Dynamic prediction models and optimization of polyacrylonitrile (PAN) stabilization processes for production of carbon fiber
Thermal stabilization process of polyacrylonitrile (PAN) is the slowest and the most energy-consuming step in carbon fiber production. As such, in industrial production of carbon fiber, this step is considered as a major bottleneck in the whole process. Stabilization process parameters are usually many in number and highly constrained, leading to high uncertainty. The goal of this paper is to study and analyze the carbon fiber thermal stabilization process through presenting several effective dynamic models for the prediction of the process. The key point with using dynamic models is that using an evolutionary search technique, the heat of reaction can be optimized. The employed components of the study are Levenberg-Marquardt algorithm (LMA)-neural network (LMA-NN), Gauss-Newton (GN)-curve fitting, Taylor polynomial method, and a genetic algorithm. The results show that the procedure can effectively optimize a given PAN fiber heat of reaction based on determining the proper values of heating ramp and temperature.