基于PSO-BP神经网络湿式摩擦元件损伤预测模型

为求解湿式离合器的多影响因素损伤关系，应用多源数据融合方法，构建一种基于PSO-BP神经网络的湿式摩擦元件损伤预测模型. 将转速和接合油压作为模型的输入参数，将提取到的摩擦片周向温度梯度、Fe和Cu元素浓度变化率、摩擦片表面粗糙度变化率作为模型输出参数, 建立了有限元仿真模型，搭建了湿式离合器摩擦磨损综合试验台，采用控制变量法研究了油压、转速对摩擦元件损伤特征参数的影响. 结果表明，输入工况与4类损伤特征参数呈非线性关系，预测值与实测值随工况变化趋势一致，损伤特征参数较油压的变化更为敏感. 对比同类模型与试验数据，预测模型具有较高的预测精度，能够有效地对湿式离合器多工况损伤进行预测. 

A Damage Prediction Model of Wet Friction Elements Based on PSO-BP Neural Network

In order to solve the multi factor damage relationship of wet clutch, a wet friction element damage prediction model based on PSO-BP neural network was constructed by using multi-source data fusion method. Taking rotational speed and joint oil pressure as input parameters of the model, taking the extracted circumferential temperature gradient of friction plate, the change rate of Fe and Cu concentration and the change rate of friction plate surface roughness Ra as output parameters of the model, a finite element simulation model was established, and the comprehensive friction and wear test-bed of wet clutch was built. The effects of oil pressure and speed on the damage characteristic parameters of friction elements were studied by using the control variable method. The results show that the input condition takes on a nonlinear relationship with the four types of damage characteristic parameters, the variation trend of the predicted value and the measured value is consistent with the working condition, and the damage characteristic parameters are more sensitive than the change of oil pressure. Compared with similar models and test data, the prediction model can provide higher prediction accuracy and can effectively predict the multi condition damage of wet clutch.