Time Frequency Domain Features Research Articles

An approach for feature extraction and selection from non-trending data for machinery prognosis

With the paradigm shift towards prognostic and health management (PHM) of machinery, there is need for reliable PHM methodologies with narrow error bounds to allow maintenance engineers take decisive maintenance actions based on the prognostic results. Prognostics is mainly concerned with the estimation of the remaining useful life (RUL) or time to failure (TTF). The accuracy of PHM methods is usually a function of the features extracted from the raw data obtained from sensors. In cases where the extracted features do not display clear degradation trends, for instance highly loaded bearings, the accuracy of the state of the art PHM methods is significantly affected. The data which lacks clear degradation trend is referred to as non-trending data. This studypresents a method for extracting degradation trends from nontrending condition monitoring data for RUL estimation. The raw signals are first filtered using a discrete wavelet transform (DWT) denoising filter to remove noise from the acquired signals. Time domain, frequency domain and timefrequency domain features are then extracted from the filtered signals. An autoregressive (AR) model is then applied to the extracted features to identify the degradation trends. Features representing the maximum health information are then selected based on a performance evaluation criteria using extreme learning machine (ELM) algorithm. The selected features can then be used as inputs in a prognostic algorithm. The feasibility of the method is demonstrated using experimental bearing vibration data. The performance of the method is evaluated on the accuracy of RUL estimation and the results show that the method can be used to accurately estimate RULwith a maximum error of 10%.

Characteristic Study and Time-Domain Discrete- Wavelet-Transform Based Hybrid Detection of Series DC Arc Faults

DC arc fault introduces major safety concerns in a wide variety of components in dc networks. However, the randomness and instability of dc arc makes it difficult to be detected. In this paper, an experimental system was designed and tested at different load current, dc source voltage, and gap length to evaluate the impact of each parameter to the dc arc. Based on the experimental results, characteristics in the electrical behavior were studied and fault detection oriented analysis was conducted. A detection algorithm utilizing both time and time-frequency domain characteristics was proposed to differentiate between dc arc fault and normal condition. The detection algorithm was then realized on a digital signal processing board and tested to verify the effectiveness. Experimental results show that the proposed algorithm can detect arc fault in a timely manner and is free of nuisance trip from normal circuit operations such as load change condition.

A rough membership neural network approach for fault classification in transmission lines

Objective: This paper presents a new approach for fault classification in extra high voltage (EHV) transmission line using a rough membership neural network (RMNN) classifier.Methods: Wavelet transform is used for the decomposition of measured current signals and for extraction of ten significant time–frequency domain features (TFDF), as well as three distinctive time domain features (TDF) particularly in terms of getting better classification performance. After extracting useful features from the measured signals, a decision of fault type of a transmission line is carried out using ten RMNN classifiers. Furthermore, to reduce the training times of the neural network, the rough neurons are used as input layer neurons, and the fuzzy neurons are utilized in hidden and output layer in each RMNN. And the Back Propagation (BP) algorithm is employed for determining the optimal connection weights between neurons of the different layers in the RMNN.Results and Conclusions: To verify the effectiveness of the proposed scheme, extensive simulations have been carried out under different fault conditions with wide variations in fault type, fault resistance, fault location and fault inception angle. Simulations results show that the proposed scheme is faster and more accurate than the back-propagation neural network (BPNN), and it is proved to be a robust classifier for digital protection.

Wavelet based Fault Classification for Rolling Element Bearing in Induction Machine

Induction motors plays the most important role in any industry. Induction motor faults results in motor failure causing breakdown and great loss of production due to shutdown of industry and also increases the running cost of machine with reduction in efficiency. This needs for early detection of fault with diagnosis of its root cause. In this research paper a wavelet based fault classification method has been developed for rolling element bearing in induction motor using vibration signal. Wavelet based vibration analysis is one of the most successful techniques used for condition monitoring of rotating machines. This paper describes a new condition monitoring method for induction motors based on wavelet transform. A robust bearing fault detection scheme has been developed by time-frequency domain feature extraction from vibration signals of healthy and defective machine. General Terms Induction motor, fault detection, ball bearing, wavelet transform, condition monitoring.

