Spectral Kurtosis Research Articles

Rolling bearing fault feature detection using nonconvex wavelet total variation

Vibration signals measured from rolling bearing are often used to judge operational condition of rotation machinery. This paper proposes a nonconvex wavelet total variation method to detect rolling bearing fault feature submerged in noise measurement. Firstly, the parametric minmax concave function is used to construct a novel wavelet total variation model to improve the accuracy of signal estimation and induce more strongly sparsity. Second, convexity parameters and regularization parameters are limited in a given region to make sure convexity of the constructed cost function. With this, an iterative algorithm with guaranteed convergence is derived to efficiently obtain the global minimum of the constructed cost function. Simulation analysis and actual application validation show that the proposed method has a good impact estimation performance and impact recoverd by the proposed method preserves more accurate amplitude than that of traditional l1-norm regularized wavelet total variation and Spectral Kurtosis.

A parameter-adaptive ACMD method based on particle swarm optimization algorithm for rolling bearing fault diagnosis under variable speed

Fault diagnosis for rolling bearing under variable speed is always a challenging topic since the vibration signal has time-varying characteristics. To overcome this difficulty, a novel method is exploited based on particle swarm optimization (PSO) and adaptive chirp mode decomposition (ACMD), named parameter-adaptive ACMD. Firstly, fast spectral kurtosis algorithm is used to get the resonance band signal. Then, the parameter-adaptive ACMD method decomposes the envelope signal to obtain the time-frequency spectrum. Next, the proposed method of fault diagnosis uses peak search algorithm to estimate instantaneous rotational frequency from the time-frequency graph processed. Finally, the measured rotational frequency is used as the phase function in the resampling process to get the order spectrum and fault characteristic order (FCO). Simulation and actual signals were analyzed and the results indicate that the proposed method can identify fault type with variable speed and has high application value.

Fault Diagnosis of Wind Turbine Bearing Based on Optimized Adaptive Chirp Mode Decomposition

Periodic impacts are considered as the important defect signatures of wind turbine bearing. However, it is difficult to separate the weak periodic impacts from collected signal under serious interference environment. Facing with this issue, the recently developed signal processing technique named adaptive chirp mode decomposition (ACMD) is firstly introduced into the fault diagnosis field of wind turbine bearing. And the influences of instantaneous frequency and weighting coefficient parameters of ACMD are researched through simulated fault signal. In additional, the L-kurtosis fluctuation spectrum (LFS) guided instantaneous frequency selection strategy and the cuckoo search algorithm (CSA) guided weighting coefficient determination strategy are respectively put forward to automatically confirm the optimal influencing parameters without requiring priori knowledge. Combined with these two specific strategies, an innovation optimized adaptive chirp mode decomposition (OACMD) method is further proposed for defect identification of wind turbine bearing. The experimental signals and the engineering signal are in turn adopted to demonstrate the reliability of this proposed diagnostic method. The obtained analysis results show that the informative target mode with obvious periodic impact signatures can be effectively extracted, and the key defect characteristic frequencies are able to be successfully identified. Furthermore, the conventional spectral kurtosis method, the minimum entropy deconvolution method and the singular spectrum decomposition method are respectively employed for performance comparison. And it is validated that the proposed OACMD method has outstanding advantage on weak defect detection of wind turbine bearing.

Identification of short-wavelength contact wire irregularities in electrified railway pantograph–catenary system

The effect of two common types of short-wavelength irregularities (local imperfections and periodic short-wavelength irregularities) of the railway catenary on the pantograph–catenary interaction performance are studied and their potential identification approaches are explored in this paper. The analysis of the intrinsic mode functions of panhead acceleration indicates that the effect of local imperfection can be reflected in the high-order deformation mode of the contact wire. The cut-off frequency is suggested to cover the wavelength smaller than 1/8 of the dropper to dropper distance, which can be used to identify the local imperfection. The instantaneous energy obtained by the Hilbert transform is used to localise the local imperfection. The effect of periodic short-wavelength irregularities can be recognised as the introduction of non-Gaussian behaviour in the contact force at specific wavelengths. Thus, spectral kurtosis is utilised to identify the deviating wavelength. The short-wavelength irregularity can be localised by the time-frequency analysis of the intrinsic mode function containing the identified deviating wavelength. The examples with measurement data indicate the validation of the present methods with some improvements to the current equipment.

