Automatic Fault Diagnosis Research Articles

A Wavelet Based Tool to Assist the Automated Analysis of Waveform Disturbances Records in Power Generators

This paper presents an automated disturbance analysis scheme for electrical power generating units. It proposes a method based on wavelet transform applied to transient (waveform) records of disturbances. The objectives are to detect faults in generators, perform a segmentation of the records and extract information that characterize the faults. The output of the method is a set of representative information of the monitored signals, which can be used in a classification system for automated fault diagnosis. The large volume of data produced by the digital recorders of power system disturbances justifies the search for automated analysis methods that can be employed to assist the utilities engineers. Tests were conducted to determine the best mother wavelet in the segmentation process. The proposed method was applied to five case studies with real disturbance data and the obtained results confirmed its efficiency.

Automatic damage identification of roller bearings and effects of sifting stop criterion of IMFs

Damage identification of roller bearings has been deeply developed to detect faults using vibration-based signal processing. Empirical mode decomposition (EMD) is one of the recent techniques adapted to this purpose; it decomposes a multi-component signal into some elementary Intrinsic Mode Functions (IMFs). Although the EMD has been applied in various applications successfully, there are some drawbacks such as lack of a mathematical base, no robust stopping criterion for sifting process, mode mixing and border effect problem. One of the most relevant drawbacks in fault diagnosis is the sifting stop criterion. Although sifting as many times as possible is needed to decompose the signal, too many sifting steps will reduce the physical meaning of IMFs which are extremely important for fault diagnosis. Thus, a precise criterion is required to identify the appropriate stage to conclude the sifting process. The proposed criteria so far are: Cauchy-type convergence, Mean fluctuations thresholds, Energy difference tracking, Resolution factor, Bandwidths, and Orthogonality criterion. Effects of sifting stop criterion on damage identification performance has not studied thoroughly yet. In this study the influence of different stopping criteria on automatic fault diagnosis is investigated. Vibration signals were acquired using the test rig assembled by Dynamics & Identification Research Group (DIRG) at Department of Mechanical and Aerospace Engineering, Politecnico di Torino. Various operating conditions were considered to obtain reliable results. By extracting feature vectors for each decomposing algorithms, the accuracy of defect detection is examined by labeling the samples whether they are healthy or faulty using support vector machine (SVM).

The Automatic Fault Diagnosis Research of Diesel Engine Load under Speed Based On Fuzzy Theory

The main reasons of diesel engine load under speed can be summed up as valve spring fracture, nozzle coking blocked hole, engine oil pipe fracture, fuel injection delay, fuel injection pump drive roll key. There are six symptoms. Using the symptoms and the reasons establishes fuzzy corresponding relation, establishing the automatic fault diagnosis fuzzy model of diesel engine load under speed, through the symptoms of diesel engine load under speed to judge its fault reasons, which could treat the faults quickly so as to ensure the normal operation of the diesel engine.

Rolling bearing fault diagnosis based on partially ensemble empirical mode decomposition and variable predictive model-based class discrimination

An automatic fault diagnosis method for rolling bearing is proposed in this paper. Partially ensemble empirical mode decomposition (PEEMD) is developed to solve the problem of mode mixing existing in empirical mode decomposition. Compared with the ensemble empirical mode decomposition, PEEMD generates much more accurate intrinsic mode functions (IMFs) and the decomposing results are complete and orthogonal. Therefore, PEEMD is utilized to preprocess the vibration signals of rolling bearing. Moreover, the features in time, frequency domains of IMFs and ones of original data in time–frequency domain are extracted to reflect the change of fault information. To avoid the high dimension of features, Laplacian score for feature selection is utilized to sort the initial features according to their significances. The pattern recognition method, variable predictive model-based class discrimination (VPMCD) is introduced to achieve an automatic fault diagnosis. Finally, the proposed fault diagnosis method for rolling bearing is applied to analyze the experimental data and the result indicates that the proposed method can effectively diagnose the fault categories and severities of rolling bearings.

