Automatic Fault Diagnosis Research Articles

A Multi-Dimension End-to-End CNN Model for Rotating Devices Fault Diagnosis on High-Speed Train Bogie

With the improvement of sensor techniques, and the urgent requirement of automatic fault diagnosis technologies, the intelligent perception system on high speed train is more popular than ever before. It records the devices’ state information through a sensor network, and services for further analysis. However, Traditional machine learning algorithms are usually constrained by massive multi-sensor data and knowledge-based feature extraction in fault diagnosis. Therefore, this paper extended fault diagnosis methodology into tensor space to deal with multi-sensor monitoring data and take full use of available information. Moreover, the convolutional neural network (CNN) is used for automatic feature learning and classification without human intervention. The effectiveness and efficiency are validated by dataset of rolling element bearings obtained in lab and real-use case. Three features can be highlighted. First of all, the proposed model showed a good adaptability and high efficiency under various working condition by taking full use of the multi-sensor data. It has powerful ability in accuracy and convergence speed. Secondly, it is not as sensitive to data quantity as other deep learning algorithms do. Such superior characteristic made the model more suitable for practical application, because of the insufficient failure data. At last, it is an intelligent End-to-End model, performing automatic fault diagnosis without manual intervention and suitable for real-use case.

Building Management System

A Building Management System is also known as Building Automation System (BAS), It is a Computer based control system installed in buildings that controls and monitors buildings mechanical and electrical equipment such as ventilation, lighting, power systems and security systems. A building management system for METRO includes inputs from the systems like HVAC (Heating Ventilation Air Conditioning), Lifts Escalators, UPS and Fire Alarm and Detection System etc. A Building Management system uses a PLC controller along with SCADA suite to monitor and control all the systems included in the Metro. BMS provides Automatic Fault Detection System and diagnosis strategies for building energy performance.

Event-based diagnosis of flight maneuvers of a fixed-wing aircraft

Despite all efforts to improve the reliability of aircraft, the possibility of failure occurrences cannot be completely eliminated, and hence, airplane safety forever remains as a post-developmental concern. In times of peril, flight crews have a minimal amount of time to detect and determine an occurred fault. An automated fault diagnosis tool can help timely detect occurred faults and increase the flight crew’s odds of recovering the plane before a possible crash. As a solution, this paper proposes a diagnosis tool that provides diagnostic information based upon the pilot’s corrective inputs to level an aircraft’s wings. For this purpose, this paper models a fixed-wing aircraft as a discrete event system (DES). The constructed model is then used to develop a diagnosis tool, a so-called diagnoser, for detecting, isolating, and identifying fault occurrences in the ailerons (left and right) of the aircraft. The developed diagnoser can be asynchronously activated during the aircraft’s operation and provides diagnostic information in real time.

Robust Interpretable Deep Learning for Intelligent Fault Diagnosis of Induction Motors

In modern manufacturing processes, motivations for automatic fault diagnosis (FD) are increasingly growing as a result of the great trends toward achieving zero breakdowns. Induction motors (IMs) represent a critical part in most of the applications. Due to its high potential of automatic feature extraction, the deep learning (DL)-based FD of IM has recently been introduced and has essentially emphasized on the diagnosis using the vibration analysis. However, this approach has not received considerable attention when using the current analysis, although it represents a cost-effective alternative. Moreover, the already implemented DL architectures are still suffering from lack of physical interpretability. In this article, a new DL architecture called deep-SincNet is implemented for a multi-FD task. The proposed end-to-end scheme automatically learns the fault features from the raw motor current and accordingly finalizes the FD process. A high accuracy for several separated and combined faults, a more physical interpretability, a high robustness against noisy environments, and a significant gain in implementation cost prove the competitive performance of the proposed approach.

Fault Detection and Diagnosis Using Self-Attentive Convolutional Neural Networks for Variable-Length Sensor Data in Semiconductor Manufacturing

Nowadays, more attention has been placed on cost reductions and yield enhancement in the semiconductor industry. During the manufacturing process, a considerable amount of sensor data called status variables identification (SVID) is collected by sensors embedded in advanced machines. This data is a valuable source for data-driven automatic fault detection and diagnosis at an early manufacturing stage to maintain competitive advantages. However, wafer processing times vary slightly from wafer to wafer, resulting in variable-length signal data. The conventional approaches use much condensed data called fault detection and classification (FDC) data made by manually designed feature extraction. Or, recent deep learning approaches assume that all wafers have the same processing time, which is impotent to the variable-length SVID. To detect and diagnose faults directly from the variable-length SVID, we propose a self-attentive convolutional neural network. In experiments using real-world data from a semiconductor manufacturer, the proposed model outperformed other deep learning models with less training time and showed robustness at different sequence lengths. Compared to FDC data, SVID data showed better fault detection performance. Without manually investigating the lengthy sensor signals, abnormal sensor value patterns were found at the time specified by the model.

