Showing papers on "Wavelet packet decomposition published in 2023"

Wavelet Packet Decomposition-Based Multiscale CNN for Fault Diagnosis of Wind Turbine Gearbox

Abstract: This article presents an intelligent fault diagnosis method for wind turbine (WT) gearbox by using wavelet packet decomposition (WPD) and deep learning. Specifically, the vibration signals from the gearbox are decomposed using WPD and the decomposed signal components are fed into a hierarchical convolutional neural network (CNN) to extract multiscale features adaptively and classify faults effectively. The presented method combines the multiscale characteristic of WPD with the strong classification capacity of CNNs, and it does not need complex manual feature extraction steps as usually adopted in existing results. The presented CNN with multiple characteristic scales based on WPD (WPD-MSCNN) has three advantages: 1) the added WPD layer can legitimately process the nonstationary vibration data to obtain components at multiple characteristic scales adaptively, it takes full advantage of WPD and, thus, enables the CNN to extract multiscale features; 2) the WPD layer directly sends multiscale components to the hierarchical CNN to extract rich fault information effectively, and it avoids the loss of useful information due to hand-crafted feature extraction; and 3) even if the scale changes, the lengths of components remain the same, which shows that the proposed method is robust to scale uncertainties in the vibration signals. Experiments with vibration data from a production wind farm provided by a company using condition monitoring system (CMS) show that the presented WPD-MSCNN method is superior to traditional CNN and multiscale CNN (MSCNN) for fault diagnosis.

Synchrosqueezed Fractional Wavelet Transform: A New High-Resolution Time-Frequency Representation

Abstract: The synchrosqueezed wavelet transform (SSWT) has been proven to be a powerful time-frequency analysis tool. However, this transform is unable to deal with signals with fast varying instantaneous frequencies. The objective of this paper is to overcome this deficiency using the fractional wavelet transform (FRWT), which is a generalization of the conventional wavelet transform. We first propose a synchrosqueezed FRWT (SSFRWT), which shares many properties of its SSWT counterpart while offering attractive new features. Then, we present a theoretical analysis of the SSFRWT, including the derivation of its basic properties. Moreover, we show that the discrete form of the SSFRWT admits efficient numerical implementation akin to that of the SSWT. Finally, the theoretical derivations are validated via simulations.

A Novel Pipeline Corrosion Monitoring Method Based on Piezoelectric Active Sensing and CNN

Abstract: In this study, a piezoelectric active sensing-based time reversal method was investigated for monitoring pipeline internal corrosion. An effective method that combines wavelet packet energy with a Convolutional Neural Network (CNN) was proposed to identify the internal corrosion status of pipelines. Two lead zirconate titanate (PZT) patches were pasted on the outer surface of the pipeline as actuators and sensors to generate and receive ultrasonic signals propagating through the inner wall of the pipeline. Then, the time reversal technique was employed to reverse the received response signal in the time domain, and then to retransmit it as an excitation signal to obtain the focused signal. Afterward, the wavelet packet transform was used to decompose the focused signal, and the wavelet packet energy (WPE) with large components was extracted as the input of the CNN model to rapidly identify the corrosion degree inside the pipeline. The corrosion experiments were conducted to verify the correctness of the proposed method. The occurrence and development of corrosion in pipelines were generated by electrochemical corrosion, and nine different depths of corrosion were imposed on the sample pipeline. The experimental results indicated that the classification accuracy exceeded 99.01%. Therefore, this method can quantitatively monitor the corrosion status of pipelines and can pinpoint the internal corrosion degree of pipelines promptly and accurately. The WPE-CNN model in combination with the proposed time reversal method has high application potential for monitoring pipeline internal corrosion.

