针对湿式离合器油液监测数据具有来源分散、 数据量大及时间轴不稳定等问题, 对光谱分析得到的多数据进行融合, 利用维纳过程的预测实时性和预测准确的优点, 建立模型开展离合器剩余寿命预测研究。 首先, 通过排列熵加权证据融合方法对离合器寿命试验获得的指示元素进行融合, 构建健康指数； 其次, 结合维纳过程建立退化模型并通过极大似然法对模型中的参数进行估计； 再次, 根据历史退化数据对参数进行更新得到离合器剩余寿命预测模型； 最后, 将预测模型与实例进行对比, 得到利用融合多元素的健康指数建立的维纳过程预测模型预测准确性相比单指示元素预测有了很大提升, 其预测点更接近试验值。 通过观察也发现, 湿式离合器运行50～60 h左右时, 预测点有了明显的变化, 而同样在220～230 h时, 预测点有了明显的偏差, 在240 h左右又重新接近试验值, 其突变点对应了离合器磨损的三个阶段, 即初期磨损, 正常磨损和剧烈磨损。 研究结果表明, 融合油液光谱数据结合维纳过程建立的预测模型用于湿式离合器的剩余寿命预测, 具有预测实时性强且预测精度高的优点, 而通过预测结果和试验值对比发现, 湿式离合器磨损状态的不同对预测结果也有一定的影响, 尤其磨损状态转变点对预测结果的影响更大。

Aiming at the problems of oil monitoring data of wet clutch, such as scattered sources, a large amount of data and unstable time axis, the multi-data obtained from the spectral analysis are fused. Using the advantages of real-time prediction and accurate prediction of the Wiener process, a model is established to predict the remaining life of the clutch. Firstly, the indicator elements obtained from the clutch life test are fused by permutation entropy weighted evidence fusion method to construct the health index; Secondly, combined with the Wiener process, the degradation model is established, and the parameters in the model are estimated by the maximum likelihood method; Thirdly, the residual life prediction model of the clutch is obtained by updating the parameters according to the historical degradation data; Finally, by comparing the prediction model with an example, it is found that the prediction accuracy of the Wiener process prediction model based on the health index of multi-element fusion is greatly improved compared with the single indicator element prediction. Its prediction point is closer to the experimental value. Through observation, it is also found that when the wet clutch operates for about 50 to 60 h, the prediction point has an obvious change, while the prediction point has an obvious deviation from 220 to 230 h, and it is close to the test value again at about 240 h. The mutation point corresponds to the three stages of clutch wear: initial wear, normal wear and severe wear. The research results show that the prediction model based on oil spectrum data and the Wiener process has the advantages of strong real-time prediction and high prediction accuracy for the remaining life of the wet clutch. The comparison between the prediction results and the test values shows that the different wear states of wet clutch also have a certain impact on the prediction results, especially the wear state transition point has a greater impact on the prediction results.