水辅助注塑的GA-LMBP逆向神经网络建模与预测  

Modeling and Prediction of Water-Assisted Injection Molding Based on GA-LMBP Inverse Neural Network

作　　者：黄汉雄[1] 何建民[1] 刘旭辉[1] 邓志武[1] 

机构地区：[1]华南理工大学工业装备与控制工程学院,广东广州510640

出　　处：《华南理工大学学报（自然科学版）》2007年第12期23-27,33,共6页Journal of South China University of Technology(Natural Science Edition)

基　　金：广东省自然科学基金资助项目(06025643);教育部留学回国人员科研启动基金资助项目

摘　　要：水辅助注塑是一种新的塑料注塑技术,由于其过程的复杂性,难以采用数学方法建立其过程的数学模型.因此,文中提出一种遗传算法(GA)与LMBP神经网络算法相结合的逆向神经网络(简称GA-LMBP),采用一系列的实验结果,建立水辅助注塑的过程模型.交叉验证表明,该模型的预测值与实验值较吻合.输入水辅助注塑制品上不同位置的壁厚,该模型可快速而准确地预测相应的加工参数,包括熔体注射量、注水压力、注水延迟时间和熔体温度.Water-assisted injection molding （WAIM） is a new injection molding technique whose mathematical model is difficult to establish by mathematical method due to the process complexity. In this work, an inverse neural network named GA-LMBP is proposed by combining the genetic algorithm （GA） with the Levenberg-Marquardt back-propagation （LMBP） neural network. Based on the proposed GA-LMBP, a model to predict the WAIM process is developed according to a series of experimental results. It is found from the cross-validation that there is a good agreement between the predicted results by the model and the experimental ones, and that, with the thickness at different locations of molded parts as the system input, the model can quickly and accurately predict such processing parameters as short-shot size, water pressure, water injection delay time, and melt temperature.

关 键 词：水辅助注塑 LMBP神经网络 遗传算法 逆向过程 建模 

分 类 号：TQ320.062[化学工程—合成树脂塑料工业]