Fault Severity Levels Research Articles

Roller element bearing fault diagnosis using singular spectrum analysis

Most of the existing time series methods of feature extraction involve complex algorithm and the extracted features are affected by sample size and noise. In this paper, a simple time series method for bearing fault feature extraction using singular spectrum analysis (SSA) of the vibration signal is proposed. The method is easy to implement and fault feature is noise immune. SSA is used for the decomposition of the acquired signals into an additive set of principal components. A new approach for the selection of the principal components is also presented. Two methods of feature extraction based on SSA are implemented. In first method, the singular values (SV) of the selected SV number are adopted as the fault features, and in second method, the energy of the principal components corresponding to the selected SV numbers are used as features. An artificial neural network (ANN) is used for fault diagnosis. The algorithms were evaluated using two experimental datasets—one from a motor bearing subjected to different fault severity levels at various loads, with and without noise, and the other with bearing vibration data obtained in the presence of a gearbox. The effect of sample size, fault size and load on the fault feature is studied. The advantages of the proposed method over the exiting time series method are discussed. The experimental results demonstrate that the proposed bearing fault diagnosis method is simple, noise tolerant and efficient.

Sweep frequency response analysis for diagnosis of low level short circuit faults on the windings of power transformers: An experimental study

This contribution is aimed at obtaining diagnosis criteria for detection of low-level short circuit faults throughout sweep frequency response analysis (SFRA) measurements on the transformer windings. Significant advantages would accrue by early detection of low level short circuit faults within the transformer, since if not quickly detected, they usually develop into more serious faults which result in irreversible damage to the transformer and the electrical network, unexpected outages and the consequential costs. A Finite Element Model (FEM) of the tested transformer has been developed to assist in justifying the modifications of the winding frequency response as a result of fault occurrence. Successful operation of the SFRA method in precisely detecting interturn faults along the transformer windings, even down to a few shorted turns on the winding, is proved through a large number of experiments and measurements. Improving the interpretation of the SFRA measurements needs complementary statistical indicators. The usage of correlation coefficient and spectrum deviation for comparison of the frequency responses obtained through SFRA measurements provides quantitative indicators of the fault presence on the transformer windings and also the fault severity level in the shorted turns.

Fault Severity Estimation of Rotating Machinery Based on Residual Signals

Fault severity estimation is an important part of a condition-based maintenance system, which can monitor the performance of an operation machine and enhance its level of safety. In this paper, a novel method based on statistical property and residual signals is developed for estimating the fault severity of rotating machinery. The fast Fourier transformation (FFT) is applied to extract the so-called multifrequency-band energy (MFBE) from the vibration signals of rotating machinery with different fault severity levels in the first stage. Usually these features of the working conditions with different fault sensitivities are different. Therefore a sensitive features-selecting algorithm is defined to construct the feature matrix and calculate the statistic parameter (mean) in the second stage. In the last stage, the residual signals computed by the zero space vector are used to estimate the fault severity. Simulation and experimental results reveal that the proposed method based on statistics and residual signals is effective and feasible for estimating the severity of a rotating machine fault.

A NEW APPROACH TO ANALYSIS OF INDUCTION MOTORS WITH ROTOR FAULTS DURING STARTUP BASED ON THE FINITE ELEMENT METHOD

The increasing popularity of a so-called transient motor current signature analysis requires the fault diagnostics parameters which could not be exposed to other factors irrelevant to the fault to make a precise assessment of the failure severity level. This challenging task needs a precise modeling of faulty motor behavior in various operating conditions at difierent fault severity levels. This paper introduces a new approach to the flnite element analysis of an induction motor with broken rotor bars during startup. The approach is based on the principle of superposition and contributes to examination of the fault rotor backward rotating magnetic fleld and current components produced by such fleld separating them from stator currents. It gives a new sight on the behavior of a faulty motor during startup for the diagnosis purposes. Further analysis of the simulation data by means of the Extended Park's Vector Approach and the continuous wavelet transform and its experimental validation is also presented in the paper.

Application of Signal Processing for Motor Condition Monitoring Based on Filtered-Signals and Eliminated-Signals

This paper proposes new procedures of motor fault detection. The proposed methods are based on filtered-signals and eliminated-signals. Generally, the raw stator phase currents collected from the motors are firstly filtered in order to get rid of measurement noises. If the new signals are called “Filtered-Signals” and the signals eliminated from the raw stator phase currents are called “Eliminated-Signals”. The first proposed procedure is to detect the motor faults by spectrum of PSD slope from the filtered-signals. The second proposed procedure is to detect the motor faults by spectrum of the eliminated-signals. The both methods are tested on 3 different motor conditions: healthy, stator fault, and rotor fault motor at full load condition. The experiments show that the both methods can differentiate conditions clearly and they also can indicate the levels of fault severity. Thus, it can be effective when the both methods are applied simultaneously to analyze the faults

A New Method of Electric Motor Fault Detection Based on Bi-Spectrum of Eliminated Signals

This paper proposes a new method of motor fault detection by applying the eliminated-signal as data sources for motor fault analysis. Bi-spectrum is used as a key method for processing the signal. The expectation is that the eliminated-signal may contain information for fault analysis. The spectrum and bi-spectrum of the signal are applied as signal processing methods to analyze the motor faults. The method is tested on 3 different motor conditions: healthy, stator fault, and rotor fault motor at full load condition. Based on experiments, the method can differentiate conditions clearly. They seem also to be able to measure fault severity levels by observing the change in among harmonic amplitudes.

