基于高斯过程机器学习的注塑过程建模及工艺参数优化设计

摘要

注塑成型过程中的工艺参数的取值对成型质量有很大影响,工艺参数之间是非线性关系,采用常规的理论分析和数值计算方法难以快速准确找到其最优解.高斯过程机器学习是一个新的预测方法,采用贝叶斯统计方法和非线性回归技术解决复杂的非线性建模问题.为获得好的成型质量,采用高斯过程机器学习的方法建立注塑成型工艺过程代理模型,可获得满意的模型后用遗传算法求得优化的工艺参数.选用聚甲醛小模数齿轮的翘曲变形实例来验证了方法的可行性.%In the process of injection molding process, parameters selection has a great influence on molding results. It's hard to find the optimal solution because of the nonlinear relation by means of conventional method such as theoretical analysis and numerical calculation. As a new prediction method, Gaussian Process (GP) machine learning employs a Bayesian statistics approach and adopts a highly nonlinear regression technique to solve this complicated nonlinear problem. In order to get good molding result, GP is proposed to construct the regression model for injection molding process. Genetic algorithm is used to optimize the process parameters with satisfactory GR model. An example withpolyformcddehyde small module gear is selected to verify this method