26 May 2023 in IEEE Transactions on Industrial Informatics
Compound bearing fault diagnosis is an essentially challenging task due to the mutual interference among multiple fault components. The state-of-the-art methods usually take the potential fault characteristic frequencies as the prior knowledge and then try to recover every fault component by exploiting the impulse signal sparsity. However, they inevitably suffer from algorithmic degradation caused by energy leakage, $l_{1}$ -norm approximation, and/or improper parameter selection. To handle these shortcomings, in this article, we propose a novel sparse Bayesian learning (SBL)-based method for the compound bearing fault diagnosis. We first present a new categorical probabilistic model to efficiently capture the truly-occurred fault components with a truncated feasible domain, which can greatly reduce the energy leakage effect. Then, we devise a more general SBL framework to recover the compound sparse impulse signal under the new categorical probabilistic model. The newly proposed method successfully avoids the $l_{1}$ -norm approximation and manual parameter selection; thus, it can yield much higher accuracy and robustness. Both simulations and experiments demonstrate the superiority of the developed method.
9 August 2023 in IEEE Transactions on Instrumentation and Measurement
This paper addresses the issue of broken rotor bar (BRB) fault detection for induction motors (IMs). The widely used fast Fourier transform (FFT) based methods are sensitive to noise and have a low resolution for short-time data; while subspace-type methods suffer from substantial performance degradation due to inappropriate model order choice and spatial smoothing operation. To effectively identify the weak fault sidebands from the rotor current signal, we cast the fault sideband identification problem as a sparse representation (SR) problem, and present a novel sparse Bayesian learning (SBL) approach for the BRB fault diagnosis. The novelties of the proposed approach are three-fold: (i) a three-stage hierarchical sparse prior is combined into the SBL framework to greatly enforce the sparsity of the current spectrum, leading to heavily sharpened fault sidebands; (ii) a new computationally efficient SR model with a truncated spectral grid is introduced to significantly reduce the computational load for Bayesian inference; and (iii) the symmetric pairing structure of fault sidebands is additionally exploited as a priori knowledge to enhance the robustness of the BRB identification performance. Both simulation and experiment results indicate the superiority of the proposed method, especially for the short sampling time and/or high noise level current signal.
13 October 2022 in IEEE Transactions on Instrumentation and Measurement
Extracting fault frequencies from noisy vibration signal is a challenging task for bearing fault diagnosis. The state-of-the-art sparse representation (SR) based methods usually consist of two steps: (i) fault impulse recovery in the time-domain and (ii) frequency transformation of the estimated signal envelope. However, any inaccurate time-domain signal recovery can cause an error accumulation problem for the following frequency transformation, and the frequency transformation itself encounters a low-resolution shortcoming especially for short-time sampling data. To handle these shortcomings, in this paper, we propose a novel sparse Bayesian learning (SBL) framework to evade the time-domain signal recovery and extract the fault frequencies directly from the frequency domain. We first present a new formulation for the sparse frequency recovery problem by utilizing the sparsity structure of the envelope spectrum, and then introduce a truncated off-grid model into the SBL framework to speed up the proposed method. Moreover, an improved grid refinement is developed to jointly combat the off-grid frequency mismatch and exploit the arithmetic sparsity structure of fault frequencies. Both simulation and experiment results indicate the effectiveness of our proposed method.
15 November 2023 in IEEE Sensors Journal
This paper investigates the fault characteristic frequency extraction from the noisy vibration signal for bearing fault diagnosis. Although sparse representation approaches are widely utilized and can achieve exceptional frequency extraction performance, they often encounter issues such as error accumulation or limited frequency resolution. State-of-the-art Bayesian learning methods can address these shortcomings, but they come with high computational complexity. Consequently, in this paper, we introduce a new low-complexity subspace-based approach for detecting fault characteristic frequencies, offering a more accurate and practical solution to bearing fault diagnosis. The novelties of the proposed method are two-fold: (i) present a subspace-based frequency extraction formulation by adopting multiple-level Hilbert transformation and spatial smoothing, which paves the way to extract the fault characteristic frequencies within frequency domain directly and efficiently; and (ii) utilize a special unitary matrix to construct a real-valued signal model and exploit the special rotation invariance under such a new real-valued model to facilitate noise suppression, computational complexity reduction and recovery performance enhancement. Results on both simulated and real datasets demonstrate the superiority of the proposed method.
3 June 2022 in IEEE Sensors Letters
The existing real-valued methods for direction-of-arrival (DOA) estimation usually utilize the orthogonality of signal subspace and noise subspace to compress the dimension of the observation matrix, which, in return, jointly reduces the computational complexity and achieves a noise suppression effect. However, it could suffer a performance loss due to the inaccurately estimated signal subspace. To address this problem, we propose a novel real-valued variational Bayesian inference (VBI)-based method, where we consider the signal subspace matrix as a random variable and adaptively update it by exploiting its row-sparsity. It is challenging to perform the Bayesian inference under the coupling effect caused by the unknown signal subspace matrix. Therefore, a column-independent VBI factorization is further introduced into the VBI framework to decouple the signal of interest. Simulation results demonstrate that our method can significantly improve the DOA estimation performance.