Optimization of CFRP Pultrusion Process with NSGA-II and ANN

Abstract

The temperature and curing degree of carbon fiber reinforced plastic (CFRP) are coupled during pultrusion. In order to figure out the real-time temperature and curing degree of CFRP, the heat transfer model and curing model for resin were established on the basis of curing kinetics and heat transfer theory, and solved by the combination of finite element, finite different and indirect decoupling methods. The fiber Bragg grating (FBG) sensors were utilized to monitor the temperature of CFRP on real-time during pultrusion, while the curing degree of CFRP was measured through Sorbitic extraction. Experimental results show that the simulation method is effective and reliable. According to the simulated results, artificial neural network (ANN) combined with fast elitist non-dominated sorting genetic algorithm (NSGA-Ⅱ) to optimize die temperature and pull speed for pultrusion process, and significant improvements were achieved.