Surrogate-assisted decomposition multi-objective evolutionary algorithm for parameters optimization in polyester fiber polymerization process

This study presents the development of a two-stage adaptive decomposition multi-objective evolutionary algorithm (TSAMOEAD) designed to optimize quality control in industrial aggregation processes, such as polyester fiber production. To address the time delay issue in quality indicator detection caused by production continuity, we first introduce an improved Informer model. This model predicts multiple quality indicators in real time from multivariate time series data, serving as a surrogate for process parameter optimization. Additionally, we enhance the CCF lag time estimation method to account for time delays in quality control, ensuring that adjustments to process parameters are made within the available time frame. In the second part of the study, we develop a two-stage adaptive decomposition-based multi-objective evolutionary algorithm to optimize polymerization process parameters. The first stage involves rapidly approximating the Pareto front using specific weight vectors and genetic operators. The second stage enhances solution diversity and convergence through the use of adaptive weight vectors and operators. To simplify the selection of optimal solutions from the Pareto front, we propose an indicator-based screening method that efficiently identifies the most suitable adjustment schemes. Experimental results demonstrate that our approach accurately predicts quality indicators and provides effective parameter adjustment strategies that meet production requirements.