Fault Symptoms Research Articles

Log-Based Fault Localization with Unsupervised Log Segmentation

Localizing faults in a software is a tedious process. The manual approach is becoming impractical because of the large size and complexity of contemporary computer systems as well as their logs, which are often the primary source of information about the fault. Log-based Fault Localization (LBFL) is a popular method applied for this purpose. However, in real-world scenarios, this method is vulnerable to a large number of previously unseen log lines. In this paper, we propose a novel method that can guide programmers to the location of a fault by creating a hierarchy of log lines with the highest rank, selected by the traditional LBFL method. We use the intuition that the symptoms of faults are in the context of normal behavior, whereas suspicious log lines grouped together are from new or additional functionalities turned on during faulty execution. To obtain this context, we used unsupervised log sequence segmentation, which has been previously used to segment log sequences into meaningful segments. Experiments on real-life examples show that our method reduces the effort to find the most crucial logs by up to 64% compared with the traditional timestamp approach. We demonstrate that context is highly useful in advancing fault localization, showing the possibility of further speeding up the process.

Association Model-Based Intermittent Connection Fault Diagnosis for Controller Area Networks

Controller Area Networks (CANs) play an important role in many safety-critical industrial systems, which places high demands on their reliability performance. However, the intermittent connection (IC) of network cables, a random and transient connectivity problem, is a common but hard troubleshooting fault that can cause network performance degradation, system-level failures, and even safety issues. Therefore, to ensure the reliability of CANs, a fault symptom association model-based IC fault diagnosis method is proposed. Firstly, the symptoms are defined by examining the error records, and the domains of the symptoms are derived to represent the causal relationship between the fault locations and the symptoms. Secondly, the fault probability for each location is calculated by minimizing the difference between the symptom probabilities calculated from the count information and those fitted by the total probability formula. Then, the fault symptom association model is designed to synthesize the causal and the probabilistic diagnostic information. Finally, a model-based maximal contribution diagnosis algorithm is developed to locate the IC faults. Experimental results of three case studies show that the proposed method can accurately and efficiently identify various IC fault location scenarios in networks.

Beyond seen faults: Zero-shot diagnosis of power circuit breakers using symptom description transfer

Power circuit breakers (CBs) are vital for the control and protection of power systems, yet diagnosing their faults accurately remains a challenge due to the diversity of fault types and the complexity of their structures. Traditional data-driven methods, although effective, require extensive labeled data for each fault class, limiting their applicability in real-world scenarios where many faults are unseen. This paper addresses these limitations by introducing symptom description transfer-based zero-shot fault diagnosis (SDT-ZSFD), a method that leverages zero-shot learning for fault diagnosis. Our approach constructs a fault symptom description (FSD) framework, which embeds a fault symptom layer between the feature layer and the label layer to facilitate knowledge transfer from seen to unseen fault classes. The method utilizes current and acceleration signals collected during CB operation to extract features. By applying sparse principal component analysis to these signals, we derive high-quality features that are mapped to the FSD framework, enabling effective zero-shot learning. Our method achieves a satisfactory recognition rate by accurately diagnosing unseen faults based on these symptoms. This approach not only overcomes the data scarcity problem but also holds potential for practical applications in power system maintenance. The SDT-ZSFD method offers a reliable solution for CB fault diagnosis and provides a foundation for future improvements in symptom-based zero-shot diagnostic mechanisms and algorithmic robustness.

A hybrid deep learning model for fault diagnosis of rolling bearings using raw vibration signals

Abstract The fault symptoms of rolling bearings are subject to various interferences in complex industrial environments, so achieving accurate, robust, and generalized fault diagnosis has become a key research direction. This article proposes a rolling bearing fault diagnosis method based on 1D-Inception-SE, which combines the 1D-Inception network model with Squeeze and Excitation Attention and can directly use the original vibration signals for fault diagnosis. The method incorporates the Adaptive Batch Normalization algorithm to enhance the model’s generalization performance in the presence of noise interference and cross-load diagnostics. Performance tests on Paderborn University Bearing and Case Western Reserve University datasets show that our approach achieves superior recognition accuracy compared to other models under similar and varied loads, as well as different signal to noise ratio. Ablation and visualization tests confirm the rationality and effectiveness of the model structure.

Research on the optimization of belief rule bases using the Naive Bayes theory

The belief rule base is crucial in expert systems for intelligent diagnosis of equipment. However, in the belief rule base for fault diagnosis, multiple antecedent attributes are often initially determined by domain experts. Multiple fault symptoms related to multiple antecedent attributes are different when an actual fault occurs. This leads to multiple antecedent attributes matching with multiple fault symptoms non-simultaneously, thereby resulting in a fault diagnosis lacking timeliness and accuracy. To address this issue, this paper proposes a method for belief rule-based optimization based on Naive Bayes theory. First, a fault sample is taken in a long enough window and divided into several interval samples, making the analysis samples approximate the overall samples. Second, using Gaussian mixture clustering and Naive Bayes optimization, iteration is performed over the threshold and limit values of fault symptoms in the belief rule base based on the requirements of the timeliness and accuracy of fault diagnosis results. Finally, the belief rule base is optimized. Using fault samples from high-pressure heaters and condensers, the validation results show that there is a there is a significant improvement in the timeliness and accuracy of fault diagnosis with the optimal belief rule base.

