Search

Machinery fault diagnosis is an attractive but challenging task, especially for low-speed conditions. Therefore, a new discriminative approach that introduces robust principal component analysis (RPCA) and multikernel to deep neural networks is proposed to perform intelligent fault diagnosis. First, RPCA is applied to extract fault signals from extreme background noise based on its sensitivity to grossly corrupted data. Second, two cascaded multikernel principal component analysis stages with additional robustness to distortions in feature extraction are used to enhance the energy of spectrum symptom and overcome the tricky issues of low-speed machinery. Especially, the multikernel is introduced into the basic PCA filters to learn the data-adapting convolution filter and gain additional robustness to nonlinearity in the signal. Finally, the proposed method is demonstrated on signals from laboratory tests (with a slightly damaged defect in a bearing) and structural fault data, outperforming those of traditional machine learning and classical deep learning methods. Moreover, hidden information of the network is visualized to analyze the reasons for its high performance. [ABSTRACT FROM AUTHOR]

Raw vibration signals poorly perform in industrial bearing fault diagnosis because impulse features are damped and masked by disturbances and noises. Fault diagnosis is more challenging due to weak features. This work presents a signal filtering and fault characteristic enhancement method based on reconstruction adaptive determinate stationary subspace filtering (Rad-SSF) and enhanced third-order spectrum to address the above problems. In particular, Rad-SSF reconstructs an adaptive self-determined, decomposed vibration signal trajectory matrix to obtain non-stationary signals. Then, the filtered signal with the best fault characteristics is extracted according to kurtosis. A 1.5-dimensional third-order energy spectrum is performed to enhance the fault characteristics by strengthening the fundamental frequency and eliminating non-coupling harmonics. Finally, the dominant frequency in the spectrum is contrasted to recognize fault diagnosis, referring to theoretical fault characteristic frequency. The feasibility and effectiveness of the proposed method are demonstrated by simulation and engineering signals under different conditions. [ABSTRACT FROM AUTHOR]

Determining the embedded dimension of a singular value decomposition Hankel matrix and selecting the singular values representing the intrinsic information of fault features are challenging tasks. Given these issues, this work presents a singular value decomposition-based automatic fault feature extraction method that uses the probability-frequency density information criterion (PFDIC) and dual beetle antennae search (DBAS). DBAS employs embedded dimension and singular values as dynamic variables and PFDIC as a two-stage objective to optimize the best parameters. The optimization results work for singular value decomposition for bearing fault feature extraction. The extracted fault signals combined with envelope demodulation can efficiently diagnose bearing faults. The superiority and applicability of the proposed method are validated by simulation signals, engineering signals, and comparison experiments. Results demonstrate that the proposed method can sufficiently extract fault features and accurately diagnose faults. [ABSTRACT FROM AUTHOR]

Fault feature extracted from fault signals plays a vital role in fault detection and diagnosis. However, in the early stage of defect, the fault signals (amplitudes and energy) are very weak and overwhelmed by large noise, and it is very difficult to extract fault feature effectively. To handle this problem, in this paper, an automatic filtering method based on an improved genetic algorithm (IGA) for extracting fault signals is presented. This method improves the binary coding generation of GA, so that each binary value matches a frequency of a fault signal frequency spectrum. Based on designed “statistical information evaluation function” as IGA fitness function, the similarity between normal state signal (noise signal) and IGA extracted noise signal is evaluated, and IGA employs genetic operations to search the optimal noise binary string. Negating noise binary string can obtain optimal fault signal binary string. It will then be multiplied by abnormal state vibration signal frequency spectrum, to yield a new frequency spectrum that only has fault components. This proposed method does not need to seek center frequency and bandwidth, and can automatically remove noise. The effective and advanced performance of the proposed method is evaluated by three different rolling bearing defects (outer race defect, inner race defect, and roller defect) in rolling bearing on rotating machine experimental bench and gearbox. Compared with rolling bearing fault characteristic frequency and results of denoising methods of bandpass filtering and wavelet decomposition, the proposed method can automatically filter noise and effectively extract fault signal. It proves that this method is feasible, effective, and advanced. [ABSTRACT FROM PUBLISHER]