Audio classification using grasshopper‐ride optimization algorithm‐based support vector machine

The accurate and robust detection of the audio has been widely grown as the speech technology in the area of audio forensics, speech recognition, and so on. However, in real time, it is a challenge to deal with the massive data arriving from distributed sources. Thus, the study introduces a method that effectively deals with the data from the distributed sources using the map-reduce framework (MRF). The map and reduce function in MRF aim at feature extraction and audio classification. The robust classification using the proposed grasshopper-ride optimization algorithm-based support vector machine (G-ROA-based SVM) uses the features, such as multiple kernel Mel frequency cepstral coefficients, spectral flux, spectral kurtosis, and delta-amplitude modulation spectrogram. The proposed G-ROA is the integration of ROA and grasshopper optimization algorithm in tuning the optimal weights of SVM and also, the kernel function in SVM is modified using the Gaussian radial basis function, Gaussian kernel, and polynomial kernels. The experimentation of the proposed method is done using two datasets, namely TUT sound event 2017 dataset and ESC dataset. TUT sound event 2017 dataset consists of eight audio recordings from a single acoustic scene. ESC dataset consists of three parts and 252,400 recordings. The analysis reveals that the proposed audio classification acquired the maximal accuracy of 0.96, minimal false alarm rate, and false rejection rate of 0.022 and 0.0119, respectively.

Transient impulses enhancement based on adaptive multi-scale improved differential filter and its application in rotating machines fault diagnosis

Transient impulses caused by local defects are critical for the fault detection of rotating machines. However, they are extremely weak and overwhelmed in the strong noise and harmonic components, making the transient features are very difficult to be extracted. This paper proposes an adaptive multi-scale improved differential filter (AMIDIF) to enhance the identification of transient impulses for rotating machine fault diagnosis. In this scheme, firstly, the AMIDIF is performed to decompose the measured signal of rotating machine into a series of multi-scale improved differential filter (MIDIF) filtered signals. Subsequently, in view of the MIDIF filtered signals exhibit varying extents of validity in revealing fault features, a weighted reconstruction method using correlation analysis is proposed in which the weighted coefficients are counted and distributed to the corresponding MIDIF filtered signals to highlight the effective MIDIF filtered signals and weaken the invalid ones. Finally, the transient impulse components of rotating machinery are obtained by multiplying the weighted coefficients and the MIDIF filtered signals under different scales. Furthermore, the fault types of rotating machines are inferred from the fault defect frequencies in the envelope spectrum of the transient impulses. Simulation analysis and experimental studies are implemented to verify the performance of the AMIDIF compared with the state-of-the-art methods including spectral kurtosis (SK), multi-scale average combination different morphological filter (ACDIF) and multi-scale morphology gradient product operation (MGPO). The results prove that the AMIDIF has excellent performance in extracting transient features for rotating machines fault diagnosis.

Are Acoustic Markers of Voice and Speech Signals Affected by Nose-and-Mouth-Covering Respiratory Protective Masks?

