Application of backward propagation network for forecasting hardness and porosity of coatings by plasma spraying

Plasma spray forming has been showing overwhelming advantages in rapid fabricating parts and molds, in which the forming quality is determined by so many processing parameters. A robust methodology that takes into account the relationship between the spray parameters and processing variables is urgently needed to control the forming quality. In this study, two backward propagation (BP) neural network models have been designed. For the first model, the temperature and velocity distribution regularities of in-flight particles have been investigated. The second model has been created to build up the nonlinear relationship between the spray parameters and the coating porosity and hardness, which has a dramatic effect on the forming quality. In order to train the BP structures and to check the validity of the methodology, plasma spraying of WC—12% Co powder was implemented according to the method of orthogonal experiments, and particle in-flight properties were monitored by an optical monitoring system of CCD camera. The predicted results were found in good agreement compared with the experimental data. The research provides an effective way to the adaptive control of the plasma spray forming and shows the effect of great significance to optimize the performance of coating.