Rapid diagnosis of corrosion beneath epoxy protective coating using non-contact THz-TDS technique

Abstract: A reliable and effective diagnosis of coating structure is important for further maintenance. For the problem of slow detection and identification using terahertz non-destructive testing technology in the industrial inspection, a rapid diagnosis algorithm based on the wavelet packet energy and support vector machine method was developed for quick evaluation of the epoxy protective coating. The process mainly included time domain signal acquisition of various epoxy protective coating samples detected by a terahertz pulse imaging system, wavelet packet energy parameters extraction as the diagnosis feature vectors, classification model establishment based on the support vector machine algorithm and coating status evaluation using a three-class classifier. The influence on classification accuracy by the various feature vectors inputs with the support vector machine classifier was analysed. Satisfying results were achieved when the relative wavelet packet energy was taken as diagnostic features. A strong defective area could be quickly identified and more detail targeted analysis could be implemented as needed. The time spent was significantly reduced compared to the terahertz imaging of the whole area along with the manual judgement. The analysis indicated that the proposed method would be very useful and can be effectively employed for the coating monitoring application.

Remote Sensing Image Fusion Based on Wavelet Transform

Abstract: A new remote sensing image fusion method based on wavelet transform has been devised for the construction of fusion multi-spectral images.The method utilizes the gradient filter given in matrix form and an improvement of the fusion algorithm proposed by Petrovic,and was applied to the fusion of Landsat Enhanced Thematic Mapper(ETM) images.The resulting images are superior to the original ETM images and gray fusion images.The method not only affords significant enhancement of spatial detail information,but also preserves the spectral information of the original multi-spectral image well.

Transformer Aging Diagnosis Method Based on Raman Spectroscopy Wavelet Packet-SPCA Feature Extraction

Abstract: Ensuring the safety and reliable operation of a power transformer is a top priority in grid work. There is an urgent need to study methods for measuring and evaluating the aging state of oil–paper insulation in older transformers. Raman spectroscopy is widely used in material detection; the spectra can represent the inner components of transformer oil and help diagnose the aging states of the transformer efficiently. In this article, thermal accelerated aging experiments were conducted, and oil–paper insulation samples with different aging states were obtained. The aging states of oil–paper insulation were tagged by the degree of polymerization (DP) value of the paper. Four wavelet packet decompositions were used for the Raman spectra, and sparse principal component analysis was used to extract the features in the wavelet packet coefficients. A 14-D feature was built and trained with a multiclassification support vector machine (SVM), and a diagnosis model for aging states was established. The results show that the accuracy of the diagnosis model reaches 94.9%. Seven samples of transformers in operation also verify the method’s effectiveness. This method has practical significance for quickly detecting the aging states of operating transformers.

Wavelet-Based Harmonic Magnitude Measurement in the Presence of Interharmonics

Abstract: The increasing proliferation of power electronic converters, nonlinear loads, and distributed generation are leading to increased levels of harmonic and interharmonics in power networks. As a consequence, power quality (PQ) has become a critical performance indicator for power utilities and end-users. This study proposes a novel harmonic estimation method based on the real-time stationary discrete wavelet packet transform (RT-SDWPT). The proposed technique decomposes an input signal into frequency bands with harmonic information at cutoff frequencies and uses a compensation strategy to estimate root mean square (RMS) values of harmonics at every sampling period. The performance and effectiveness of the proposed method are assessed using real measurement data from field cases and experimental setup. The real measurements include challenging scenarios with harmonics, subharmonics, interharmonics, frequency deviation, and non-stationary PQ events. The proposed method outperforms the harmonic estimation provided by the discrete Fourier transform (DFT)-based approach and existing wavelet packet-based methods in terms of accuracy and speed.

Denoising method of weak fault acoustic emission signal under strong background noise of engine based on autoencoder and wavelet packet decomposition

Abstract: The weak fault acoustic emission (AE) signals collected in the actual operating conditions of the engine are often submerged in the strong background noise. This paper proposes a denoising method of AE signals based on the combination of autoencoder and wavelet packet decomposition (AE-WPD) to address the above problem. Firstly, the wavelet packet is used to decompose engine background noise signals and noise-containing fault AE signals to enhance the local analysis capability of the autoencoder. Then, the dataset of each frequency band after decomposition is created. Among them, background noise signals are regarded as normal datasets. Noise-containing fault signals are treated as outlier datasets. The difference between each frequency band of background noise signals and noise-containing fault signals is analyzed. The autoencoder model is trained, validated and tested for effectiveness. In addition, a comparison is made with other commonly used denoising methods. Four types of evaluation indexes are introduced to quantitatively assess various methods. Finally, the real engine background noise signals with different signal-to-noise ratio (SNR) are added to the fault AE signals to verify the robustness of the proposed AE-WPD method. The experimental results show that the proposed AE-WPD method outperforms other denoising methods at different SNR. This lays the foundation for engine structural condition monitoring and subsequent fault identification and localization.

