Application of Multi-fidelity Surrogate Models to the Noisy Optimization Problems of Carbon Fiber Polymerization Process Parameters

Carbon fiber is an innovative and strategic material for aeros-pace and other critical areas, and the polymerization process is one of the most critical processes for the production of carbon fiber. To improve the quality of precursor fibers and save costs, the optimization of process parameters is one of the important aspects. Considering the complex production plant environment and other circumstances of the carbon fiber polymerization process, a noisy multi-objective optimization problem model on the molecular weight of polyacrylonitrile (PAN) and the monomer conversion rate is developed. In order to solve the noisy optimization problems of carbon fiber polymerization process parameters (NOPCFPPP), this paper proposes an improved multi-fidelity optimization (IMFO) algorithm, aiming to reduce the impact of the noise on the evolution of a smaller evaluation budget, so as to further solve and optimize the process parameters, and hope to be able to provide some theoretical guidance in the actual polymerization process of carbon fibers. Experimental results show that the algorithm proposed in this paper is highly competitive with existing algorithms for noisy multi-objective optimization problems.