硅灰石纤维填充超高分子量聚乙烯基复合材料干滑动摩擦磨损的BP神经网络分析

超高分子量聚乙烯及其复合材料由于其优良的自润滑和防黏性能可用于农业工程装备中的滑动接触部件和触土部件。基于人工神经网络在复杂系统建模问题上的优越性，考察了几种因素对硅灰石纤维增强复合材料的摩擦和磨损性能影响的模型。考虑到输入、输出数据个数，调试设计了一个3×10×2的BP神经网络，其输入层由3个神经元构成，分别为硅灰石纤维的处理方法、硅灰石纤维的加入量和试验过程中的法向载荷。隐含层有10个神经元。输出层2个神经元分别为材料的摩擦系数和磨损量。基于上述BP神经网络对硅灰石纤维增强超高分子量聚乙烯基复合材料的干滑动摩擦磨损性能进行了模拟和预测。对神经网络的训练和检验表明该BP神经网络能够较好地预测影响因素对复合材料的干滑动摩擦和磨损的作用，大部分数据的预测值与试验值的误差在10%以内，其仿真精度能够满足实际的摩擦磨损预测要求。

BP neural network analysis of dry sliding friction and wear of UHMWPE composites

Ultra high molecular weight polyethylene(UHMWPE) and its composites can be used for the sliding contact parts and soil-working components of agricultural engineering equipment since their excellent self-lubrication and anti-adhesion. Dry sliding friction and wear of UHMWPE composites filled with wollastonite fibers were simulated and estimated based on BP neural network. The BP neural network with a 3×10×2 structure was designed, trained and checked. It is shown that the designed neural network can estimate the effects of the coupling treatment method of wollastonite fibers, fiber content in the composites and the normal load during tests on the dry sliding friction and wear properties of the composites effectively. Mostly, the estimated values of the friction coefficients and worn volumes were close to the corresponding experimental data and the errors between the estimated values and the corresponding experimental data were less than 10%.