Life grade recognition method based on supervised uncorrelated orthogonal locality preserving projection and K-nearest neighbor classifier

A novel life grade recognition method based on Supervised Uncorrelated Orthogonal Locality Preserving Projection (SUOLPP) and K-nearest neighbor classifier (KNNC) is proposed in this paper. A time–frequency domain feature set is first constructed to completely extract the feature of different life grades, then SUOLPP is proposed to automatically compress the high-dimensional time–frequency domain feature sets of training and test samples into the low-dimensional eigenvectors with better discrimination, and finally the low-dimensional eigenvectors of training and test samples are input into KNNC to conduct life grade recognition. SUOLPP algorithm considers both local information and label information in designing the similarity matrix, and requires the output basis vectors to be statistically uncorrelated and orthogonal in order to improve the life grade feature extraction power of OLPP. KNNC ranks the test samples׳ neighbors among the training samples and uses the class labels of similarity neighbors to classify the unknown input test samples, so that it has such advantages as less calculation amount, finer timeliness and higher pattern recognition accuracy compared with support vector machine (SVM) and Fuzzy C-Means Clustering (FCM). The life grade recognition example on deep groove ball bearings demonstrated the effectivity of the proposed life grade recognition method.

Vibration sensor based tool condition monitoring using ν support vector machine and locality preserving projection

Reliable online monitoring of the tool condition is paramount for automatic machining process. C-support vector machine (C-SVM) has got many successful applications in the field of tool wear monitoring. However, the selection of penalty parameter C is usually realized based on optimization process, which increases the training time of the classifier greatly. In this paper, ν support vector machine (ν-SVM) is presented to realize multi categories tool wear classification. In this model, C is replaced by a new parameter ν which represents an upper bound on the fraction of training errors and a lower bound of the fraction of support vectors. At the same time, the nearest neighbor (NN) based rule is proposed to realize the fast selection of ν based on training samples. In addition, to further improve training speed and classification accuracy, locality preserving projection (LPP) method is utilized to reduce the dimension of feature vectors by extracting the lower dimensional manifold characteristics. To testify the effectiveness of the proposed method, milling experiment of Ti6Al4V alloy was carried out and vibration signals corresponding to four kinds of tool wear status were collected. Time domain and frequency domain features are extracted based on wavelet packet decomposition and dimension reduction is realized by using LPP algorithm. Based on the selected features, both C-SVM and ν-SVM are utilized to realize the classification of multi categories tool wear status. The analysis shows that the combination of NN based ν-SVM with LPP can realize faster training of classifier without sacrificing the classification accuracy.

Fault Feature Analysis of a Cracked Gear Coupled Rotor System

Considering the misalignment of gear root circle and base circle and accurate transition curve, an improved mesh stiffness model for healthy gear is proposed, and it is validated by comparison with the finite element method. On the basis of the improved method, a mesh stiffness model for a cracked gear pair is built. Then a finite element model of a cracked gear coupled rotor system in a one-stage reduction gear box is established. The effects of crack depth, width, initial position, and crack propagation direction on gear mesh stiffness, fault features in time domain and frequency domain, and statistical indicators are investigated. Moreover, fault features are also validated by experiment. The results show that the improved mesh stiffness model is more accurate than the traditional mesh stiffness model. When the tooth root crack appears, distinct impulses are found in time domain vibration responses, and sidebands appear in frequency domain. Amplitudes of all the statistical indicators ascend gradually with the growth of crack depth and width, decrease with the increasing crack initial position angle, and firstly increase and then decrease with the growth of propagation direction angle.

Methods of Fault Diagnosis in Fiber Optic Current Transducer Based on Allan Variance

To ensure low failure and high reliability of fiber optic current transducers (FOCTs), it is urgent to study methods of condition monitoring and fault diagnosis in FOCT. Faults in FOCT have statistical characteristics. With the analyzing of time domain and frequency domain features in fiber optic current transformers’ measurement data, we establish correspondence between the physical characteristics of key components in transformer and data features and then build diagnostic analysis model based on Allan variance. According to the Allan variance calculation results, we can diagnose fiber optic current transformer’s health state and realize faults location. Experiment results show that diagnostic methods based on Allan variance are accurate and effective to identify fault features.

