Expensive Multi-Objective of Pre-Oxidation Process Parameter Optimization Considering Heterogeneity

The precursor pre-oxidation process, as a critical link in carbon fiber production, stabilizes the properties of carbon fiber precursors, lays the foundation for the carbonization process, and is also a step that can directly determine the structural properties of carbon fiber, so it has been widely concerned by academic circles. Evaluating the density and strength of pre-oxidized filaments is expensive, time-consuming, and has different cost consumption, leading to heterogeneous problems in optimizing pre-oxidized process parameters. An optimization algorithm of pre-oxidation process parameters based on collaborative migration agent genetic programming is proposed to solve this problem. This method establishes a single-objective multi-precision surrogate model of the pre-oxidation process and a collaborative model between the two targets. The cooperative model predicts the high-cost target to reduce the operating cost in the calculation process. Our experimental results show that the proposed algorithm performs well in optimizing pre-oxidation process parameters.