基于模糊RBF神经网络的选域激光烧结尺寸精度预测建模.pdfVIP

学兔兔 Oct．2008 VoI．57 NO．1O ·1029· 基于模糊RBF神经网络的选域激光烧结 尺寸精度预测建模 李小飞，董俊慧，张永志 (内蒙古x-大学材料科学与工程学院，内蒙古呼和浩特010051) 摘要：将减法聚类与伪逆法相结合建立了模糊RBF(RadialBasisFunctionneuralnetwork，径向基神经网络)模型。通 过正交试验，获得不同激光功率、扫描速度、预热温度和切片厚度参数条件下SLS试件的尺寸误差，在此基础上获得 训练与预测样本数据，并对该模型进行仿真。结果显示预测平均总误差为2．16％，表明该模型具有建模快、模型简单、 训练速度快、预测精度高，泛化能力强的优点，可根据不同烧结工艺参数对SLS制件尺寸精度进行较准确地预测，以 便指导实际生产。 关键词：模糊神经网络；预测；尺寸精度 中图分类号：TF124 文献标识码：A 文章编号：1001—4977(2008)10—1029—04 Modeling for Predicting SLS Dimension Precision Based 0n Fu77y RBF Neural Network LI Xiao—fei，DONG Jun-hui，ZHANG Yong—zhi (College of Materials Science and Engineering，Inner Mongolia University of Technology， Hohhot 01 0051，Inner Mongolia，China) Abstract：With combining subtractive clustering and pseudo．inverse method．fuzzy RBF neural network iS established．By orthogonal test，dimension error of selective Iaser sintering《SLS) specimens under different parameters including laser power，scan speed，preheating temperature and slice thickness are obtained．Training and predicting swatch data are obtained on this base．The results of network simulation show that the average error of prediction iS 2．16％．It indicates that this modeI has excellences of building model quickly，simple structure，training fast，high precision and good generalization．It can be used in predicting dimension precision of SLS part exactly according to different sinter parameter in order to direct practice production． Key words：fuzzy neural network；predict；dimension precision 选域激光烧结 (Selective Laser Sintering，简称SLS) 产中加工参数发生变化时，制件的尺寸精度也随之改 是借助精确引导的激光束使材料粉末烧结或熔融后凝