Thruster fault diagnosis method based on Gaussian particle filter for autonomous underwater vehicles

Autonomous Underwater Vehicles (AUVs) generally work in complex marine environments. Any fault in AUVs may cause significant losses. Thus, system reliability and automatic fault diagnosis are important. To address the actuator failure of AUVs, a fault diagnosis method based on the Gaussian particle filter is proposed in this study. Six free-space motion equation mathematical models are established in accordance with the actuator configuration of AUVs. The value of the control (moment) loss parameter is adopted on the basis of these models to represent underwater vehicle malfunction, and an actuator failure model is established. An improved Gaussian particle filtering algorithm is proposed and is used to estimate the AUV failure model and motion state. Bayes algorithm is employed to perform robot fault detection. The sliding window method is adopted for fault magnitude estimation. The feasibility and validity of the proposed method are verified through simulation experiments and experimental data.

FD4C: Automatic Fault Diagnosis Framework for Web Applications in Cloud Computing

The large-scale dynamic cloud computing environment has raised great challenges for fault diagnosis in Web applications: First, fluctuating workloads cause traditional application models to change over time; second, modeling the behaviors of complex applications usually requires domain knowledge which is difficult to obtain; third, managing large-scale applications manually is impractical for operators. To address these issues, this paper proposes an automatic fault (F) diagnosis (D) framework for (4) Web applications in cloud (C) computing (FD4C). In this paper, we propose an online incremental clustering method to recognize access behavior patterns. We also use correlation analysis to model the correlations between the workloads and application performance/resource utilization metrics in a specific access behavior pattern. FD4C detects faults by discovering the abrupt changes of correlation coefficients with control charts. Then, FD4C identifies the fault-related metrics using a feature selection method. To evaluate our proposal, we inject typical faults into TPC-W benchmark and apply FD4C to diagnose the injected faults. The experimental results show that FD4C can effectively detect the typical faults and accurately locate the metrics related to the faults.

Bearing Fault Diagnosis Using a Novel Classifier Ensemble Based on Lifting Wavelet Packet Transforms and Sample Entropy

In order to improve the fault detection accuracy for rolling bearings, an automated fault diagnosis system is presented based on lifting wavelet packet transform (LWPT), sample entropy (SampEn), and classifier ensemble. Bearing vibration signals are firstly decomposed into different frequency subbands through a three-level LWPT, resulting in a total of 8 frequency-band signals throughout the third layers of the LWPT decomposition tree. The SampEns of all the 8 components are then calculated as feature vectors. Such a feature extraction paradigm is expected to depict complexity, irregularity, and nonstationarity of bearing vibrations. Moreover, a novel classifier ensemble is proposed to alleviate the effect of initial parameters on the performance of member classifiers and to improve classification effectiveness. Experiments were conducted on electric motor bearings considering various set of fault categories and fault severity levels. Experimental results demonstrate the proposed diagnosis system can effectively improve bearing fault recognition accuracy and stability in comparison with diagnosis methods based on a single classifier.

A Novel Characteristic Frequency Bands Extraction Method for Automatic Bearing Fault Diagnosis Based on Hilbert Huang Transform.

Because roller element bearings (REBs) failures cause unexpected machinery breakdowns, their fault diagnosis has attracted considerable research attention. Established fault feature extraction methods focus on statistical characteristics of the vibration signal, which is an approach that loses sight of the continuous waveform features. Considering this weakness, this article proposes a novel feature extraction method for frequency bands, named Window Marginal Spectrum Clustering (WMSC) to select salient features from the marginal spectrum of vibration signals by Hilbert–Huang Transform (HHT). In WMSC, a sliding window is used to divide an entire HHT marginal spectrum (HMS) into window spectrums, following which Rand Index (RI) criterion of clustering method is used to evaluate each window. The windows returning higher RI values are selected to construct characteristic frequency bands (CFBs). Next, a hybrid REBs fault diagnosis is constructed, termed by its elements, HHT-WMSC-SVM (support vector machines). The effectiveness of HHT-WMSC-SVM is validated by running series of experiments on REBs defect datasets from the Bearing Data Center of Case Western Reserve University (CWRU). The said test results evidence three major advantages of the novel method. First, the fault classification accuracy of the HHT-WMSC-SVM model is higher than that of HHT-SVM and ST-SVM, which is a method that combines statistical characteristics with SVM. Second, with Gauss white noise added to the original REBs defect dataset, the HHT-WMSC-SVM model maintains high classification accuracy, while the classification accuracy of ST-SVM and HHT-SVM models are significantly reduced. Third, fault classification accuracy by HHT-WMSC-SVM can exceed 95% under a Pmin range of 500–800 and a m range of 50–300 for REBs defect dataset, adding Gauss white noise at Signal Noise Ratio (SNR) = 5. Experimental results indicate that the proposed WMSC method yields a high REBs fault classification accuracy and a good performance in Gauss white noise reduction.

