基于神经网络的聚四氟乙烯复合材料摩擦系数实时识别  

Real-time identification of friction coefficient of PTFE composite based on neural network

作　　者：任华新[1] Ren Huaxin(Liaoning University of International Business,Dalian 116033)

机构地区：[1]辽宁对外经贸学院

出　　处：《化工新型材料》2019年第12期130-134,共5页New Chemical Materials

基　　金：2018年度辽宁省普通高等学校本科教学改革研究项目（10841625）

摘　　要：基于神经网络,对聚四氟乙烯(PTFE)复合材料摩擦系数实时识别进行了研究,在此阶段中,可以对材料的物理性质进行实时监控,通过人工神经网络模型结构来对PTFE复合材料的摩擦系数进行实时识别,从而为后续复合材料的加工工艺提供实时监控的理论依据和支撑。采用BP算法训练的神经网络均可进行任意精度逼近,但在实际网络训练中,非线性作用函数类型、训练样本数量等均可能降低网络模型精度。采用非线性函数最小二乘曲线拟合的线性算法表明,该线性化处理方法是正确的,为已知材料性能时,快速识别摩擦系数提供了一种有效的手段。Based on neural network,the real-time identification of friction coefficient of PTFE composites was studied.Through on-line monitoring of physical parameters,the real-time identification of friction coefficient of PTFE can be realized by using artificial neural network model,which provided basic support for real-time control and real-time prediction of intelligent deep drawing.Neural networks trained by BP algorithm can be approximated with arbitrary accuracy,but in actual network training,the type of non-linear function and the number of training samples may reduce the accuracy of network model.The linear algorithm of least squares curve fitting with non-linear function shown that the linearization method was correct and provided an effective means for fast identification of friction coefficient when material properties were known.

关 键 词：摩擦系数 聚四氟乙烯复合材料 BP算法 实时识别 

分 类 号：TB332[一般工业技术—材料科学与工程] TP183[自动化与计算机技术—控制理论与控制工程]