Blind denoising of structural vibration responses with outliers via principal component pursuit (original) (raw)
Related papers
Journal of Civil Structural Health Monitoring
Due to the increasing quest of adopting low-cost sensors in structural health monitoring (SHM) processes, which may lead to detecting signals contaminated by significant levels of noise, the need to devise appropriate and effective denoising strategies, at the post-processing stage, is becoming more and more essential. Among several approaches proposed in the literature, it has been demonstrated that the employment of discrete wavelet transform (DWT) as a multi-rate filter bank, as well as the use of singular value decomposition (SVD), may result to be quite effective in signal denoising within various research fields, as biological, acoustic and mechanical. Here, DWT- and SVD-based denoising techniques are first independently reconsidered and reimplemented, aiming at exploring their optimal calibration in purifying noise-corrupted vibration response signals encountered in civil engineering applications. Then, a systematic performance evaluation is provided within a comparative fram...
Shock and Vibration, 2022
The precise detection of building vibration signals is a crucial problem for the identification of building vibration sources and characteristics. However, the building vibration signal is usually accompanied by complex high-frequency noise. The present study proposed a novel building vibration signal denoising method based on improved empirical modal decomposition coupled with interwoven Fourier decomposition (IEMD-IWFD). The noise-embed building vibration signal is first decomposed by the IEMD-IWFD. Then, the intrinsic mode function (IMF) components with useful information are extracted from the original building vibration signal using the energy criterion of the autocorrelation function. After that, the building vibration signal is formed by reconstructing the IMF component using the Hilbert transform. Based on the comparison of similarity coefficient and mean square error between the reconstructed signal from IEMD-IWFDM and EMD and target signal, it is indicated that the IEMD-IW...
Anomaly detection for a vibrating structure: A subspace identification/tracking approach
The Journal of the Acoustical Society of America, 2017
Mechanical devices operating in noisy environments lead to low signal-to-noise ratios creating a challenging signal processing problem to monitor the vibrational signature of the device in real-time. To detect/classify a particular type of device from noisy vibration data, it is necessary to identify signatures that make it unique. Resonant (modal) frequencies emitted offer a signature characterizing its operation. The monitoring of structural modes to determine the condition of a device under investigation is essential, especially if it is a critical entity of an operational system. The development of a model-based scheme capable of the on-line tracking of structural modal frequencies by applying both system identification methods to extract a modal model and state estimation methods to track their evolution is discussed along with the development of an on-line monitor capable of detecting anomalies in real-time. An application of this approach to an unknown structural device is di...
Journal of Vibration and Control
Blind component separation (BCS) aims to decompose a single-channel vibration signal mixture into periodic components and random transient components. In addition to periodic components, random transient components with a high degree of impulsiveness are signals of interest in practice. An adaptive signal processing method called empirical mode decomposition (EMD) decomposes a nonlinear and non-stationary signal into the sum of simple components termed intrinsic mode functions (IMFs). Ensemble empirical mode decomposition (EEMD) was an improvement of the EMD and aims to relieve a mode mixing problem existing in the EMD. However, there is no a universal standard formula that can be used to select appropriate parameters of the EEMD. Improper parameters of the EEMD still cause a mode mixing problem that makes a signal of a similar scale reside in some successive IMFs. An enhanced EEMD for blind component separation is developed in this paper to respectively extract periodic components and random transient components from a single-channel vibration signal mixture. A revised spectral coherence (RSC) is proposed to measure the spectral dependence between two successive IMFs. The closer the revised spectral coherence is to one, the higher the spectral dependence of two successive IMFs is. Additionally, a fusion rule based on locations of local minima of the revised spectral coherence is proposed to automatically fuse successive IMFs with similar characteristics into a new IMF, called an enhanced IMF (EIMF). Vibration signals including simulated and real multi-fault signals were used to verify the enhanced EEMD. A comparison with the EEMD was conducted to show the superior of the enhanced EEMD. The results demonstrate that the enhanced EEMD has better performance than the EEMD for automatically extracting periodic components and random transient components from single-channel vibration signal mixtures.
