基于径向基函数神经网络的橡胶配方性能预测  

Performance prediction of rubber formulation based on radial basis function neutral network

作　　者：于增顺[1] 高齐圣[1] 杨方[1] 

机构地区：[1]青岛大学自动化工程学院,山东青岛266071

出　　处：《橡胶工业》2010年第6期335-338,共4页China Rubber Industry

基　　金：国家自然科学基金资助项目(70571041)

摘　　要：介绍径向基函数(RBF)神经网络的结构和常用算法。以SBR胶料配方为例,利用RBF神经网络拟合配合剂用量与胶料性能间的非线性关系,以15组样本数据为训练样本进行网络训练,对15组训练样本和1组非训练样本的预测值与实测值进行比较。结果表明,误差在容许范围之内,说明RBF神经网络算法适用于橡胶配方性能预测。The widely used algorithm and structure of radial basis function（RBF） neutral networks were introduced.Taking SBR compound formulation as example,the nonlinear relationship of level of compounding agents and the properties of compound was fitted by using RBF neutral network.Network training was processed with 15 groups of sample data,and the predicated values and measured values of 15 groups of training samples and one new sample were compared.The results showed that the error was within allowable range and the neutral network algorithm was suitable for the performance predication of rubber formulation.

关 键 词：径向基函数 神经网络 胶料配方 

分 类 号：TQ330.61[化学工程—橡胶工业]