Improving Prediction Accuracy and Robustness in Injection Mechanism Based on Simplified Pareto and Updated Training Set Hybrid Metamodel

In squeeze casting, the injection parameters including fit clearance, speed, temperature, and their uncertainties significantly impact the forming quality. Robust optimization can improve the design reliability and reduce the influence of uncertainty while using a suitable metamodel is beneficial for prediction accuracy and efficiency. This paper proposes a robust optimization method based on a hybrid metamodel with Simplified Pareto and Updated Training Set (SPUTS) to improve the prediction accuracy along the Pareto front and in the whole design space. After the first round of robust optimization based on a general metamodel, the training set is updated by simplifying the Pareto solution set. A finite element simulation is performed to construct a high-precision metamodel that combines the kriging and radial basis function (RBF) models to run a new robust optimization. The proposed method was validated by application to the robust optimization of an injection mechanism with a large inner diameter. The results indicated that the SPUTS hybrid metamodel greatly reduced the prediction errors in the test set. The optimized design showed better reliability and robustness and had a greater clearance ratio than the initial design.