Unsupervised feature selection using swarm intelligence and consensus clustering for automatic fault detection and diagnosis in Heating Ventilation and Air Conditioning systems

Various sensory and control signals in a Heating Ventilation and Air Conditioning (HVAC) system are closely interrelated which give rise to severe redundancies between original signals. These redundancies may cripple the generalization capability of an automatic fault detection and diagnosis (AFDD) algorithm. This paper proposes an unsupervised feature selection approach and its application to AFDD in a HVAC system. Using Ensemble Rapid Centroid Estimation (ERCE), the important features are automatically selected from original measurements based on the relative entropy between the low- and high-frequency features. The materials used is the experimental HVAC fault data from the ASHRAE-1312-RP datasets containing a total of 49 days of various types of faults and corresponding severity. The features selected using ERCE (Median normalized mutual information (NMI)=0.019) achieved the least redundancies compared to those selected using manual selection (Median NMI=0.0199) Complete Linkage (Median NMI=0.1305), Evidence Accumulation K-means (Median NMI=0.04) and Weighted Evidence Accumulation K-means (Median NMI=0.048). The effectiveness of the feature selection method is further investigated using two well-established time-sequence classification algorithms: (a) Nonlinear Auto-Regressive Neural Network with eXogenous inputs and distributed time delays (NARX-TDNN); and (b) Hidden Markov Models (HMM); where weighted average sensitivity and specificity of: (a) higher than 99% and 96% for NARX-TDNN; and (b) higher than 98% and 86% for HMM is observed. The proposed feature selection algorithm could potentially be applied to other model-based systems to improve the fault detection performance.

Knowledge extraction using data mining for multi-class fault diagnosis of induction motor

The feature extraction capability of rough set theory (RST) and genetic algorithm (GA) are used to extract knowledge from radial frame vibration signal for fault diagnosis of induction motor. This knowledge can assist in selecting scales for continuous wavelet transform (CWT) and mono-components required for Hilbert transform (HT) to extract fault related information from the vibration signal. Thus, the computational complexity of the signal processing tools is considerably reduced making both CWT and HT hardware friendly and suitable for real-time applications. For machine learning based automatic multi-class fault diagnosis, the performance of the classifiers are also considerably improved with significant reduction in computational burden since the redundant and irrelevant information can be effectively removed. The information obtained using data mining technique is successfully used to detect six types of induction motor faults. The results obtained are also verified in presence of high level of noise which has not been attempted earlier. The main contribution of the paper is to combine the advantages of two powerful signal processing tool like CWT and HT to extract hidden information from vibration signal in conjunction with data mining technique making them computationally efficient and easy to implement.

Online fault diagnosis for nonlinear power systems

This paper considers the problem of automatic fault diagnosis for transmission lines in large scale power networks. Since faults in transmission lines threatens stability of the entire power network, fast and reliable fault diagnosis is an important problem in transmission line protection. This work is the first paper exploiting sparse signal recovery for the fault-diagnosis problem in power networks with nonlinear swing-type dynamics. It presents a novel and scalable technique to detect, isolate and identify transmission faults using a relatively small number of observations by exploiting the sparse nature of the faults. Buses in power networks are typically described by second-order nonlinear swing equations. Based on this description, the problem of fault diagnosis for transmission lines is formulated as a compressive sensing or sparse signal recovery problem, which is then solved using a sparse Bayesian formulation. An iterative reweighted ℓ1-minimisation algorithm based on the sparse Bayesian learning update is then derived to solve the fault diagnosis problem efficiently. With the proposed framework, a real-time fault monitoring scheme can be built using only measurements of phase angles at the buses.

