基于BP网络的MEM200快速成形机工艺参数优化  

Process Parameter Optimization of MEM200 Rapid Prototyping Machine Based on BP Neural Network

作　　者：黄泳波[1] 许光辉[1] 黄道平[2] 四库[1] 

机构地区：[1]东莞理工学院机械工程学院,广东东莞523808 [2]华南理工大学自动化科学与工程学院,广东广州510641

出　　处：《顺德职业技术学院学报》2012年第1期4-8,共5页Journal of Shunde Polytechnic

摘　　要：对影响FDM制件翘曲变形的各种工艺参数进行深入分析,结合MEM200快速成形机的特点,确定了影响翘曲变形的五个主要的工艺参数,利用正交试验对主要影响因素进行了实验研究,为神经网络预测模型提供了学习样本。采用贝叶斯正则化算法,通过分析比较确定网络的最佳结构,以提高网络对翘曲变形和成形时间的预测能力。在此基础上,采用优化算法,利用Matlab编制程序实现了工艺参数的优化,经成形验证,表明了该方法能够减少制件的翘曲变形。To combine the characteristics of MEM200 rapid prototyping machine and determine five major process parameters of buckling deformation with further analysis of various process parameters infected by buckling deformation of FDM workpieces. The main factors were studied for the neural network modeling as the corresponding samples by orthogonal experiment. Use Bayesian regularization algorithm and comparison to determine the optimal network structure to increase predictive ability of the forming time and warping deformation. Based on the method, we use optimization algorithms to achieve the optimization of process parameters by Matlab. The results show that the method is feasible through the forming effect test.

关 键 词：熔融堆积成形 翘曲变形 正交试验 BP神经网络 工艺优化 

分 类 号：TH161[机械工程—机械制造及自动化]