Compared with traditional plastics, glass fiber-reinforced plastic (GFRP) has
more outstanding performance advantages, which is more and more widely used. To
improve the quality of the products manufactured by the GFRP injection molding,
the injection parameters are optimized in two stages. In the first stage, the
range of optimization parameters including the glass fiber content and six
molding parameters is selected by the Moldflow recommendation. The warpage and
shrinkage of each orthogonal experiment are obtained by the Moldflow simulation.
Then, a comprehensive evaluation method called GRA-TOPSIS and the range analysis
method are utilized to identify the optimal level values of all optimization
parameters. According to the order of influence of each parameter, the range of
these parameters is adjusted for the second stage. In the second stage, the
orthogonal array table is also arranged for the training samples, and the Latin
hypercubic sampling (LHS) table is arranged for the prediction samples. The
regular extreme learning machine based on the improved particle swarm
optimization (IPSO-RELM) is utilized to construct the surrogate models of the
warpage and shrinkage, which replaces expensive experimental time and cost.
Then, the multi-objective firefly algorithm (MOFA) is performed to find the
Pareto-optimal front, and the GRA-TOPSIS method is performed to obtain the final
injection scheme. Through the simulation verification, the warpage and shrinkage
are reduced by 0.0857 mm and 0.1893% compared with the scheme of the first
stage, which indicates the effectiveness of the proposed multi-objective
optimization method.