Long-term Activities Segmentation using Viterbi Algorithm with a k-minimum-consecutive-states Constraint

In the last years, several works have made use of acceleration sensors to recognize simple physical activities like: walking, running, sleeping, falling, etc. Many of them rely on segmenting the data into fixed time windows and computing time domain and/or frequency domain features to train a classifier. A long-term activity is composed of a collection of simple activities and may last from a few minutes to several hours (e.g., shopping, exercising, working, etc.). Since long-term activities are more complex and their duration varies greatly, generating fixed length segments is not suitable. For this type of activities the segmentation should be done dynamically. In this work we propose the use of the Viterbi algorithm on a Hidden Markov Model with the addition of a k-minimum-consecutive-states constraint to perform the long-term activity recognition and segmentation from accelerometer data. This constraint allows the algorithm to perform a more informed search by incorporating prior knowledge about the minimum duration of each long-term activity. Our experiments showed good results for the activity recognition task and it was demonstrated that the accuracy was significantly increased by adding the k-minimum-consecutive-states constraint.

基于BP神经网络的滚动轴承故障监测研究

滚动轴承是机械设备中最常用的零件之一，它能否正常运行关系到整台机器的安全，所以对滚动轴承进行故障诊断具有重大的意义。本文搭建状态监测系统平台采集正常轴承和故障轴承的振动信号，根据特征选取原则，提取时域和频域特征并将特征值归一化。根据已知状态的轴承特征值训练BP神经网络，之后利用已建立的网络识别未知状态的轴承，经过试验验证，该方法得到了非常好的监测效果。 The rolling bearing is one of the most common parts in machinery and equipment. Its normal running is related to the safety of the whole machine. So the fault diagnosis of rolling bearing is of great significance. In this paper, the platform system to collect normal bearing and the fault bearing vibration signals is established. According to the principle of feature selection, the time domain and frequency domain features are extracted and the eigenvalues are normalized. BP neural network has been trained based on the bearing characteristics of known state. Bearing of unknown state is identified by using the established network, and it is proved that this method has got a very good effect.

The analysis of surface EMG signals with the wavelet-based correlation dimension method.

Many attempts have been made to effectively improve a prosthetic system controlled by the classification of surface electromyographic (SEMG) signals. Recently, the development of methodologies to extract the effective features still remains a primary challenge. Previous studies have demonstrated that the SEMG signals have nonlinear characteristics. In this study, by combining the nonlinear time series analysis and the time-frequency domain methods, we proposed the wavelet-based correlation dimension method to extract the effective features of SEMG signals. The SEMG signals were firstly analyzed by the wavelet transform and the correlation dimension was calculated to obtain the features of the SEMG signals. Then, these features were used as the input vectors of a Gustafson-Kessel clustering classifier to discriminate four types of forearm movements. Our results showed that there are four separate clusters corresponding to different forearm movements at the third resolution level and the resulting classification accuracy was 100%, when two channels of SEMG signals were used. This indicates that the proposed approach can provide important insight into the nonlinear characteristics and the time-frequency domain features of SEMG signals and is suitable for classifying different types of forearm movements. By comparing with other existing methods, the proposed method exhibited more robustness and higher classification accuracy.

A Novel Local Time-Frequency Domain Feature Extraction Method for Tool Condition Monitoring Using S-Transform and Genetic Algorithm

This paper investigates an online effective method for tool condition monitoring. Acoustic emission signal of a system which is acquired by a sensor mounted to the spindle of the milling machining center is used as the fault indicator because it is easily to be installed, inexpensive and practical for use in industrial environment. Time-frequency analysis is selected for signal processing step based on its ability to reveal time and frequency variant characteristics of faulty signal. S-transform is used as a powerful time-frequency method for this purpose. Because of the high dimension of the time-frequency results, it is desirable to use a local region of interest in time-frequency domain instead of using the entire information, for fast and accurate monitoring and detection when any abnormal/fault operating condition might occur. Such a strategy also helps to reduce the computation cost which is necessary for online applications and improves the interpretation resolution for law quality signals. An optimization method based on genetic algorithm is used for finding the most discriminative local area as the region of interest in time-frequency domain. For feature generation step, a correlation coefficient between each signal and the healthy signal is assigned to the signal using a 2-D correlation analysis. Curve fitting approach is then used to determine a function to approximate the fault value based on the correlation coefficients. Experimental results based on a milling machine under different operating conditions show that this method has a high accuracy for fault detection. It is also concluded that the accuracy of the local feature extraction is higher than the conventional ways.

