基于深度学习的缓变故障早期诊断及寿命预测

发布时间：2018-04-27 21:17

本文选题：缓变故障 + 早期诊断　；参考：《山东大学学报(工学版)》2017年05期

【摘要】：为了克服传统的早期微小故障诊断方法不能区分多个不同时刻发生故障的不足,提出一种将深度学习和PCA相结合的方法实现微小缓变故障早期诊断及寿命预测。对采集的数据进行深度学习实现逐层特征抽取,学习早期微小故障特征,建立微小缓变故障早期诊断模型,结合PCA方法将深度学习所抽取的高维故障特征向量集成为一个故障特征变量,根据历史故障数据特征变量演化规律定义数据驱动的故障演变标尺,并通过指数型非线性拟合方法建立寿命预测模型。选取TE平台数据进行算法有效性检验,并与其他算法对比,从而验证了所提出算法的有效性。
[Abstract]:In order to overcome the shortcoming that the traditional early micro fault diagnosis method can not distinguish the fault at many different times, a method combining depth learning and PCA is proposed to realize the early diagnosis and life prediction of small and slow variable faults. The data are deeply studied to extract features from each layer, to learn the features of small faults in the early stage, and to establish a model for early diagnosis of small and slowly changing faults. Combined with PCA method, the high dimensional fault feature vector extracted by depth learning is integrated into a fault feature variable, and a data-driven fault evolution scale is defined according to the evolution rule of historical fault data feature variables. The life prediction model is established by exponential nonlinear fitting method. The data of te platform are selected to verify the validity of the algorithm, and compared with other algorithms, the validity of the proposed algorithm is verified.
【作者单位】：河南大学计算机与信息工程学院;杭州电子科技大学自动化学院;
【基金】：国家自然科学基金资助项目(U1604158) 河南省教育厅科学技术研究重点资助项目(16A413002)
【分类号】：TP18;TP277

【相似文献】