【摘要】 滚动轴承的健康状态直接影响机械设备的性能和安全。传统滚动轴承智能故障诊断方法存在高质量状态数据缺失、模型泛化能力不足等问题。为解决上述问题，提出一种联合分布自适应并行卷积神经网络的滚动轴承变工况故障诊断方法。首先，采用快速傅里叶变换对原始振动信号进行预处理；其次，利用并行卷积神经网络提取不同层次、不同尺度的特征，从而充分捕捉输入数据中的信息；然后，引入最大均值差异和相关对齐结合的领域自适应方法，在均值和协方差角度减小源域与目标域间的分布差异；最后，利用凯斯西储大学轴承数据集不同工况之间的多种迁移学习任务对所提方法进行实验验证，证明了该方法在不同迁移任务中的有效性和优越性。更多还原

【Abstract】 The health state of rolling bearings directly affects the performance and safety of mechanical equipment. The traditional rolling bearing intelligent fault diagnosis methods have problems such as lack of high-quality state data and insufficient model generalization ability. In order to solve the above problems, a fault diagnosis method for rolling bearings under different working conditions based on parallel convolutional neural network with joint domain adaptation(PCNNJDA) is proposed. Firstly, the original vibration signal is preprocessed using the fast Fourier transform(FFT). Then, the parallel convolutional neural network(PCNN) is used to extract features of different levels and scales, so as to fully capture the information in the input data. A joint domain adaptation(JDA) method combining the maximum mean discrepancy and the correlation alignment is introduced to reduce the distribution difference between the source domain and the target domain in terms of mean and covariance. Finally, the proposed method is validated by a variety of transfer learning tasks between different working conditions of the Case Western Reserve University bearing dataset. The experimental results show that the proposed method is effective and superior in different transfer tasks.更多还原