Two-stage feature selection for bearing fault diagnosis based on dual-tree complex wavelet transform and empirical mode decomposition

This paper presents an automatic fault diagnosis of different rolling element bearing faults using a dual-tree complex wavelet transform, empirical mode decomposition, and a novel two-stage feature selection technique. In this method, dual-tree complex wavelet transform and empirical mode decomposition were used to preprocess the original vibration signal to obtain more accurate fault characteristic information. Then, features in the time domain were extracted from each of the original signals, the coefficients of the dual-tree complex wavelet transform, and some useful intrinsic mode functions to generate a rich combined feature set. Next, a two-stage feature selection algorithm was proposed to generate the smallest set of features that leads to the superior classification accuracy. In the first stage of the two-stage feature selection, we found the candidate feature set using the distance evaluation technique and a k-nearest neighbor classifier. In the second stage, a genetic algorithm-based k-nearest neighbor classifier was designed to obtain the superior combination of features from the candidate feature set with respect to the classification accuracy and number of feature inputs. Finally, the selected features were used as the input to a k-nearest neighbor classifier to evaluate the system diagnosis performance. The experimental results obtained from real bearing vibration signals demonstrated that the method combining dual-tree complex wavelet transform, empirical mode decomposition, and the two-stage feature selection technique is effective in both feature extraction and feature selection, which also increase classification accuracy.

Automated Fault Diagnosis and Accommodation Control for Mechanical Systems

In this paper, a fault diagnosis and accommodation control scheme is designed, which is applicable to mechanical systems. First, the fault detection is developed by designing a nonlinear observer. Second, the fault isolation is achieved by using multiple fault observers. Third, when a fault is diagnosed, an adequate fault accommodation control is triggered to compensate the effects of the fault function. Finally, case study is given to illustrate the effects of the proposed method.

Design of the Architecture of Power Grid’s Device Monitoring System

Under the Large Operation mode of State Grid of china, the concentrated monitoring work on power devices is facing some serious problems and challenges. They are, monitoring information is not well structured, the amount of alarm information is quite huge and automatic fault diagnosis function is not smart enough. To handle these problems, this paper carried out comprehensive research on device monitoring oriented application system architecture, and proposed an overall solution on solving device monitoring problems from both substation and main stations. The realization of key modules of this architecture could greatly optimize the device monitoring oriented application system, and guide the incoming construction work on power dispatching and monitoring integration system.

Model-based fault localization in bottling plants

The bottling of beverages is carried out in complex plants that consist of several machines and material flows. To realize an efficient bottling process and high quality products, operators try to avoid plant downtimes. With actual non-productive times of between 10% and 60%, the operators require diagnosis tools that allow them to locate plant components that cause downtime by exploiting automatically acquired machine data.This paper presents a model-based solution for automatic fault diagnosis in bottling plants. There are currently only a few plant-specific solutions (based on statistical calculations or artificial neural networks) for automatic bottling plant diagnosis. In order to develop a customizable solution, we followed the model-based diagnosis approach which allows the automatic generation of diagnosis solutions for individual plants. The existing stochastic and discrete-event models for bottling plants are not adequate for model-based diagnosis. Therefore, we developed new first-principle models for the relevant plant components, validated them numerically, and abstracted them to qualitative diagnosis models. Based on the diagnosis engine OCC’M Raz’r, application systems for two real plants and one virtual plant (based on discrete-event simulation) were generated and evaluated. Compared to the reasons for downtime identified by experts, we obtained up to 87.1% of compliant diagnosis results. The diagnosis solution was tested by practitioners and judged as a useful tool for plant optimization.

Qualitative Event-Based Fault Isolation under Uncertain Observations


 
 
 For many systems, automatic fault diagnosis is critical to ensuring safe and efficient operation. Fault isolation is performed by analyzing measured signals from the system, and reasoning over the system behavior to determine which faults have occurred, based on models of predicted faulty behavior. For dynamic systems, reasoning may be performed using qualitative analysis of the differences between measured signals and their predicted values, in which observations take the form of qualitative symbols. Such an approach is quick to isolate faults, but depends critically on correct generation of the qualitative symbols from the signals. In this paper, we develop an approach to qualitative event-based fault isolation for dynamic systems that is robust to incorrect qualitative observations. Observations are treated as uncertain, where multiple interpretations of an observation, each with its own probability, are considered. By interpreting observed symbols in a probabilistic manner, the approach degrades gracefully as the number of incorrectly-generated symbols increases. The approach is demonstrated on an electrical power system testbed, and experiments using real data obtained from the hardware demonstrate the improved fault isolation performance in the presence of incorrect symbol generation.
 