Electric Motor Fault Analysis Based on Windowed-Zeropadded FFT

This paper proposes a new method for motor fault analysis. Windowed-Zeropadded FFT is applied as a signal processing method. The method is based on both windowing and zero-padding of the signal. The expectation is that the method can provide harmonic amplitude more visible for the purpose of motor fault analysis. The method is tested on 3 different motor conditions: healthy, stator fault, and rotor fault motor at full load condition. The method can provide more visible harmonic amplitudes than other methods, because it can eliminate the leakages and provide smoother plotting. Thus, it can help improve accuracy of motor condition classification and the prediction of the fault severity levels.

Fault Prediction of Induction Motor Based on Time-Frequency Analysis

Because the faults happening in the motor (such as the stator and the rotor faults) can distort the sinusoidal response of the motor RPM and the main frequency, hence the spectrum method has previously been introduced which it relates to both amplitudes and phases among harmonics in a signal. The method popularly applied for fault detection is based on frequency analysis by observing the side band, its harmonics around main frequencies or its other harmonics. Based on the present experiments, the spectrum method by FFT function has ability to distinguish the motor condition. But, the fault severity levels seem to not able to analyze. Hence the time-frequency Analysis (or spectrogram) of the stator phase currents is proposed here. The method is expected to show relation between the phase current signals and the fault levels which make it can detect the faults and also indicate the fault levels. The experiments show that the proposed method can provide good accuracy for fault prediction and fault level quantification. Hence it can conclude that the propose method can be an effective tool for motor fault prediction.

Bearing fault diagnosis using multi-scale entropy and adaptive neuro-fuzzy inference

A bearing fault diagnosis method has been proposed based on multi-scale entropy (MSE) and adaptive neuro-fuzzy inference system (ANFIS), in order to tackle the nonlinearity existing in bearing vibration as well as the uncertainty inherent in the diagnostic information. MSE refers to the calculation of entropies (e.g. appropriate entropy, sample entropy) across a sequence of scales, which takes into account not only the dynamic nonlinearity but also the interaction and coupling effects between mechanical components, thus providing much more information regarding machinery operating condition in comparison with traditional single scale-based entropy. ANFIS can benefit from the decision-making under uncertainty enabled by fuzzy logic as well as from learning and adaptation that neural networks provide. In this study, MSE and ANFIS are employed for feature extraction and fault recognition, respectively. Experiments were conducted on electrical motor bearings with three different fault categories and several levels of fault severity. The experimental results indicate that the proposed approach cannot only reliably discriminate among different fault categories, but identify the level of fault severity. Thus, the proposed approach has possibility for bearing incipient fault diagnosis.

Empirical validation of object-oriented metrics for predicting fault proneness models

Empirical validation of software metrics used to predict software quality attributes is important to ensure their practical relevance in software organizations. The aim of this work is to find the relation of object-oriented (OO) metrics with fault proneness at different severity levels of faults. For this purpose, different prediction models have been developed using regression and machine learning methods. We evaluate and compare the performance of these methods to find which method performs better at different severity levels of faults and empirically validate OO metrics given by Chidamber and Kemerer. The results of the empirical study are based on public domain NASA data set. The performance of the predicted models was evaluated using Receiver Operating Characteristic (ROC) analysis. The results show that the area under the curve (measured from the ROC analysis) of models predicted using high severity faults is low as compared with the area under the curve of the model predicted with respect to medium and low severity faults. However, the number of faults in the classes correctly classified by predicted models with respect to high severity faults is not low. This study also shows that the performance of machine learning methods is better than logistic regression method with respect to all the severities of faults. Based on the results, it is reasonable to claim that models targeted at different severity levels of faults could help for planning and executing testing by focusing resources on fault-prone parts of the design and code that are likely to cause serious failures.

A Novel Strategy for the Fault Detection and Diagnosis of Centrifugal Chiller Systems

This paper presents a new fault detection and diagnosis (FDD) strategy for centrifugal chillers in building air-conditioning systems. The strategy is implemented using the fuzzy modeling and the artificial neural network techniques. Based on the sensitivity analysis of performance indices (PIs) to the chiller faults, the PI residuals between the model estimations using the normal data and the model estimations using the fault data are chosen to be normalized and to form a diagnostic classifier. The chosen PI risiduals are further quantified by the proposed fuzzy models. The fault identification is realized by neurasl network, using the quantitative diagnostic classifier. The use of the quantitative diagnostic classifier in this FDD strategy overcomes the problem faced by classical qualitative fault classifiers when different faults have the same linguistic rule pattern. The strategy implementation is based on the data at normal and faulty operating conditions at a range of load levels and fault severity levels. The strategy is validated using the laboratory chiller data provided by ASHRAE Research Project RP-1043 (Comstock and Braun 1999c).

Use of time domain features for the neural network based fault diagnosis of a machine tool coolant system

A procedure is presented for the fault diagnosis of a machine tool coolant system by the use of artificial neural networks (ANNs). The characteristic features of time domain signals of the system with normal and faulty conditions have been used as inputs to ANNs, consisting of input, hidden and output layers. The input layer consisted of nodes for features extracted from the time domain response of the coolant system with simulated faults and severity levels. The inputs were normalized within the range 0.0 and 1.0. The output layer consisted of nodes indicating the status of the system: normal or defective, with classification of faults and severity levels. Two hidden layers with different numbers of neurons were used. Two schemes were used to develop ANNs for the separate and simultaneous detection of fault type and severity level. The ANNs were trained using a back-propagation algorithm with the normalized set of extracted features for known conditions. The ANNs were tested using the data of simulated test faults of both trained and novel types. The diagnostic system detected the fault type and severity level quite accurately in both schemes. The reduced number of inputs lead to faster training, requiring far fewer iterations compared with the case of using the entire time response curve.