Early fault detection for rolling bearings: A meta‐learning approach

AbstractEarly fault detection (EFD) of rolling bearings aims at detecting the early symptoms of faults by monitoring small deviations of health states. Accurate EFD enables predictive maintenance and contributes to the stability of mechanical systems. In recent years, machine learning based methods have shown impressive performance on EFD. Most of the current machine learning‐based methods assume the availability for a large amount of data. However, in practice, the authors may only have a very limited amount of training data, which makes it hard to learn a reliable machine learning model. To address this concern, in this work, the authors propose to tackle EFD via meta learning. Specifically, the authors first formulate EFD as a few‐shot learning problem and then propose to tackle this problem with a metric‐based meta learning method. Furthermore, ensemble learning is further leveraged to improve the detection robustness. For the proposed method, the distribution difference from the working conditions and the bearings are considered. The experimental results on two bearing datasets show that the proposed method can achieve better EFD performance, that is, detecting incipient faults earlier while bringing in lower false alarms, compared with several frequently used EFD methods.

Deep exponential excitation networks: toward stronger attention mechanism for weak fault diagnosis

Considering that large mechanical equipment often has various excitation sources, the signals generated by these excitation sources are often not simply added or multiplied together, but nonlinearly mixed, which exhibit complex non-stationary characteristics, making classical algorithms difficult to extract fault features. Especially when faults just occur, the fault symptom is often weak and submerged by noise, resulting in low diagnosing accuracy. Accordingly, this article develops a new deep attention method, namely deep exponential excitation networks, which improves diagnosing performance by amplifying important information, that is, the discriminative information between normal and weak fault conditions. The developed method introduces an exponential function into the attention mechanism, yielding a stronger focus on important features. Here, our “stronger” means that the attention mechanism has larger weights, and wider weight ranges, which are achieved via three paradigms of exponential excitation blocks. Meanwhile, the weights are automatically learned in deep networks, which can adaptively amplify the information related to early weak faults according to different severities of faults. Finally, extensive experiments on the marine engine datasets containing different noises demonstrate the effectiveness of the developed method.

Improving compound fault diagnostic accuracy of drilling permanent magnet synchronous motor using information fusion

The diagnostic accuracy of compound faults is low due to the similarity of fault symptoms. Accurate diagnostic results can only be obtained by fusing all available information on compound faults. However, a single information fusion method cannot fuse all available information well because the characteristics of fault information are different. A method combining three kinds of information fusion technology is proposed to improve the compound fault diagnostic accuracy of the drilling permanent magnet synchronous motor (DPMSM). The fault symptoms are fused by Bayesian inference using the Bayesian Network. The additional information is fused by fuzzy logic operation using the Fuzzy Theory. The fusion results obtained from the above two methods are fused as evidence by mathematical logic using Evidence Theory to obtain the final diagnosis results. The effectiveness of the proposed method is verified by Cases. After fusion, “RoEc” fault in Case A rises to 0.933 from 0.826, “ItSc” fault in Case B rises to 0.524 from 0.262, and “RoEc” fault in Case C rises to 0.344 from 0.046 while “ItSc” down to 0.172 from 0.458. Cases indicate it can effectively improve the fault diagnostic accuracy and correct the results of missed diagnosis and misdiagnosis in compound faults of the DPMSM in drilling engineering. The effect of this method is influenced by the selection of specific experts and the accuracy of statistical data.

Analysis of PMSM Short-Circuit Detection Systems Using Transfer Learning of Deep Convolutional Networks

Abstract Modern permanent magnet synchronous motor (PMSM) diagnostic systems are now combined with advanced artificial intelligence techniques, such as deep neural networks. However, the design of such systems is mainly focussed on a selected type of damage or motor type with a limited range of rated parameters. The application of the idea of transfer learning (TL) allows the fully automatic extraction of universal fault symptoms, which can be used for various diagnostic tasks. In the research, the possibility of using the TL idea in the implementation of PMSM stator windings fault-detection systems was considered. The method is based on the characteristic symptoms of stator defects determined for another type of motor or mathematical model in the target diagnostic application of PMSM. This paper presents a comparison of PMSM motor inter-turn short circuit fault detection systems using TL of a deep convolutional network. Due to the use of direct phase current signal analysis by the convolutional neural network (CNN), it was possible to ensure high accuracy of fault detection with simultaneously short reaction time to occurring fault. The technique used was based on the use of a weight coefficient matrix of a pre-trained structure, the adaptation of which was carried out for different sources of diagnostic information.

