Regular ArticleFEATURE EXTRACTION BASED ON MORLET WAVELET AND ITS APPLICATION FOR MECHANICAL FAULT DIAGNOSIS
Abstract
The vibration signals of a machine always carry the dynamic information of the machine. These signals are very useful for the feature extraction and fault diagnosis. However, in many cases, because these signals have very low signal-to-noise ratio (SNR), to extract feature components becomes difficult and the applicability of information drops down. Wavelet analysis in an effective tool for signal processing and feature extraction. In this paper, a denoising method based on wavelet analysis is applied to feature extraction for mechanical vibration signals. This method is an advanced version of the famous “soft-thresholding denoising method” proposed by Donoho and Johnstone. Based on the Morlet wavelet, the time-frequency resolution can be adapted to different signals of interest. In this paper, this denoising method is introduced in detail. The results of the application in rolling bearing diagnosis and gear-box diagnosis are satisfactory.
References (12)
- C.J. LI et al.
Wavelet decomposition of vibrations for detection of bearing-localized defect
NDT&E International
(1997) - Z. GENG et al.
Vibrational diagnosis of machine parts using the wavelet packet technique
The British Journal of NDT
(1994) - W.J. STASZEWSKI
Wavelet based compression and feature selection for vibration analysis
Journal of Sound and Vibration
(1998) - J. ZHAO
Study for practical diagnosis techniques based on wavelets and neural networks
Doctoral dissertation, School of Mechanical Engineering, Xi'an Jiaotong University
(1997) - S. MALLAT et al.
Matching pursuits with time–frequency dictionaries
IEEE Transactions on Signal Processing
(1993)
Cited by (889)
Time–space–frequency feature Fusion for 3-channel motor imagery classification
2024, Biomedical Signal Processing and ControlLow-channel EEG devices are crucial for portable and entertainment applications. However, the low spatial resolution of EEG presents challenges in decoding low-channel motor imagery. This study introduces TSFF-Net, a novel network architecture that integrates time–space–frequency features, effectively compensating for the limitations of single-mode feature extraction networks based on time-series or time–frequency modalities. TSFF-Net comprises four main components: time–frequency representation, time–frequency feature extraction, time–space feature extraction, and feature fusion and classification. Time–frequency representation and feature extraction transform raw EEG signals into time–frequency spectrograms and extract relevant features. The time–space network processes time-series EEG trials as input and extracts temporal–spatial features. Feature fusion employs Maximum Mean Discrepancy (MMD) loss to constrain the distribution of time–frequency and time–space features in the Reproducing Kernel Hilbert Space, subsequently combining these features using a weighted fusion approach to obtain effective time–space–frequency features. Moreover, few studies have explored the decoding of three-channel motor imagery based on time–frequency spectrograms. This study proposes a shallow, lightweight decoding architecture (TSFF-img) based on time–frequency spectrograms and compares its classification performance in low-channel motor imagery with other methods using two publicly available datasets. Experimental results demonstrate that TSFF-Net not only compensates for the shortcomings of single-mode feature extraction networks in EEG decoding, but also outperforms other state-of-the-art methods. Overall, TSFF-Net offers considerable advantages in decoding low-channel motor imagery and provides valuable insights for algorithmically enhancing low-channel EEG decoding.
Mid-frequency vibration analysis of built-up structures using WFE-SEA method
2024, International Journal of Mechanical SciencesThe idea of classifying built-up structures into hybrid deterministic-statistical subsystems is formed as a framework to deal with the dynamical response in the mid-frequency region. This paper explores the potential combination of the wavelet finite element method and SEA, resulting in the hybrid WFE-SEA method, to achieve a solution with reduced computational burdens and increased accuracy. The development of the hybrid WFE-SEA method commences with the formulation of a 3-D B-spline wavelet theory, utilizing the semi-orthogonality of B-spline wavelet. A multi-resolution analysis of this theory facilitates the creation of a 3-D B-spline wavelet on an interval element model, incorporating nodes distributed on both the surface and inside of the hexahedron. Subsequently, the 3-D B-spline wavelet on an interval element is integrated with SEA theory, producing in a hybrid WFE-SEA framework suitable for examining dynamic problems of built-up structures. The energy-balance equation is derived by coupling the 3-D B-spline wavelet theory with diffuse field reciprocity relation to estimate the mid-frequency dynamic response. Furthermore, the validity of the proposed approach is rigorously assessed by comparing it against detailed mid-frequency analyses, including the hybrid FE-SEA method and the hybrid ES-FE-SEA method. The conclusion highlights the substantial agreement of the present hybrid method with other approaches, achieved at a small cost in terms of degrees of freedom.
