Automated and Robust Multi-Objective Optimal Design of Thin-Walled Product Injection Process Based on Hybrid RBF-MOGA
QingQing Feng,Xionghui Zhou
DOI: https://doi.org/10.1007/s00170-018-3084-5
2018-01-01
Abstract:An automated and robust multi-objective optimal design system has been developed and carried out on the optimization of injection process parameters. A hybrid multi-objective optimization of process parameters in plastic injection molding (PIM) is implemented for a synchronous decrease in multi objectives. The main defects indicating plastic part quality including warpage, volume shrinkage, and weldline are the objectives of optimization. Melt temperature, injection time, cooling time, mold temperature, packing pressure, and packing time are considered as optimization parameters. Orthogonal array-Latin hypercube sampling (OA-LHS) method is adopted to generate the initial set of parameter points. Metamodeling technique, radial basis functions (RBF), is utilized to construct the response surface fitting process parameters and simulation responses. RBF replaces expensive simulation experiments and reduces the time and computation cost. For a quick search of optimal point, Pareto-ranking-based multi-objective genetic algorithm (MOGA) is performed to make a trade-off among three objectives. Accuracy analysis shows a lower prediction error of the proposed algorithm, and sensitivity analysis identifies the parameters which have significant influence on response. Simulation experiments are implemented for verification of the optimization. The proposed optimization is automated with the help of visual basic scripts, python, Applications Programming Interface (API) of Moldflow and Dakota. The system is applied to an industrial application in thin-walled automobile air-conditioner vent injection process which increases efficiency and accuracy in optimal design compared to conventional trial-and-error design process.