Optimization model for isothermal CVI process parameters for C/C composites based on genetic algorithm and neural network

An optimization model of the process parameters during a chemical vapor infiltration (CVI) was established based on genetic algorithm and back propagation (GA-BP) neural network. The experimental data from the novel isothermal CVI process of carbon/carbon (C/C) composites were selected as the samples to evaluate the model. The effect of the main controllable factors, such as infiltration temperature, part pressure of precursor gas and resident time etc, on the density and uniformity of C/C composites were analyzed. Under the guidance of the model, the maximum errors between the desired densities and the tested densities of the experiment samples are not larger than 6.2% and those between their density differences were not larger than 8.2%. The results show that the established optimization model has high precision and good generalization. It can be efficiently applied for optimizing CVI process parameters.