A Multi-Classifier Approach to MUAP Classification for Diagnosis of Neuromuscular Disorders

The shapes and sounds of isolated motor unit action potentials (MUAPs) in an electromyographic (EMG) signal provide a significant source of information for diagnosis, treatment and management of neuromuscular disorders. These parameters can be analyzed qualitatively by an expert or quantitatively by using pattern recognition techniques. Due to the advantages of quantitative EMG method, developing robust automated MUAP classifiers have been explored and several systems have been developed for this purpose by now, but the accuracy of the existing methods is not high enough to be used in clinical environments. In this paper, a novel classification strategy based on ensemble of support vector machines (SVMs) classifiers in hybrid serial/parallel architecture is proposed to determine the class label (myopathic, neuropathic, or normal) for a given MUAP. The developed system employs both time domain and time-frequency domain features of the MUAPs extracted from an EMG signal using an EMG signal decomposition system. Different classification strategies including single classifier and multiple classifiers with several subsets of features were investigated. Experimental results using a set of real EMG signals showed robust performance of multi-classifier methods proposed here. Of the methods studied, the multi-classifier that uses multiple features sets and a combination of both trainable and nontrainable fusion techniques to aggregate base classifiers showed the best performance with average accuracy of 97% which is significantly higher than the average accuracy of single SVM-based classifier system (i.e., 88%).

Baseline characteristics of cervical auscultation signals during various head maneuvers

Cervical auscultation (CA) is an emerging method of assessing swallowing disorders that is both non-invasive and inexpensive. This technique utilizes microphones to detect acoustic sounds produced by swallowing activity and characterize its behavior. Though some properties of swallowing sounds are known, there is still a need for a complete understanding of the baseline characteristics of cervical auscultation signals as well as how they change due to the patient's head motion, age, and sex. In order to examine these parameters, data was collected from 56 healthy adult participants that performed six different head movement tasks without swallowing. After preprocessing the signal, features were extracted. Dependent variables were time domain, frequency domain and time-frequency domain features. Statistical tests showed that only the skewness and peak frequency were not statistically different for all tasks. The peak frequency results indicate that head movement does not significantly affect the microphone signal, and that it is unnecessary to filter out the lowest frequency components. No sex differences were observed in the extracted features, but several features exhibited age dependence.

Fault Diagnosis of Rolling Bearing Based on the PSO-SVM of the Mixed-Feature

The rolling bearing is one of the most important and widely used parts in the rotating machinery. It is necessary to establish a reliable condition monitoring program which can avoid serious fault in the runtime and diagnose failure timely and accurately when it happens. This paper puts forward to a fault diagnosis method of rolling bearing based on the PSO-SVM of the mixed-feature. Firstly, we extract features in time domain, frequency domain, and order quenfrency domain. Secondly, select both Support Vector Machine (SVM) parameters by Particle Swarm Optimization (PSO) algorithm and kernel function of SVM classification model. Finally, classification model of SVM is designed by using the extracted salient features, kernel function and optimal parameter of PSO. The result verifies the effectiveness of the proposed method.

