基于支持向量机算法的微注射成型工艺参数优化  

Process Parameter Optimization of Micro-Injection Molding Based on Support Vector Machine

作　　者：杨金领[1] 尹自强[1] 关朝亮[1] 铁贵鹏[1] 

机构地区：[1]国防科技大学机电工程与自动化学院,湖南长沙410073

出　　处：《塑料工业》2009年第5期27-30,共4页China Plastics Industry

基　　金：国家自然科学基金资助项目(50735007)

摘　　要：为了控制菲涅尔透镜在注塑过程中的翘曲变形量,采用支持向量机算法建立了菲涅尔透镜的翘曲预测模型,并对该模型的预测精度进行了研究。采用正交试验法获取注塑工艺参数,各组注塑工艺经Moldflow仿真得出模型的训练样本及检验样本数据。然后,对支持向量机算法建立的翘曲预测模型进行样本学习,训练完毕后由检验样本验证该模型的预测精度。实验结果表明:采用支持向量机算法建立的预测模型预测误差比较稳定,均在0.2%以内。因此,采用支持向量机算法建立菲涅尔透镜的翘曲预测模型可有效地预测菲涅尔透镜的最大翘曲量,且预测的精度与稳定性较高。In order to control the warpage of Fresnel lens in micro-injection molding process, the model of non-linear relationship between the injection process parameters and the maximal warpage of Fresnel lens was established through the support vector machine （SVM）. Process parameters were designed as orthogonal arrays, and then, the training and testing data could be obtained by numerical simulation. After the simulation, the maximal warpage of Fresnel lens could be predicted using a SVM program based on leaning data extracted from simulation results. Then, the prediction preeision of the model could be proved by testing data. Experimental results indicate that the prediction error of the model was very steadily, and all the error was below 0. 2%. Henee, such a SVM-based model eould accurately predict the maximum product warpage quality with accuration and stability.

关 键 词：微注射成型 支持向量机 翘曲 预测 菲涅尔透镜 

分 类 号：TQ320.662[化学工程—合成树脂塑料工业] TP183[自动化与计算机技术—控制理论与控制工程]