Vibration-based monitoring and diagnostics using compressive sensing
Journal of Sound and Vibration, 2017
Vibration data from mechanical systems carry important information that is useful for characterization and diagnosis. Standard approaches rely on continually streaming data at a fixed sampling frequency. For applications involving continuous monitoring, such as Structural Health Monitoring (SHM), such approaches result in high volume data and rely on sensors being powered for prolonged durations. Furthermore, for spatial resolution, structures are instrumented with a large array of sensors. This paper shows that both volume of data and number of sensors can be reduced significantly by applying Compressive Sensing (CS) in vibration monitoring applications. The reduction is achieved by using random sampling and capitalizing on the sparsity of vibration signals in the frequency domain. Preliminary experimental results validating CS-based frequency recovery are also provided. By exploiting the sparsity of mode shapes, CS can also enable efficient spatial reconstruction using fewer spatially distributed sensors. CS can thereby reduce the cost and power requirement of sensing as well as streamline data storage and processing in monitoring applications. In well-instrumented structures, CS can enable continued monitoring in case of sensor or computational failures.
Mechanical Systems and Signal Processing, 2005
Under the only hypothesis of independent sources, blind source separation (BSS) consists of recovering these sources from several observed mixtures of them. As it extracts the contributions of the sources independently of the propagation medium, this approach is usually used when it is too difficult to modelise the transfer from the sources to the sensors. In that way, BSS is a promising tool for non-destructive machine condition monitoring by vibration analysis. Principal component analysis (PCA) is applied as a first step in the separation procedure to filter out the noise and whiten the observations. The crucial point in PCA and BSS methods remains that the observations are generally assumed to be noisefree or corrupted with spatially white noises. However, vibration signals issued from electro-mechanical systems as rotating machine vibration may be severely corrupted with spatially correlated noises and therefore the signal subspace will not be correctly estimated with PCA.
2017
Vibration-based monitoring of mechanical structures often involves continuous monitoring that result in high data volume and instrumentation with a large array of sensors. Previously, we have shown that Compressive Sensing (CS)-based vibration monitoring can significantly reduce both volume of data and number of sensors in temporal and spatial domains respectively. In this work, further analysis of CS-based detection and localization of structural changes is presented. Incorporating damping and noise handling in the CS algorithm improved its performance for frequency recovery. CS-based reconstruction of deflection shape of beams with fixed boundary conditions is addressed. Formulation of suitable bases with improved conditioning is explored. Restricting hyperbolic terms to lower frequencies in the basis functions improves reconstruction. An alternative is to generate an augmented basis that combines harmonic and hyperbolic terms. Incorporating known boundary conditions into the CS problem is studied.
Journal of Sound and Vibration, 2013
Using different transform basis, the vibration signal components, such as impulses, harmonics and modulated components, can be effectively separated by basis pursuit. For multicomponent vibration signals, this paper proposes a new vibration signal component separation approach by iteratively using basis pursuit. The signal is firstly denoised by basis pursuit denoising and the impulsive component can be separated by using identity matrix and redundant Fourier basis. Then signal components with different frequency ranges can be separated by using short-time Fourier transform basis with different window lengths, and the components with high frequency are separated after each iteration. To choose the best basis from the dictionary, the separation coefficient is proposed for evaluating the separating performance of the transform basis. The proposed approach, empirical mode decomposition (EMD) and matching pursuit are respectively applied to analyzing a simulated faulty vibration and a simulated faulty gear vibration signal. The comparison results demonstrate its superiority to EMD and matching pursuit in separation accuracy. Finally, the proposed approach is used for detecting the faults of a rotor and rolling bearings. The results further show that this proposed method can be effectively applied to mechanical fault diagnosis. & 2013 Elsevier Ltd. All rights reserved. cases, it is very difficult to directly obtain the fault information from the original vibration signal. Therefore, we usually first separate the useful vibration components from the measured vibration signal. Then, several methods, such as frequency spectrum, envelope spectrum and time-domain statistics, can be used for extracting the fault feature. Wavelet transform is an early approach for separating the signal into its components. It is successfully used for processing mechanical vibration signal [1-3]. However, typical wavelet transforms, such as orthogonal discrete wavelet transform [4], double density discrete wavelet transform [5], higher density dyadic wavelet transform [6], separate the Contents lists available at SciVerse ScienceDirect
2017
This paper assesses two different approaches for efficient output - only Vibration - based Structural Health Monitoring (V - SHM) in large - scale civil engineering structures, promoting the use of dense arrays of low - power wireless sensors. Firstly, a non - uniform deterministic sub - Nyquist multi - coset sampling scheme is considered to acquire ambient stationary structural response signals. This sampling scheme is coupled with a power spectrum blind sampling technique along with the frequency domain decomposition algorithm of operational modal analysis to obtain structural modal properties. This is accomplished without necessitating either signal reconstruction in the time - domain or signal sparsity assumption . Secondly, a spectro-temporal compressive sensing approach is considered applicable to cases where sign al reconstruction in time - domain is desired. The latter approach considers non-uniform in time random sampling at sub - Nyquist average rates informed by prior kno...