Research on fault diagnosis of motor sound signal based on RBF neural network

Abstract: In order to achieve accurate diagnosis of motor faults, a technique based on wavelet analysis and RBF neural networks is used. The wavelet thresholding method is first used to reduce the noise of the motor sound and improve the signal-to-noise ratio in order to further extract fault features. Then the wavelet packet method is used to analyze the sound signals of the three-phase asynchronous motor in three states to extract the band energy, and finally the band energy is fed into the neural network for training to build a classifier for fault diagnosis. The experimental results show that the method of combining wavelet packet technology and RBF neural network has less time consumption and higher accuracy in diagnosing motor faults. It has the potential for further development.

Short Circuit Failure Detection in Induction Motor Using Wavelet Transform and Fuzzy C-Means

Abstract: Induction motors need to be monitored regularly because it involves the company's productivity. The induction motor monitoring method in this study uses a motor current variable which is transformed using the Discrete Wavelet Transform. Discrete Wavelet Transform (DWT) is used in this study because the results are satisfactory for detecting a short circuit in the stator winding of an induction motor. Of the many types and levels of discrete wavelet transforms, the haar wavelet transform at the third level is used in this study. Furthermore, the results of the discrete wavelet transform are processed using the Fuzzy C-means method. Fuzzy C-Mean (FCM) is the grouping approach that each part has a member degree of cluster according to the fuzzy logic algorithm. Motor modeling is shown in this article as normal condition, final fault current, and initial fault current. For this analysis, a combination of wavelet transform and Fuzzy C-means is used to classify motor currents into three motor states. The motor current is processed by Haar DWT level 3 to generate a high frequency signal. Then the high frequency signal is processed to get the energy signal. The energy signal is then fed to Fuzzy C-means to identify its condition. The results show that fuzzy C-means produces an error of 0% for the normal case, 33.3% for the initial error case and 0% for the final error case.

Recognition of Hand Gestures using Wavelet Packet Transform and Cascaded Feed Forward Neural Networks

Abstract: Abstract This paper presents a noval framework that classifies finger movements automatically using Wavelet Transform and its derivatives by capturing statistical features from the discrete time Electromyogram (EMG) signals. In the suggested method, wavelet-based denoising is used to separate out the subject's EMG signals, and then Discrete Wavelet Transform (DWT) and Wavelet Packet Transform are used to decompose the signals and extract their key characteristics (WPT). The derivatives of the feature sets are employed to analyse the correlation among them. This method is motivated by the surveillance that there exists a distinctive correlation between the different features of the samples of the signals extracted at various frequency levels. Experimentally, it was perceived that this correlation varies from signal to signal. Both Feed forward and Cascaded Feed forward Artificial Neural Networks (ANN) are used for classification. Experiments show that the proposed method significantly improves the classification rate. The performance of the suggested wavelet-based features and their derivatives in combination with ANN and trained with the Levenberg-Marquardt algorithm was evaluated by comparing the simulation results for various sets of features. Comparing the new method benefits to earlier traditional methods in terms of classification performance helped to further highlight their advantages. These experimental findings demonstrate that the suggested approach performs admirably in classifying finger movements based on EMG signal patterns. The suggested methodology also helps clinicians increase the reliability of myoelectric pattern recognition.

On Some Properties of a New PoisN Wavelet Family

Abstract: This paper presents a family of new real valued continuous wavelet functions derived from Poisson kernel. Properties like Admissibility condition, vanishing properties, frequency response and time-frequency bandwidth are studied. A comparison is made between new wavelet family “PoisN” with Gaussian wavelet family. Finally, the de-noising capability of the proposed new wavelet is demonstrated with a noisy version of Gaussian function with different values of SNR.

