Scholarlink

<p>Complex distillation processes can often be effectively optimized using meta-heuristic algorithms. However, during optimization procedure, a large number of infeasible solutions are generated, hindering efficient exploration of feasible, high-performance regions of the search space. In this study, we propose a data-driven identification and adaptive directed correction strategy for handling infeasible solutions, and on this basis, develop an efficient multi-objective optimization framework (MO-DIDC) for complex distillation processes. By identifying infeasible solutions that closely resemble high-performance ones, the framework leverages them to accelerate convergence to optimal designs. A surrogate model is trained to distinguish high- and low-performance solutions and is then used to identify potentially high-performance candidates within the infeasible set. Through similarity analysis, the most influential variable is selected for correction to generate new promising solutions. This strategy reduces unnecessary exploration of infeasible regions and concentrates computational effort on feasible, high-quality solutions. Demonstrated on a side-stream double-column extractive distillation system and a four-column extractive distillation system, the proposed optimization framework outperforms a widely used genetic algorithm while substantially improving computational efficiency, achieving optimization time reductions of 35.3% and 20.8%, respectively. Overall, the proposed MO-DIDC framework provides an effective and computationally efficient tool for the optimization of complex distillation processes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An efficient multi-objective optimization framework based on data-driven identification and adaptive directed correction for complex distillation processes

  • 期刊论文

  • 发表日期：

    2026年4月1日

查看全文
  • Fucheng Xu,
  • Lu Yang,
  • Zihao Wang,
  • Tao Shi,
  • Rongsheng Lin,
  • Bohong Wang,
  • Hengcong Tao,
  • Zhiliang Cheng,
  • Weifeng Shen

摘要

Complex distillation processes can often be effectively optimized using meta-heuristic algorithms. However, during optimization procedure, a large number of infeasible solutions are generated, hindering efficient exploration of feasible, high-performance regions of the search space. In this study, we propose a data-driven identification and adaptive directed correction strategy for handling infeasible solutions, and on this basis, develop an efficient multi-objective optimization framework (MO-DIDC) for complex distillation processes. By identifying infeasible solutions that closely resemble high-performance ones, the framework leverages them to accelerate convergence to optimal designs. A surrogate model is trained to distinguish high- and low-performance solutions and is then used to identify potentially high-performance candidates within the infeasible set. Through similarity analysis, the most influential variable is selected for correction to generate new promising solutions. This strategy reduces unnecessary exploration of infeasible regions and concentrates computational effort on feasible, high-quality solutions. Demonstrated on a side-stream double-column extractive distillation system and a four-column extractive distillation system, the proposed optimization framework outperforms a widely used genetic algorithm while substantially improving computational efficiency, achieving optimization time reductions of 35.3% and 20.8%, respectively. Overall, the proposed MO-DIDC framework provides an effective and computationally efficient tool for the optimization of complex distillation processes.