基于GK-SVR和自适应NSGA-II的涡轮盘锻件工艺参数多目标优化方法

Forging forming process directly determines the quality of turbine disk forgings, unsuitable process parameters will lead to unstable or even unqualified quality of turbine disk forgings. To solve this problem, a multi-objective optimization method for turbine disk forging process parameters based on Gaussian kernel support vector regression (GK-SVR) and adaptive NSGA-II optimization algorithm (ANSGA-II) is proposed, in which a multi-objective fitness function is established using GK-SVR for minimizing the deviation of the predicted turbine disk forging quality from the set target value, and ANSGA-II is used to iteratively find the optimal set of process parameters for the established multi-objective optimization model. The method improves the prediction accuracy of high-dimensional few-sample turbine disk forgings by introducing Gaussian kernel in SVR; meanwhile, it adopts Latin hypercube design for initializing the population and cosine probability-based parent individual selection to effectively enhance the population diversity, and designs an adaptive crossover operator that integrates simulated binary crossover operator and Lévy distribution crossover operator, which can better balance the global search and local search ability to better find the global optimal solution. The effectiveness and superiority of GK-SVR and ANSGA-II algorithms are verified through experiments, and the ranking of the superior and inferior performance of Pareto\'s non-inferior solution set is accomplished by approximating the ideal solution ranking method (TOPSIS) to achieve the acquisition of the optimal forging process parameter combinations.