BackgroundWorldwide use of nose-and-mouth-covering respiratory protective mask (RPM) has become ubiquitous during COVID19 pandemic. Consequences of wearing RPMs, especially regarding perception and production of spoken communication, are gradually emerging. The present study explored how three prevalent RPMs affect various speech and voice sound properties.MethodsPre-recorded sustained [a] vowels and read sentences from 47 subjects were played by a speech production model (‘Voice Emitted by Spare Parts’, or ‘VESPA’) in four conditions: without RPM (C1), with disposable surgical mask (C2), with FFP2 mask (C3), and with transparent plastic mask (C4). Differences between C1 and masked conditions were assessed with Dunnett's t test in 26 speech sound properties related to voice production (fundamental frequency, sound intensity level), voice quality (jitter percent, shimmer percent, harmonics-to-noise ratio, smoothed cepstral peak prominence, Acoustic Voice Quality Index), articulation and resonance (first and second formant frequencies, first and second formant bandwidths, spectral center of gravity, spectral standard deviation, spectral skewness, spectral kurtosis, spectral slope, and spectral energy in ten 1-kHz bands from 0 to 10 kHz).ResultsC2, C3, and C4 significantly affected 10, 15, and 19 of the acoustic speech markers, respectively. Furthermore, absolute differences between unmasked and masked conditions were largest for C4 and smallest for C2.ConclusionsAll RPMs influenced more or less speech sound properties. However, this influence was least for surgical RPMs and most for plastic RPMs. Surgical RPMs are therefore preferred when spoken communication is priority next to respiratory protection.

A neural network approach for improved bearing prognostics of wind turbine generators

Condition monitoring of High-Speed Shaft Bearing (HSSB) in Wind Turbine Generators (WTGs) remains a challenging subject for industrial and academic studies. The investigation of mechanical vibration signals presents the most popular method in the literature. Consequently, this work involves a novel data-driven approach for direct HSSB prognosis using the vibration analysis. The proposed method is based on the computation of traditional statistical metrics derived both from the time-domain and frequency-domain via Spectral Kurtosis (SK). Then, the selection of the most suitable features was made using three metrics (monotonicity, trendability, prognosablity) to guarantee a better generalization of the trained Elman Neural Network (ENN). The validation of the proposed method was done using the benchmark of the center for Intelligent Maintenance Systems (IMS) for training and real measured Green Power Monitoring Systems (GPMS) data for testing. We have provided two links for downloading these data sets. The experimental results show that the proposed approach presents a powerful prediction tool. Comparative results with previous work show several advantages for the proposed combination of statistical metrics and ENN, such as the external prediction and real online estimation of the Remaining Useful Life (RUL). Also, some new practical findings are provided in the discussion.

The Enkurgram: A characteristic frequency extraction method for fluid machinery based on multi-band demodulation strategy

Modulation frequency extraction of fluid machinery is significant, not only for the condition monitoring and fault diagnosis of cooperative targets in industrial fields but also for object detection and information prejudgment of non-cooperative targets in military applications. However, most existing demodulation methods tend to show poor performance due to the ubiquitous heavy Gaussian and non-Gaussian noise. Especially, narrowband demodulation methods such as Kurtogram are in a dilemma due to the signal characteristic of multi-wideband carrier wave in fluid machinery. In this paper, the Enkurgram is proposed for multiple demodulation frequency bands selection based on the combination of energy factor and shape factor. This method shows excellent demodulation capability in both simulation analysis and applications to fluid machinery. Firstly, an Amplitude-Modulated (AM) signal model with multi-wideband carrier wave is established. Unlike other narrowband demodulation methods, multiple non-overlapping frequency bands are selected by Enkurgram for demodulation based on the composite criterion of Spectral Energy (SE) and Spectral Kurtosis (SK). Moreover, the demodulation performance is quantified by the proposed Signal Characterization Ratio (SCR). Secondly, the effectiveness of Enkurgram is validated by simulation signals under different noise interference situations, including white Gaussian noise, stochastic impulse interference, periodic impulse interference, and contaminated modulation wave. Finally, the proposed method is verified by the actual signals from centrifugal pump and propeller, respectively. The analysis results prove that the Enkurgram has satisfactory demodulation capability in dealing with signals under low Signal-to-Noise Ratio (SNR) levels or with non-Gaussian noise, which is the Achilles’ heel of SK-based methods.