Research on detection of welding penetration state during robotic GTAW process based on audible arc sound

PurposePenetration state is one of the most important factors for judging the quality of a gas tungsten arc welding (GTAW) joint. The purpose of this paper is to identify and classify the penetration state and welding quality through the features of arc sound signal during robotic GTAW process.Design/methodology/approachThis paper tried to make a foundation work to achieve on‐line monitoring of penetration state to weld pool through arc sound signal. The statistic features of arc sound under different penetration states like partial penetration, full penetration and excessive penetration were extracted and analysed, and wavelet packet analysis was used to extract frequency energy at different frequency bands. The prediction models were established by artificial neural networks based on different features combination.FindingsThe experiment results demonstrated that each feature in time and frequency domain could react the penetration behaviour, arc sound in different frequency band had different performance at different penetration states and the prediction model established by 23 features in time domain and frequency domain got the best prediction effect to recognize different penetration states and welding quality through arc sound signal.Originality/valueThis paper tried to make a foundation work to achieve identifying penetration state and welding quality through the features of arc sound signal during robotic GTAW process. A total of 23 features in time domain and frequency domain were extracted at different penetration states. And energy at different frequency bands was proved to be an effective factor for identifying different penetration states. Finally, a prediction model built by 23 features was proved to have the best prediction effect of welding quality.

Bearing degradation process prediction based on the PCA and optimized LS-SVM model

Bearing degradation process prediction is extremely important in industry. This paper proposed a new method to achieve bearing degradation prediction based on principal component analysis (PCA) and optimized LS-SVM method. Firstly, the time domain, frequency domain, time–frequency domain features extraction methods are employed to extract the original features from the mass vibration signals. However, the extracted original features still with high dimensional and include superfluous information, the multi-features fusion technique PCA is used to merge the original features and reduce the dimension, the typical sensitive features are extracted. Then, based on the extracted features, the LS-SVM model is constructed and trained for bearing degradation process prediction. The pseudo nearest neighbor point method is used to determine the input number of the model. The particle swarm optimization (PSO) is used to selected the LS-SVM parameters. An accelerated bearing run-to-failure experiment was carried out, the results proved the effectiveness of the methodology.

Feature Analysis of Advanced Radar Emitter Signals Based on Continuous Wavelet Transform

For further study the recognition problem of radar emitter signals (RES), the theory of continuous wavelet transform (CWT) and gray moment are introduced into the feature extraction of RES. A new approach for RES feature extraction was proposed based on CWT and gray moment. By using the time-frequency domain characteristics of wavelet analysis and the moment-based method, the CWT coefficients of RES and the changing rules of RES gray moment were researched. The experiment results shows that the wavelet gray moments of the RES take on a rising trend along with the increase of the order, and the four-order wavelet gray moment vector can express the local information of RES more exactly. This is useful for further research on RES feature extracting.

Fault Feature Extraction Method of Large Rotating Machinery

Working condition of large rotating machine is varying. The collected vibration signals contain fault information as well as non-fault information, like load change and noise. The traditional fault feature extraction method which based on energy changing has certain limitations; therefore, new fault feature extraction method based on Fast ICA will be researched. Separating independent signals from blind signals by adopting Independent Component Analysis(ICA), purifying fault information and suppressing interference information; and the lifting wavelet packet is used for acquiring time frequency domain feature band from signals, so as to solve the difficulty that the fault information of the variable working condition rotating machine is always submerged by irregular working condition changing information such that effective fault prediction is hard to carry out.

Spectral Regression Based Fault Feature Extraction for Bearing Accelerometer Sensor Signals

Bearings are not only the most important element but also a common source of failures in rotary machinery. Bearing fault prognosis technology has been receiving more and more attention recently, in particular because it plays an increasingly important role in avoiding the occurrence of accidents. Therein, fault feature extraction (FFE) of bearing accelerometer sensor signals is essential to highlight representative features of bearing conditions for machinery fault diagnosis and prognosis. This paper proposes a spectral regression (SR)-based approach for fault feature extraction from original features including time, frequency and time-frequency domain features of bearing accelerometer sensor signals. SR is a novel regression framework for efficient regularized subspace learning and feature extraction technology, and it uses the least squares method to obtain the best projection direction, rather than computing the density matrix of features, so it also has the advantage in dimensionality reduction. The effectiveness of the SR-based method is validated experimentally by applying the acquired vibration signals data to bearings. The experimental results indicate that SR can reduce the computation cost and preserve more structure information about different bearing faults and severities, and it is demonstrated that the proposed feature extraction scheme has an advantage over other similar approaches.