 

Kuiper test and autoregressive model-based approach for wireless sensor network fault diagnosis

Wireless sensor networks (WSNs) have recently received increasing attention in the areas of defense and civil applications of sensor networks. Automatic WSN fault detection and diagnosis is essential to assure system's reliability. Proactive WSNs fault diagnosis approaches use embedded functions scanning sensor node periodically for monitoring the health condition of WSNs. But this approach may speed up the depletion of limited energy in each sensor node. Thus, there is an increasing interest in using passive diagnosis approach. In this paper, WSN anomaly detection model based on autoregressive (AR) model and Kuiper test-based passive diagnosis is proposed. First, AR model with optimal order is developed based on the normal working condition of WSNs using Akaike information criterion. The AR model then acts as a filter to process the future incoming signal from different unknown conditions. A health indicator based on Kuiper test, which is used to test the similarity between the training error of normal condition and residual of test conditions, is derived for indicating the health conditions of WSN. In this study, synthetic WSNs data under different cases/conditions were generated and used for validating the approach. Experimental results show that the proposed approach could differentiate WSNs normal conditions from faulty conditions. At last, the overall results presented in this paper demonstrate that our approach is effective for performing WSNs anomalies detection.

Monte Carlo analysis of the effect of uncertainties on model-based HVAC fault detection and diagnostics

Faults in HVAC systems can have a significant negative impact on energy consumption, indoor thermal comfort, and air quality. Automatic fault detection and diagnosis tools can help commissioning providers, operators, and facility managers efficiently detect and diagnose faults. They also can help satisfy the increasing demand for commissioning. A model-based fault detection and diagnosis (FDD) method was developed to detect faults by comparing model prediction and measurement, and to diagnose faults using a rule-based fuzzy inferencing system. The method includes Monte Carlo analysis to improve the robustness of the fault detection and diagnosis and reduce false alarms. The Monte Carlo analysis is employed not only to predict uncertainties in reference model outputs, based on estimates of uncertainty in each of the measured inputs, but also to determine the confidence levels of fault diagnosis by combining the effects of input uncertainties at different operating points. A simulated variable-air-volume (VAV) system, including detailed component models that can simulate different faults as well as correct operation, was used to test the diagnostic rules and the Monte Carlo analysis included in the method. The effect of uncertainties on fault diagnosis is illustrated for various types of faulty operation.

Application of High Frequency Acceleration Envelope Power Spectrum and ANN for Fault Diagnosis on Journal Bearing Under Various Loading Conditions

Journal bearings are widely used to support the shaft of industrial machinery with heavy loads, such as compressors, turbines and centrifugal pumps. The major problem in journal bearing is catastrophic failure due to corrosion and erosion, results in economic loss and creates high safety risks. So, it is necessary to provide condition monitoring technique to detect and diagnose failures, to achieve cost benefits to industry. High frequency acceleration enveloping facilitates the extraction of low amplitude, high frequency signals associated with repetitive impacts in journal bearings, providing a key tool for early detection in the onset of bearing damage and similar machinery health problems when coupled with standard FFT analysis. The DEWESOFT software-based methods for implementing and interpreting high frequency acceleration enveloping are presented and compared. In this paper the application of STFT (Short Time Fourier Transform) and Autocorrelation through FFT are used for processing vibration signal to detect faults in journal bearing is presented. A bearing testing apparatus is used for experimental studies to obtain vibration signal from a healthy bearing and fault bearing. Further application of ANN for extracting features like RMS, mean, kurtosis, shape factor, crest factor and impulse factor is investigated as a model for automated fault diagnosis is presented.

A Modeling and Probabilistic Reasoning Method of Dynamic Uncertain Causality Graph for Industrial Fault Diagnosis

Online automatic fault diagnosis in industrial systems is essential for guaranteeing safe, reliable and efficient operations. However, difficulties associated with computational overload, ubiquitous uncertainties and insufficient fault samples hamper the engineering application of intelligent fault diagnosis technology. Geared towards the settlement of these problems, this paper introduces the method of dynamic uncertain causality graph, which is a new attempt to model complex behaviors of real-world systems under uncertainties. The visual representation to causality pathways and self-relied “chaining” inference mechanisms are analyzed. In particular, some solutions are investigated for the diagnostic reasoning algorithm to aim at reducing its computational complexity and improving the robustness to potential losses and imprecisions in observations. To evaluate the effectiveness and performance of this method, experiments are conducted using both synthetic calculation cases and generator faults of a nuclear power plant. The results manifest the high diagnostic accuracy and efficiency, suggesting its practical significance in large-scale industrial applications.