An Improved Dual-Kurtogram-Based T 2 Control Chart for Condition Monitoring and Compound Fault Diagnosis of Rolling Bearings

Condition monitoring and compound fault diagnosis are crucial key points for ensuring the normal operation of rotating machinery. A novel method for condition monitoring and compound fault diagnosis based on the dual-kurtogram algorithm and multivariate statistical process control is established in this study. The core idea of this method is the capability of the dual-kurtogram in extracting subbands. Vibration data under normal conditions are decomposed by the dual-kurtogram into two subbands. Then, the spectral kurtosis (SK) of Subband I and the envelope spectral kurtosis (ESK) of Subband II are formulated to construct a control limit based on kernel density estimation. Similarly, vibration data that need to be monitored are constructed into two subbands by the dual-kurtogram. The SK of Subband I and the ESK of Subband II are calculated to derive T 2 statistics based on the covariance determinant. An alarm will be triggered when the T 2 statistics exceed the control limit and suitable subbands for square envelope analysis are adopted to obtain the characteristic frequency. Simulation and experimental data are used to verify the feasibility of the proposed method. Results confirm that the proposed method can effectively perform condition monitoring and fault diagnosis. Furthermore, comparison studies show that the proposed method outperforms the traditional T 2 control chart, envelope analysis, and empirical mode decomposition.

Vibration signatures in ball bearings as a function of lubricant viscosity ratio κ, under alternating lubrication conditions

This paper analyses the vibrations emitted by a set of grease lubricated ball bearings run under load, when lubrication conditions alternate between good and insufficient. Conditions are varied by adjusting the operational parameters rotational speed and temperature. The dependency of the vibration signature on the lubrication ratio κ is considered. The results show a good correlation between κ and spectral kurtosis for a given frequency band, a: when the bearing is new, and b: when κ is below 0.6.

Condition Evaluation in Steel Truss Bridge with Fused Hilbert Transform, Spectral Kurtosis, and Bandpass Filter

This study is concerned with the diagnosis of discrepancies in a steel truss bridge by identifying dynamic properties from the vibration response signals of the bridges. The vibration response signals collected at bridges under three different vehicular speeds of 10 km/hr, 20 km/hr, and 30 km/hr are analyzed using statistical features such as kurtosis, magnitude of peak-to-peak, root mean square, crest factor as well as impulse factor in time domain, and Stockwell transform in the time-frequency domain. The considered statistical features except for kurtosis show uncertain behavior. The Stockwell transform showed low-resolution outcomes when the presence of noise in the recorded vibration responses. The elimination of noise and extraction of meaningful dynamic properties from the vibration responses is done by applying a new method which comes from the fusion of Hilbert transform with Spectral kurtosis and bandpass filtering. The outcomes obtained from Hilbert transform processed residual signals which are further filtered using bandpass filter show more robustness and accuracy in characterizing bridge modal frequencies from the noisy vibration responses. The proposed method produces a high-resolution frequency response which can unveil the joint discrepancy in the bridge structure.

Bearing Defect Recognition Based on Optimal Impulse Response Wavelet Denoising Technique

A rolling bearing defect detection and recognition technique based on optimal impulse response continuous wavelet transform (IRCWT) de-noising technique is proposed. In the proposed technique, the IRCWT is used as a tool for local time and frequency characteristics analysis, and the spectral kurtosis (SK) is used as the impulse index to extract transient pulse components from the processed rolling defect vibration data. Firstly, the IRCWT is used to process the sampling vibration data. And then SK is calculated to choose the best signal band-pass band from the time frequency analysis of vibration data, in order to restrain the interference noise and highlight the transient shock feature. Finally, the envelope spectrum is computed and the rolling bearing defect feature is extracted. The rolling element bearing defect recognition results show that the proposed approach can effectively detect the transient characteristics and distinguish the rolling element bearing localized defect.