The Sparsity of Cycle Spinning for Wavelet-Based Solutions of Linear Inverse Problems

Abstract: The usual explanation of the efficacy of wavelet-based methods hinges on the sparsity of many real-world objects in the wavelet domain. Yet, standard wavelet-shrinkage techniques for sparse reconstruction are not competitive in practice, one reason being that the lack of shift-invariance of the wavelet transform produces blocky artifacts. The standard remedy is cycle spinning, which results in a substantial reduction of these artifacts. In this letter, we propose a theoretical investigation of the sparsity of solutions to the cycle-spinning variant of wavelet-based resolutions of linear inverse problems. We derive a representer theorem that provides a complete characterization of the solution set. Our theorem indicates that the solutions are typically not sparse, where sparsity is measured with respect to the wavelet dictionary. This exposes that the role of sparsity in the success of wavelet-based solutions of linear inverse problems requires further investigation. We corroborate our theoretical results with numerical examples for the problem of image denoising.

Research on Bearing Fault Diagnosis Based on 1DCNN with Fault Compound Features

Abstract: Spectral kurtosis analyses and wavelet packet transform are particularly suitable for fault feature extraction of non-stationary signals of rotating machinery. This paper proposes a composite analysis algorithm based on spectral kurtosis analysis and wavelet packet transform to study the deep learning fault diagnosis method combined with the bearing database of Western Reserve University. The feature data set is constructed based on the dimensionless and non-dimensional signal features such as signal mean value and waveform index, frequency domain features such as spectrum, spectral kurtosis analysis, and time-frequency domain comprehensive features of wavelet packet transform. Then feature data set is used as CNN deep learning network for fault learning and recognition. The experimental results show that the algorithm can achieve more than 97% fault recognition rate and is effectively applied to bearing fault diagnosis tasks.

Edge Detection Algorithm for Biological Slice Images Based on Empirical Wavelet Transform and Morphology

Abstract: Edge detection is important in extracting image features, and microscopic slice images consist of closed-loop structures and complex internal textures, and extracting the corresponding features has an important role in biology, epidemiology, pathology and other fields. In this study, an edge detection algorithm for slice images based on empirical wavelet transform (EWT) and morphology is proposed. The empirical wavelet divides the Fourier spectrum of the signal into successive intervals, and then constructs a wavelet filter bank for filtering in the corresponding interval segments, and finally obtains the amplitude modulation frequency components by signal reconstruction. The empirical wavelet transform overcomes the modal aliasing problem caused by the scale discontinuity in the time domain, which reflects the characteristics of the empirical wavelet transform. The image components extracted by the empirical wavelet are then enhanced using a morphological algorithm, which can effectively extract the closed-loop edges of the sliced image as well as the significant textures inside. In this paper, the proposed method is tested on locust slice images as an example. The proposed algorithm can also be effectively applied to other biological cross-sectional images.

Loose bolt detection method based on wavelet packet decomposition with CNN

Abstract: Loose bolts in PV mounts can affect the smooth operation of the system. If the loose bolts are not detected in time, the PV modules may fall off and cause the mounts to collapse in serious cases. The ultrasonic wavelet packet decomposition (WPD) combined with convolutional neural network (CNN) is proposed as a detection method to address the problem that the ultrasonic signal collected when the bolt is loose indistinctively. First, the ultrasonic signal with bolt loosening information is decomposed into four layers of wavelet packets to extract the energy composition of the feature vectors of each sub-band signal; second, the CNN model is designed and the network is trained with the feature vectors as samples; finally, the discrimination of bolt loosening conditions is achieved. The feasibility of the method is verified through experiments, and the experimental results show that the accuracy of the model is improved from 95.64% to 99.92% compared with the original data training model, and from 86.96% to 99.92% compared with the support vector machine (SVM).