Deep residual network based fault detection and diagnosis of photovoltaic arrays using current-voltage curves and ambient conditions

Automatic fault detection and diagnosis techniques for photovoltaic arrays are crucial to promote the efficiency, reliability and safety of photovoltaic systems. In recent decades, many conventional artificial intelligence approaches have been successfully applied to automatically establish fault detection and diagnosis model using fault data samples, but most of them rely on manual feature extraction or expert knowledge to build diagnosis models, which is inefficient and may ignore some potential useful features. In addition, they usually use shallow neural networks with limited performance. Addressing the issues, this paper proposes a novel intelligent fault detection and diagnosis method for photovoltaic arrays based on a newly designed deep residual network model trained by the adaptive moment estimation deep learning algorithm, which can automatically extract features from raw current-voltage curves and ambient irradiance and temperature, and effectively improve the performance with a deeper network. In order to validate the proposed fault diagnosis model, a Simulink based simulation model is designed for a real laboratory photovoltaic array, and both fault simulation and real experiments are carried out to obtain simulation and experimental fault datasets. Furthermore, two other popular deep learning based models are used for comparison, including convolution neural network and convolutional auto-encoder. Both of simulation and real experimental comparison results demonstrate that the proposed deep residual network based method achieves high and best overall performance in terms of accuracy, generalization performance, reliability and training efficiency.

A complete diagnosis of faulty sensor modules in a wireless sensor network

Wireless sensor network (WSN) is a collection of sensor devices, which collects information from the environment and sends to the base station by single-hop or multi-hop communication towards achieving some predefined objectives. Due to the multi-functional applications in the adventurer environment, the sensor modules are damaged by various external and internal sources, which leads to a failure sensor network. Therefore, automated fault diagnosis and fault tolerance are of utmost importance in WSN. In this article, a complete fault diagnosis methodology is proposed for faulty sensors in the failed WSN. The proposed methodology consists of four phases of functioning such as initialization, fault detection, fault classification, and fault tolerance phase. The hard fault and link fault are detected by checksum and Fletcher’s checksum method. The soft faults (permanent, intermittent, and transient) are detected by Mann-Whitney U statistical test. The Gaussian transformation function with a classification factor is used for soft faults classification. The fault tolerance is attained in the sensor system by the stepwise regressional method. The proposed methodology is implemented in a real testbed experiment with simulation setup. The parameters like fault detection accuracy, false alarm rate, false positive rate, fault classification rate, false classification rate, R-squared value, standard error, fault probability, and degree of sensors are considered for the evaluation of the proposed diagnosis methodology.

Novel fault diagnosis scheme utilizing deep learning networks

Information availability and processing advances have allowed automated fault diagnosis methods to gain popularity in nuclear industry. However, stringent safety requirements coupled with nuclear industry requires that such methods avoid any kind of misdiagnosis. In manuscript, a novel fault diagnosis model is being proposed which can not only diagnose faults in an early time frame but also has the ability to identify fault scenarios that cannot be confidently classified. Moreover, for each type of fault identified, severity of the fault can also be calculated. The proposed method utilizes Principle Component Analysis & Deep Learning networks of Long Short Term Memory and Convolution Neural Network. The study was conducted on simulation model of IP-200 nuclear power plant using RELAP5/MOD 4.0 thermal-hydraulic code. Proposed fault diagnosis model was verified using different operational states along-with 4 different fault types and 1 blind case. Simulation data suggests that proposed model is a feasible scheme for fault diagnosis of nuclear power plant.

Fault Isolability Analysis and Optimal Sensor Placement for Fault Diagnosis in Smart Buildings

Faults and anomalies in buildings are among the main causes of building energy waste and occupant discomfort. An effective automatic fault detection and diagnosis (FDD) process in buildings can therefore save a significant amount of energy and improve the comfort level. Fault diagnosability analysis and an optimal FDD-oriented sensor placement are prerequisites for effective, efficient and successful diagnostics. This paper addresses the problem of fault diagnosability for smart buildings. The method used in the paper is a model-based technique which uses Dulmage-Mendelsohn decomposition. To the best of our knowledge, this is the first time that this method is used for applications in smart buildings. First a dynamic model for a zone in a real-case building is developed in which faults are also introduced. Then fault diagnosability is investigated by analyzing the fault isolability of the model. Based on the investigation, it was concluded that not all the faults in the model are diagnosable. Then an approach for placing new sensors is implemented. It is observed that for two test scenarios, placing additional sensors in the model leads to full diagnosability. Since sensors placement is key for an effective FDD process, the optimal placement of such sensors is also studied in this work. A case study of campus building OU44 at the University of Southern Denmark is considered. The results show that as the system gets more complicated by introducing more faults, additional sensors should be added to achieve full diagnosability.