Spectral kurtosis based on evolutionary digital filter in the application of rolling element bearing fault diagnosis

Spectral kurtosis based on evolutionary digital filter in the application of rolling element bearing fault diagnosis

Early Fault Diagnosis of Shaft Crack Based on Double Optimization Maximum Correlated Kurtosis Deconvolution and Variational Mode Decomposition

Aiming at the problem that the vibration energy of shaft crack early weak fault is small and the feature extraction was difficult, the fault diagnosis method of cuckoo search, maximum correlation kurtosis deconvolution and variational mode decomposition (CS-MCKD-VMD) was proposed to diagnose the early weak fault of shaft cracks. Firstly, considering the periodic characteristics of fault signal, MCKD was used to highlight the fault signal, and the kurtosis correlation of deconvolution signal processed by MCKD were enhanced. Then, VMD could be used to decompose the harmonic signal, and VMD was used to process the deconvolution signal to generate some intrinsic mode functions (IMF), and the correlation coefficient method was used to calculate the components with the maximum correlation. At the same time, in order to improve the correlation of parameters in MCKD and VMD, CS was used to optimize the parameters to reduce the error of human selection. Thirdly, the optimal component was calculated by the fast spectral kurtosis (FSK), and the envelope spectrum was obtained by filtering the optimal component with the parameters obtained. Finally, the simulation and experimental results show that the transfer frequency and its 2 and 3 times frequency were prominent in the envelope spectrum, which was consistent with the analysis of shaft crack fault mechanism, Therefore, it was proved that the CS-MCKD-VMD method can effectively extract the early weak fault features of shaft cracks.

An FSK-MBCNN based method for compound fault diagnosis in wind turbine gearboxes

Accurate identification of compound fault in wind turbine gearbox is very important for condition monitoring of wind turbine systems. This paper proposes a novel compound fault diagnosis method named FSK-MBCNN for wind turbine gearboxes. The proposed method combines the fast spectral kurtosis (FSK) with a multi-branch convolutional neural network (MBCNN). First, FSK transforms the one dimensional vibration signal into the fast kurtogram of a two dimensional image as the input feature map of the convolutional neural network model. Second, a multi-branch module is proposed to provide suitable network architecture adapting to different learning rate and maximum training epoch. The experiment results reveal that the proposed FSK-MBCNN method can diagnose compound fault of wind turbine gearbox with more than 97% accuracy. The fault identification accuracies of noised signal and variable load signal are close to 90%, proving the robustness of the proposed method. Compared with other three methods, the proposed method can recognize 9 different statuses of the wind turbine gearboxes more accurately in the aforementioned cases. This fascinating discovery indicates that the proposed FSK-MBCNN method can be applied to actual condition monitoring of wind turbines to reduce the maintenance cost and the downtime.

Reweighted Dual Sparse Regularization and Convex Optimization for Bearing Fault Diagnosis

Performing early and accurate bearing fault diagnosis is of great significance. However, accurate extraction of the repetitive transients from noisy vibration signals is a critical issue. In this article, a reweighted dual sparse regularization (RDSR) method is proposed for bearing fault diagnosis, and the RDSR is a unified framework of the reweighted generalized minimax-concave (reGMC) penalty and the ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> -norm regularization. To be specific, the reGMC penalty and ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> -norm are both introduced in the sparse coding model. The reGMC is utilized to enhance the sparsity and overcome the underestimation deficiency of the ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm while restraining the interference. Moreover, the ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> -norm is employed to simplify the model and overcome the overfitting problem, which can make the optimization process rapid and stable. Therefore, it can observably enhance the accuracy of bearing fault diagnosis and preserve the amplitude of the fault transient component. Comparisons with ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm minimization model and the conventional spectral kurtosis (SK) show that the proposed method can preserve the spectral peak of the periodic transient components and improve the signal-to-noise ratio (SNR). Experimental case studies further verify that the proposed method can accurately estimate the periodic transient component from vibration signals, which demonstrates that the proposed method is feasible for bearing fault diagnosis.

