Dynamic Constrained Robust Optimization over Time for Operational Indices of Pre-oxidation Process

The pre-oxidation process of carbon fiber, as one of its important production processes, can affect its physical properties directly. In this paper, in order to provide more valuable theoretical guidance for the actual pre-oxidation process, we construct a dynamic constrained multi-objective optimization model of the pre-oxidation process. For the treatment of dynamic constraints, we use the strategy based on penalty function, extending the constraint violation value in the objective space from a single feasibility deviation value to a weighted sum of feasibility deviation value and non-dominated deviation value, and propose the dynamic constrained multi-objective evolutionary algorithm considering non-dominated deviation (DCMOEA-ND). We then incorporate the new robustness definition and propose the dynamic constrained robust optimization over time (DCROOT), which is designed to obtain high-quality filaments and reduce energy consumption while reducing the switching cost of solutions. Experimental results show that DCMOEA-ND can obtain Pareto optimal set (POS) with better convergence and distribution, and the robust solutions obtained by DCROOT have better performance than other algorithms.