Energy Efficiency Solutions for Buildings: Automated Fault Diagnosis of Air Handling Units Using Generative Adversarial Networks

Automated fault diagnosis (AFD) for various energy consumption components is one of the main topics for energy efficiency solutions. However, the lack of faulty samples in the training process remains as a difficulty for data-driven AFD of heating, ventilation and air conditioning (HVAC) subsystems, such as air handling units (AHU). Existing works show that semi-supervised learning theories can effectively alleviate the issue by iteratively inserting newly tested faulty data samples into the training pool when the same fault happens again. However, a research gap exists between theoretical AFD algorithms and real-world applications. First, for real-world AFD applications, it is hard to predict the time when the same fault happens again. Second, the training set is required to be pre-defined and fixed before being packed into the building management system (BMS) for automatic HVAC fault diagnosis. The semi-supervised learning process of iteratively absorbing testing data into the training pool can be irrelevant for industrial usage of the AFD methods. Generative adversarial network (GAN) is well-known as an unsupervised learning technique to enrich the training pool with fake samples that are close to real faulty samples. In this study, a hybrid generative adversarial network (GAN) is proposed combining Wasserstein GAN with traditional classifiers to perform fault diagnosis mimicking the real-world scenarios with limited faulty training samples in the training process. Experimental results on real-world datasets demonstrate the effectiveness of the proposed approach for fault diagnosis problems of AHU subsystem.

Texture-and-Shape Based Active Contour Model for Insulator Segmentation

Insulator segmentation is a critical step for automatic insulator fault diagnosis in high voltage transmission systems. Existing methods fail to segment insulators when they have a low contrast with the surroundings. Considering the unique shape and texture characteristics of insulators, a texture-and-shape based active contour model is proposed for insulator segmentation. The segmentation is achieved by evolving a curve iteratively by the texture features and shape priors. In the texture-driven curve evolution, a semi-local region descriptor is used to extract the texture features of insulators and a new convex energy functional is defined based on the extracted features with the topology-preserving term. The topology-preserving term keeps the curve's topology unchanged as the curve topology is determined by the shape template. In the shape-driven curve evolution, the shape context descriptor is used to align the shape template with the current curve. The semantic transformation between the shape template and the current curve is obtained by Procrustes analysis and then adopted to update the current curve to resemble the shape prior. The proposed method is applied to a set of images, and the experimental results confirm the efficacy and effectiveness of the proposed method for segmenting insulators in cluttered backgrounds.

Support vector machine based fault detection and diagnosis for HVAC systems

Various faults occurred in the heating, ventilation and airconditioning (HVAC) systems usually lead to more energy consumption and worse thermal comfort inevitably. This paper presents a feasible and valid solution of HVAC fault detection and diagnosis (FDD) problem based on statistical machine learning technology. It learns the consistent nature of different types of faults of HVAC operation based on support vector machine (SVM), and then identify types of fault in all subsystems using the statistical relationships between groups of measurements. In order to speed up the learning process, principle component analysis (PCA) has been applied to compress the training data. Our approach models the dynamical sub-systems and sequence data in HVAC system. The learnt models can then be used for automatic fault detection and diagnosis. The approach has been tested on commercial HVAC systems. It had successfully detected and identified a number of typical AHU faults.

Support vector machine based fault detection and diagnosis for HVAC systems

Various faults occurred in the heating, ventilation and airconditioning (HVAC) systems usually lead to more energy consumption and worse thermal comfort inevitably. This paper presents a feasible and valid solution of HVAC fault detection and diagnosis (FDD) problem based on statistical machine learning technology. It learns the consistent nature of different types of faults of HVAC operation based on support vector machine (SVM), and then identify types of fault in all subsystems using the statistical relationships between groups of measurements. In order to speed up the learning process, principle component analysis (PCA) has been applied to compress the training data. Our approach models the dynamical sub-systems and sequence data in HVAC system. The learnt models can then be used for automatic fault detection and diagnosis. The approach has been tested on commercial HVAC systems. It had successfully detected and identified a number of typical AHU faults.

