针对机械伺服系统因温度变化产生的非线性摩擦变化，提出了一种反映温度因素的摩擦建模方法来实现对伺服系统的摩擦补偿。首先，分析了温度和摩擦的关系，并结合修正黏性摩擦的LuGre模型，讨论了该模型各参数与温度之间的关系。利用单隐层BP神经网络描述了随温度变化的各个参数，并确定了神经网络的输入、输出以及传递函数。然后，通过神经网络训练获得神经网络参数，从而得到与温度相关的摩擦模型。最后，改变运行条件，验证了提出的模型对摩擦的估计能力。建立的摩擦模型在不同运行条件、不同温度状态下的最大相对估计偏差小于2.5%，表明其能很好地估计系统摩擦力矩，满足高精度摩擦补偿。

In consideration of the nonlinear friction from changed temperatures of a mechanical servo system, a nonlinear friction modeling related to the temperature change was proposed to achieve high precision friction compensation of the servo system. Firstly, the relationship between temperature and friction was analyzed, and the dependence of the parameters on the temperature in the model was further analyzed based on the modified LuGre model of viscous friction. Then, the single hidden layer BP neural network was used to describe the parameters changed with temperature and to determine the input, output parameters and transfer functions of the neural network . Furthermore, an experiment was designed and parameters of the neural network were obtained by training the neural network, by which the friction model related to temperature change was implemented . Finally, the ability of the model for friction estimation was verified by changing operation conditions. The experimental results indicate that the maximum relative estimation error of the frictional model is less than 2.5% when it is applied under different operation and temperature conditions. The friction model related to the temperature change estimates system friction torques in various operating conditions accurately and satisfies the need of high precision friction compensation.