Taylor-DBN: A new framework for speech recognition systems

Speech recognition is a rapidly emerging research area as the speech signal contains linguistic information and speaker information that can be used in applications including surveillance, authentication, and forensic field. The performance of speech recognition systems degrades expeditiously nowadays due to channel degradations, mismatches, and noise. To provide better performance of speech recognition, the Taylor-Deep Belief Network (Taylor-DBN) classifier is proposed, which is the modification of the Gradient Descent (GD) algorithm with Taylor series in the existing DBN classifier. Initially, the noise present in the speech signal is removed through the speech signal enhancement. The features, such as Holoentropy with the eXtended Linear Prediction using autocorrelation Snapshot (HXLPS), spectral kurtosis, and spectral skewness, are extracted from the enhanced speech signal, which is fed to the Taylor-DBN classifier that identifies the speech of the impaired persons. The experimentation is done using the TensorFlow speech recognition database, the real database, and the ESC-50 dataset. The accuracy, False Acceptance Rate (FAR), False Rejection Rate (FRR), and Mean Square Error (MSE) of the Taylor-DBN for TensorFlow speech recognition database are 96.95%, 3.04%, 3.04%, and 0.045, respectively, and for real database, the accuracy, FAR, FRR, and MSE are 96.67%, 3.32%, 3.32%, and 0.0499, respectively. Similarly, for the ESC-50 dataset, the accuracy, FAR, FRR, and MSE are 96.81%, 3.18%, 3.18%, and 0.047, respectively. The results imply that the Taylor-DBN provides better performance as compared to the existing conventional methods.

Correlation dimension and approximate entropy for machine condition monitoring: Revisited

The sparsity of signals is of great concern in various research domains. In mechanical systems and signal processing, repetitive transients are the symptoms of localized gear and bearing faults and they are sparse signals. During the recent years, sparsity measures, such as kurtosis and Shannon entropy, have been thoroughly studied to quantify repetitive transients for machine condition monitoring. Spectral kurtosis and spectral negative Shannon entropy are two typical examples of sparsity measures for machine condition monitoring. Besides sparsity measures, complexity measures including correlation dimension (CD) and approximate entropy (AE) have been experimentally studied during the recent years. However, theoretical investigations on these two complexity measures for machine condition monitoring are seldom reported. This paper aims to fill this research gap and propose some new theorems and proofs to show that CD and AE have a “bilateral reduction” effect, which is a proper measure of entropy. Specifically, CD with any dimension and AE with smaller dimension become smaller when a signal is getting sparser or more deterministic, which is significantly different from sparsity measures that are monotonically increasing when a signal is getting from more deterministic to sparser. This new discovery is able to help readers fully understand the main difference between sparsity measures and complexity measures. In view of this discovery, it is suggested that the concept of blind fault component separation should be used to separate low-frequency periodic components (a deterministic signal) from high-frequency repetitive transients (a sparse signal) before complexity measures are used for machine condition monitoring. This suggestion aims to avoid the uncertainty of machine condition monitoring caused by low-frequency periodic components and high-frequency repetitive transients.

Phonetic correlates of laryngeal and place contrasts of Burushaski

Burushaski is an endangered language isolate spoken in Hunza, Nager, and Yasin valleys of Gilgit, Northern Pakistan. The present study investigates the acoustic correlates of Hunza Burushaski’s three-way stop laryngeal contrast (voiceless unaspirated, voiceless aspirated, and voiced unaspirated) across five places of articulation (bilabial, dental, retroflex, palatal, and velar). A wide range of acoustic correlates were measured, including Voice Onset Time (VOT), fundamental frequency (f0), first four formants (F1, F2, F3, and F4), spectral moments of stop release bursts (spectral center of gravity, spectral standard deviation, spectral skewness, and spectral kurtosis), and spectral tilt (H1*–H2*, H1*–A1*, H1*–A2*, and H1*–A3*). The data were collected from four Burushaski speakers. The findings indicated that voiceless aspirated consonants exhibited longer Voice Onset Time and higher spectral tilt onsets than their voiceless unaspirated and voiced unaspirated counterparts. The fundamental frequency was raised in both voiceless series. Some acoustic measures (e.g., Voice Onset Time and fundamental frequency) were better indicators of the laryngeal contrasts while others (spectral moments) more reliably distinguished the place contrasts. The results of Linear Discriminant Analysis (LDA) showed that a combination of spectral tilt and Voice Onset Time are the best descriptors of the three laryngeal categories of Burushaski.