A Rotating Machinery Fault Diagnosis Method Based on Feature Learning of Thermal Images

The rotating machinery plays a vital role in industrial systems, in which unexpected mechanical faults during operation can lead to severe consequences. For fault prevention, many fault diagnostic methods based on vibration signals are available in the literature. However, the vibration signals are obtained by using different types of sensors, which can cause sensor installation issues and damage the rotating machinery. In addition, this kind of data acquisition through vibration signal induces a large amount of signal noise during machine operation, which will challenge the later fault diagnosis. A recent fault detection method based on infrared thermography (IRT) for rotating machinery avoids these issues. However, the corresponding literature is limited by the fact that the characteristics of the manual design cannot characterize the fault completely so that the diagnostic accuracy cannot exceed the diagnostic method based on the vibration signals. This paper introduces a popular image feature extraction method into the fault diagnosis of rotating machinery based on IRT for the first time. First, capturing the IRT images of the rotating machinery in different states, and then two popular feature extraction methods for IRT images, bag-of-visual-word, and convolutional neural network, are tested in turn. Finally, the extracted features are classified to implement the automatic fault diagnosis. The developed method is applied to analyze the experimental IRT images collected from bearings, and the results demonstrate that the developed method is more effective than the traditional methods based on vibration signals.

Automatic Diagnosis of Rolling Element Bearing Under Different Conditions Based on RVMD and Envelope Order Capture

Fault characteristic frequency is the main basis for rolling element bearing diagnostics but finding a suitable frequency band for demodulation and searching for the fault characteristic frequencies consume a lot of time and manpower in practice. A data-driven method based on recursive variational mode decomposition (RVMD), and an envelope order capture is proposed to realize the automatic fault diagnosis of bearing under different operating conditions. The process starts with a new proposed RVMD of the vibration signal, where the mode with maximum kurtosis of the unbiased autocorrelation of the envelope is selected to get envelope order spectrum. Thereafter, an order capture algorithm is designed to automatically search for the fault characteristic orders in theory, which are used for constructing feature vectors for diagnosis. The proposed method is tested on two test-beds which both contain the same type of bearing (SKF6205) but operate in different conditions, and gets good performance in bearing diagnosis. In addition, the fault diagnosis of test-bed two using training samples that are from test-bed one is investigated. This method reveals well generalization capability in the fault diagnosis of the same type of rolling element bearing under different operating conditions.

Fault severity recognition based on self-adaptive particle swarm optimisation using wavelet kernel extreme learning machine

Gear fault levels are an important issue and can show the operating status of gears. In this paper, the extreme learning machine (ELM) learning method is regarded as a basic algorithm for classifying gear fault levels. In order to improve the generalisation of the ELM and obtain a better classification performance, the wavelet kernel concept was introduced to the ELM and thereby a structure called a wavelet kernel extreme learning machine (WK-ELM) was constructed. In terms of the WK-ELM, because the dilation factor of the wavelet kernel and the number of hidden neurons of the ELM are important for the classification performance, particle swarm optimisation (PSO) was used to search for the most appropriate values of the solved problem. What is more, PSO was also optimised by a self-adaptive inertia weight that formed a novel structure, that is, self-adaptive inertia weight particle swarm optimisation (SAPSO). So far, an automatic fault diagnosis system for gear fault levels based on the SAPSO-WK-ELM has been proposed in this paper. Aimed at recognising different gear fault levels at different rotational speeds, ranking mutual information (RMI) and standardised mutual information (SMI) techniques were applied to select sensitive features for gear fault levels. Compared with other relational algorithms previously used, the results of this study, including simulation and experimental data, have demonstrated that the SAPSO-WK-ELM can achieve an outstanding classification performance.

Association Rule‐Based Feature Mining for Automated Fault Diagnosis of Rolling Bearing

Effective and efficient diagnosis methods are highly demanded to improve system reliability. Comparing with conventional fault diagnosis methods taking a forward approach (e.g., feature extraction, feature selection, and fusion, and then fault diagnosis), this paper presents a new association rule mining method which provides an inverse approach unearthing the underlying relation between labeled defects and extracted features for bearing fault analysis. Instead of evenly dividing methods used in traditional association rule mining, a new association rule mining approach based on the equal probability discretization method is presented in this study. First, a series of extracted features of signal data are discretized following the guideline of equalized probability distribution of the data in order to avoid excessive concentration or decentralized data. Next, the data matrix composed of arrays of discretized features and defect labels is exploited to generate the association rules representing the relation between the features and fault types. Experimental study on a bearing test reveals that the proposed method can generate a series of underlying association rules for bearing fault diagnosis, and the related features selected by the proposed method can be used directly to analyze bearing signals for fault classification and defect severity identification. As a new feature selection method, it possesses prominent superiority compared to traditional PCA, KPCA, and LLE dimension reduction methods.

