏设计。