【摘要】 为了提升鲁棒性和泛化性，并且考虑各种深度自动编码器的互补性能，提出了一种基于动态加权的集成深度自动编码器的旋转机械故障诊断。结合稀疏深度自动编码器，降噪深度自动编码器和收缩深度自动编码器三种模型来构造集成深度自动编码器，提升处理冗余信息、噪声破坏和信号扰动的能力。为了增强识别性能，提出了一种动态加权平均方法来聚合学习特征。在自吸离心泵数据集和电机轴承数据集上进行了实验验证，结果显示提出方法的测试精度分别达到100%、99.69%和99.92%。通过与其他方法的比较，证明了提出的故障诊断方法的有效性。更多还原

【Abstract】 In order to improve the robustness and generalization,and consider the complementary performance of various depth automatic encoders,a fault diagnosis method of rotating machinery based on dynamic weighting integrated depth automatic encoder was proposed.Combined with sparse depth automatic encoder,noise reduction depth automatic encoder and shrinking depth automatic encoder,the integrated depth automatic encoder was constructed to improve the ability of processing redundant information,noise damage and signal disturbance.In order to enhance the recognition performance,a dynamic weighted average method was proposed to aggregate learning features.The experimental verification is carried out on the data sets of centrifugal pump and motor bearing,and the results show that the test accuracy of the proposed method is 100%,99.69% and 99.92% respectively.Compared with other methods,the effectiveness of the proposed fault diagnosis method is proved.更多还原