采用非线性变换单元组成的多层前馈神经网络建立了丝杆螺母摩擦副表面边界膜温度特性的磨损自补偿数学模型 ,该模型可用于准确地预测边界膜对摩擦学特性的影响 .采用 L- M规则进行神经网络学习训练使网络收敛快且误差小 ,所得网络输出结果与实验结果有较好的吻合性 .该神经网络可为工程设计人员进行摩擦学设计提供有效的计算工具 The multi-layer feedforward neural network consisting of nonlinear transformation elements is used to establish a self-compensated mathematical model for the temperature dependence of the boundary membrane temperature on the surface of the screw-nut friction pair. The model can be used to accurately predict the influence of the boundary membrane on the tribological properties. - M rules of neural network learning and training to make the network convergence and error small, the output of the network output results and experimental results have good agreement. The neural network for engineering design tribology design provides an effective calculation tool