Automatic Fault Diagnostic System for Induction Motors under Transient Regime Optimized with Expert Systems

Induction machines (IMs) power most modern industrial processes (induction motors) and generate an increasing portion of our electricity (doubly fed induction generators). A continuous monitoring of the machine’s condition can identify faults at an early stage, and it can avoid costly, unexpected shutdowns of production processes, with economic losses well beyond the cost of the machine itself. Machine current signature analysis (MCSA), has become a prominent technique for condition-based maintenance, because, in its basic approach, it is non-invasive, requires just a current sensor, and can process the current signal using a standard fast Fourier transform (FFT). Nevertheless, the industrial application of MCSA requires well-trained maintenance personnel, able to interpret the current spectra and to avoid false diagnostics that can appear due to electrical noise in harsh industrial environments. This task faces increasing difficulties, especially when dealing with machines that work under non-stationary conditions, such as wind generators under variable wind regime, or motors fed from variable speed drives. In these cases, the resulting spectra are no longer simple one-dimensional plots in the time domain; instead, they become two-dimensional images in the joint time-frequency domain, requiring highly specialized personnel to evaluate the machine condition. To alleviate these problems, supporting the maintenance staff in their decision process, and simplifying the correct use of fault diagnosis systems, expert systems based on neural networks have been proposed for automatic fault diagnosis. However, all these systems, up to the best knowledge of the authors, operate under steady-state conditions, and are not applicable in a transient regime. To solve this problem, this paper presents an automatic system for generating optimized expert diagnostic systems for fault detection when the machine works under transient conditions. The proposed method is first theoretically introduced, and then it is applied to the experimental diagnosis of broken bars in a commercial cage induction motor.

An efficient feature extraction approach based on manifold learning for analogue circuits fault diagnosis

Automated fault diagnosis of analogue circuits makes use of dimension reduction (DR) technologies to extract fault features. Traditional DR methods such as the principal component analysis (PCA) and kernel based DR methods have achieved valuable results. However, these methods give inadequate consideration to the geometric structure of fault data which is embedded in high dimensional spaces (HDS). This negatively affects the low-dimensional features. Recently introduced, manifold learning methods have the ability to recover the intrinsic geometric structure of a data set in low dimensional spaces (LDS). Locally linear embedding (LLE) and diffusion maps (DM) are two popular approaches of manifold learning methods. The former preserves the local linear structure of a data set and the latter is resistant to outlier effects. We present a novel DR algorithm using both LLE and DM to extract fault features based on the analysis of the intrinsic geometric structure of fault data. First, by the correlation analysis of the original fault data, we conclude that they are locally linear dependent, and then use LLE to obtain mid-level fault features. The calculation of local outlier factor (LOF) values confirms the existence of outliers in the fault data with mid-level features. Then, we apply DM to eliminate the negative effect of outliers and derive visual data with high-level features. Our experimental results show that the proposed algorithm offers performance superior to the traditional DR methods regardless of whether the circuit is linear or non-linear.

Multi-fault diagnosis of rolling bearing using fuzzy entropy of empirical mode decomposition, principal component analysis, and SOM neural network

The condition monitoring and multi-fault diagnosis of rolling bearing is a very important research content in the field of the rotating machinery health management. Most researches widely used empirical mode decomposition in tandem with principal component analysis which is applied for feature extraction. But this method may lead to imprecise classification. In this paper, we propose a new method of rolling bearing multi-fault diagnosis, by combining the fuzzy entropy of empirical mode decomposition, principal component analysis, and self-organizing map neural network. The empirical mode decomposition process allows the vibration signal to be decomposed into a series of intrinsic mode functions. For each intrinsic mode function, we obtained the fault feature information. The proposed approach combines the fuzzy function and sample entropy to obtain fuzzy entropy. By this combination, we can reflect the complexity and the irregularity in each intrinsic mode function component. The fuzzy entropy of empirical mode decomposition used to construct the vectors is defined as the input of the principal component analysis. This principal component analysis is used to reduce the dimension of the feature vectors. Finally, the reduced feature vectors are chosen as input of self-organizing map network for automatic fault diagnosis and fault classification. The obtained results show that the proposed approach makes it possible to correctly assess the degradation of rolling bearing and to obtain recognition of high-sensitivity defects for different types of bearing faults.