Morse wavelet transform-based features for voice liveness detection
2024, Computer Speech and LanguageThe need for Voice Liveness Detection (VLD) has emerged particularly for the security of Automatic Speaker Verification (ASV) systems. Existing Spoofed Speech Detection (SSD) systems rely on attack-specific approaches to detect spoofed speech. However, to safeguard ASV systems against all the kinds of spoofing attacks (known as well as unknown attacks), determining whether a speech is uttered live (genuine) or not, is important. To that effect, in this work, we propose the detection of pop noise using Morse wavelet for VLD task. Pop noise is a discriminative acoustic cue that is present in live speech and is absent/diminished in spoofed speech. It is captured by the microphone in the form of sudden bursts of air from a live speaker’s mouth due to the close proximity of the speaker with the microphone. To validate this hypothesis, we present an analysis of pop noise energy w.r.t. distance and found that it decreases exponentially with distance. Furthermore, pop noise is said to be present in very low frequency regions. To capture the pop noise effectively, we propose to exploit the excellent frequency resolution of Continuous Wavelet Transform (CWT) using Generalized Morse Wavelets (GMWs). GMWs are a superfamily of analytic wavelets. To that effect, in this work, we have analysed the suitability of GMWs for pop noise detection for VLD task using the POp noise COrpus (POCO). The wavelet parameters are fine-tuned according to the VLD task. Furthermore, the performance of VLD system is evaluated for various subband frequencies, and it is observed that the subband of 1 to gives the best performance accuracy of 90.55% and 88.43% on the Dev and Eval sets, respectively. In addition, phoneme-based analysis shows the dependence of the performance of the VLD system on the type of phonemes in the utterances. It is shown that phonemes, such as plosives and fricatives show distinct pop noise as compared to other phonemes. Furthermore, the extension of the POCO dataset is used for experiments where simulated reverberation is added to spoofed signals, assuming the attacker (or the recording device) is positioned at various distances. This leads to the studying the effect of speaker-attacker distance. Similar to the previous results, it is observed that for the reverberated case too, the optimal frequency subband for VLD task is 1 to , across all the distances. Furthermore, the proposed feature set is evaluated using three classifiers, namely, Convolutional Neural Network (CNN), Light CNN (LCNN), and Residual Neural Network (ResNet), for POCO dataset as well as reverberated POCO dataset. It is observed that CNN gives the highest accuracy of 88.43% on Eval set of the POCO dataset. Furthermore, the proposed features are also evaluated under the assumptions of two ideal scenarios — when the ASV system is strictly under attack, and when it is strictly not under attack. It is observed that the proposed Morse wavelet-based VLD system rejected 89% of the spoofed utterances, and accepted 88.30% of the genuine utterances.
Blast wave induced strain measurements in polymers using FBG sensor inside shock tube
2024, Measurement: Journal of the International Measurement ConfederationPolypropylene and polytetrafluoroethylene (Teflon) have several applications in the aerospace industry and the study of their behaviour in high speed impact events is of great significance. This paper demonstrates the use of FBG sensors for measuring blast wave induced strain in aerospace grade polymer samples inside a Vertical Shock Tube (VST). FBG strain measurement assembly incorporating the polymer test sample and embedded FBG sensor has been designed and developed for suitably mounting it in the driven section of the VST. Experiments are carried out by impacting the polymer samples with blast waves of different peak pressures generated inside VST. PCB pressure transducer is used to record the blast pressure profile in the VST. A systematic study of strain profiles developed in the polymer test samples and their dependence on incident blast wave parameters has been carried out. The strain profiles measured by FBG in both the polymer samples are analysed using FFT and continuous wavelet transform analyses. FEM analysis is carried out using ABAQUS dynamic explicit to simulate the blast wave induced compression in polymers for different blast wave peak pressures and compared with experimental results. The application of FBG sensors to measure blast wave induced dynamic strain inside polymer materials and their capability to withstand the harsh environment of impulse test facilities like shock tubes has been demonstrated.
Online damage monitoring of inelastic seismic response of a reinforced concrete structure
2024, Mechanical Systems and Signal ProcessingMethods of early warning in damage detection of civil engineering structures represent a particular challenge in Structural Health Monitoring (SHM). To investigate these problems, vibration records of a full-size, reinforced concrete frame subjected to seismic damages on a shaking table, were investigated. The seismic recordings of excitations and response were accompanied by images of cracks associated with each state of the damage. This paper presents the results of the damage identification from non-linear response during the damaging phases of the shaking table tests. Time-dependent base shear forces were observed, together with wavelet spectral analyses of the responses. Three damage-sensitive features accounting for decreasing stiffness, energy dissipation, and the wavelet changes during inelastic seismic responses were analyzed as functions of horizontal peak table velocities. The paper aims to study seismic response records of moment-resisting reinforced concrete structure frames in search of prospective simple online damage indicators which could be applied to start alarms, right after excessive damage loads, e.g., in instantaneous earthquake rapid damage assessments of buildings.
Total process of fault diagnosis for wind turbine gearbox, from the perspective of combination with feature extraction and machine learning: A review
2024, Energy and AIWith the increasing of the installed capacity of wind power, the condition monitoring and maintains technique is becoming more important. Wind Turbines (WT) gearbox is one of the key wind power components as it plays the role of power transmission and speed regulation. Towards this, a number of scholars have pay attention to the fault diagnosis of WT gearbox. The efficiency of Machine Learning (ML) algorithms is highly correlated with signal type, data quality, and extracted features employed. The implementation of ML techniques has proven to be advantageous in simplifying the comprehension prerequisites for fault diagnosis technology concerning fault mechanisms. More and more current studies predominantly concentrate on the utilization and fine-tuning of ML algorithms, while providing limited insights into the features of the acquired data. Therefore, it is necessary to review the research in recent years from the perspective of the combination of feature extraction and ML algorithms, and provide a detailed direction for future WT gearbox fault diagnosis technology research. In this paper, data processing algorithms and typical fault diagnosis methods based on ML methods for WT gearbox are reviewed. For the using of ML method in WT gearbox fault diagnosis, the data prepared for training is very important. The paper firstly reviewed the data analysing method which will support the ML method. The data analysing methods include data acquisition, data preprocessing and feature extraction method. Feature extraction plays a pivotal role in the realm of gearbox fault diagnosis, as it serves as the essence of effective detection. This review will primarily focus on exploring methods that enable the utilization of efficient features in combination with ML techniques to achieve accurate gearbox fault diagnosis. Then typical ML method for WT gearbox fault diagnosis are carefully reviewed. Moreover, some prospects for future research directions are discussed in the end.