Existence of uncertainty minimizers for the continuous wavelet transform

Abstract: Continuous wavelet design is the endeavor to construct mother wavelets with desirable properties for the continuous wavelet transform (CWT). One class of methods for choosing a mother wavelet involves minimizing a functional, called the wavelet uncertainty functional. Recently, two new wavelet uncertainty functionals were derived from theoretical foundations. In both approaches, the uncertainty of a mother wavelet describes its concentration, or accuracy, as a time-scale probe. While an uncertainty minimizing mother wavelet can be proven to have desirable localization properties, the existence of such a minimizer was never studied. In this paper, we prove the existence of minimizers for the two uncertainty functionals.

Comparison of decomposition levels for wavelet transform based energy management in multi-fuel cell system

Abstract: In a hybrid energy storage system consisting of multi-fuel cell systems and super-capacitors, the wavelet transform is adopted to decompose the load power into multiple levels and assign the low-frequency component for fuel cell (FC) systems to reduce the impact caused by frequent variations in power demand. Two strategies—wavelet-based direct strategy and wavelet-quadratic strategy—are simulated in order to explore the system output characteristics with an increasing decomposition level. It is shown that the wavelet-quadratic strategy decreases the frequency of FC fluctuations as the decomposition level increases, which facilitates delayed degradation. The direct strategy improves the efficiency but sacrifices the FC stack health due to frequent load changes.

Extraction of Fault Characteristic Frequency Band of Photovoltaic DC Series Arc Based on Wavelet Packet Transform and Tsallis Entropy

Abstract: The fault characteristics of photovoltaic (PV) DC series arc are scattered in the MHz-level broadband, so extracting the fault characteristic frequency band to enhance the characteristic information is of great significance for the efficient detection of arc faults. In this paper, a Cassie model-based photovoltaic DC series arc fault simulation is first established in PSACD. The output current of the photovoltaic array with two power frequency cycles before and after the fault is used as the signal analysis unit. The combination of mother wavelet and decomposition layer number is optimized by using Tsallis wavelet packet singular entropy (TWPSE). Then, based on the optimal combination, the signal analysis unit is decomposed and reconstructed by wavelet packet, and the Tsallis entropy ratio before and after the fault of each frequency band is calculated to determine the fault characteristic frequency band. Finally, the energy ratio analysis is carried out on the characteristic frequency bands obtained by the optimal and non-optimal combinations, and it is found that the fault characteristic frequency band extracted by the former has the largest energy ratio. The results show that the fault characteristic frequency band extracted by optimal combination can effectively enhance the fault feature, and is more conducive to efficient detection of photovoltaic DC series arc faults.

A useful method of signal de-noising

Abstract: Signal de-noising process is the process of signal characteristic components extraction. In view of the singular value decomposition de-noising method is difficult to separate under low SNR, through observing the singular value distribution under different decomposition order, select the appropriate number of decomposition, and combining the wavelet packet de-noising ability and good frequency resolution, put forward a kind of effective signal de-noising method. Firstly, based on different SNR to choose the appropriate number of singular value decomposition, select bigger singular value reconstructing the signal, preliminary to remove noise, and then make three layer wavelet packet decomposition, compare the energy distribution of all nodes, and make selection of energy concentration of several nodes, reconstruct the signal. Simulation results verify the validity of this method.

Performance Analysis of Real and Complex Wavelet Transform Techniques used for Wavelet-Based Image Denoising

Abstract: Image de noising is a principal technique majorly used for original image restoration, segmentation and image classification. It is basically used to refine the images by eliminating noise embedded. In the current work, authors present a denoising technique based on Wavelet Domain Filtering. Denoising of images after domain transform helps in separating the noise and data components. The discrete wavelet transform and dual tree complex wavelet transforms work on the analysis and synthesis filter banks to filter and further segment the noisy input signal to low frequency and high frequency components constituting data artifacts and noise respectively. The progressive decomposition of data to a particular number of levels finally results in a noise-free output after filtering, considering a particular threshold. A comparative analysis of thresholding techniques is presented and evaluated for the parameters Signal to Noise Ratio (SNR) and lowest Root Mean Square Error Value (RMSE). The simulation results indicate superior performance of dual tree complex wavelet transform(DTCWT) when compared to the discrete wavelet transform.