Adaptive Particle Swarm Optimization-Based Neural Network in Quality Prediction for Plastic Injection Molding

This paper presents an innovative neural network-based quality prediction system for a plastic injection molding process. The particle swarm optimization algorithm (PSO) is analyzed and an adaptive parameter-adjusting PSO algorithm based on velocity information (APSO-VI) is put forward. Compared with other PSO algorithms using several test functions, experimental results indicate that APSO-VI algorithm can easier find global optimization solution, performance of APSO-VI are significantly improved, especially while the optimization problems are complex nonlinear, high-dimensional, multi-peak and etc. The APSO-VI is used to train neural network, a back-propagation neural network based on APSO-VI model (APSO-VINN) is proposed for creating a dynamic quality predictor. The APSO-VINN is compared with back-propagation neural network model (BPNN) and back-propagation neural network based on PSO model (PSONN) which has the same contracture as the APSO-VINN. Experimental results show that APSO-VINN can better predict the product quality (volume shrinkage and weight) and can likely be used for various practical applications.