Using Non-stationary Time-Domain Statistical Measures for Predictive Maintenance of Large Centrifugal Compressors in the Absence of Component Faults

Time-domain statistical measures such as standard deviation, skewness, and kurtosis are simple yet effective tools for assessing the health of rotating machinery based on measured vibration signals. These statistical quantities are, therefore, widely used in industry when designing condition-based maintenance (CBM) systems, especially for rotating machinery. Despite the growing adoption of CBM systems, determining the overhaul schedule can still be challenging, particularly when the rotating machinery in question does not exhibit any fault symptoms. As a result, the overhaul schedule is often determined using the traditional time-based maintenance (TBM) approach. In this short communication, we examine the time-varying characteristics of statistical quantities immediately before and after the overhaul of a large centrifugal compressor. It was found that the non-stationarity of the standard deviation increases in cases where an overhaul is due, while the variability of skewness and kurtosis does not change significantly in the absence of specific fault components.

Fault Diagnosis of Airborne Electronic Equipment Based on Dynamic Bayesian Networks

With the rapid development of the aerospace industry, the structure of airborne electronic equipment has become more complex, which to some extent increases the difficulty of fault detection and maintenance of airborne electronic equipment. Traditional manual fault diagnosis methods can no longer fully meet the diagnostic needs of airborne electronic equipment. Therefore, this chapter uses dynamic Bayesian network to diagnose the faults of airborne electronic equipment. The basic idea of using a dynamic Bayesian network-based fault diagnosis method for airborne electronic devices is to mine data based on historical fault data of airborne electronic devices, and obtain fault symptoms and training data of airborne electronic devices. For non-essential fault symptoms, rough set theory was introduced to reduce their attributes and obtain the simplest attribute set, thereby simplifying the network model.

Fault diagnosis-based SDG transfer for zero-sample fault symptom

The traditional fault diagnosis models cannot achieve good fault diagnosis accuracy when a new unseen fault class appears in the test set, but there is no training sample of this fault in the training set. Therefore, studying the unseen cause-effect problem of fault symptoms is extremely challenging. As various faults often occur in a chemical plant, it is necessary to perform fault causal-effect diagnosis to find the root cause of the fault. However, only some fault causal-effect data are always available to construct a reliable causal-effect diagnosis model. Another worst thing is that measurement noise often contaminates the collected data. The above problems are very common in industrial operations. However, past-developed data-driven approaches rarely include causal-effect relationships between variables, particularly in the zero-shot of causal-effect relationships. This would cause incorrect inference of seen faults and make it impossible to predict unseen faults. This study effectively combines zero-shot learning, conditional variational autoencoders (CVAE), and the signed directed graph (SDG) to solve the above problems. Specifically, the learning approach that determines the cause-effect of all the faults using SDG with physics knowledge to obtain the fault description. SDG is used to determine the attributes of the seen and unseen faults. Instead of the seen fault label space, attributes can easily create an unseen fault space from a seen fault space. After having the corresponding attribute spaces of the failure cause, some failure causes are learned in advance by a CVAE model from the available fault data. The advantage of the CVAE is that process variables are mapped into the latent space for dimension reduction and measurement noise deduction; the latent data can more accurately represent the actual behavior of the process. Then, with the extended space spanned by unseen attributes, the migration capabilities can predict the unseen causes of failure and infer the causes of the unseen failures. Finally, the feasibility of the proposed method is verified by the data collected from chemical reaction processes.

Expert system based fault diagnosis for railway point machines

To meet the increasing demands for availability at reasonable cost, operators and maintainers of railway point machines are constantly looking for innovative techniques for switch condition monitoring and prediction. This includes automated fault root cause diagnosis based on measurement data (such as motor current curves) and other information. However, large, comprehensive sets of labeled data suitable for standard machine learning are not yet available. Existing data-driven approaches focus only on the differentiation of a few major fault categories at the level of the measurement data (i.e., the “fault symptoms”). There is great potential in hybrid models that use expert knowledge in combination with multiple sources of information to automatically identify failure causes at a much more detailed level. This paper discusses a Bayesian network diagnostic model for determining the root causes of faults in point machines, based on expert knowledge and few labeled data examples from the Netherlands. Human-interpretable current curve features and other information sources (e.g., past maintenance actions) are used as evidence. The result of the model is a ranking of the most likely failure causes with associated probabilities in terms of fuzzy multi-label classification, which is directly aimed at providing decision support to maintenance engineers. The validity and limitations of the model are demonstrated by a scenario-based evaluation and a brief analysis using information theoretic measures. We present the information sources used, the detailed development process and the analysis methodology. This article is intended to be a guide to developing similar models for various complex technical assets.

A study on the decoupling of helical gear compound uniform wear and tooth root crack faults via gear condition indicators based on dynamic modeling

Diagnosis of compound faults plays a significant role in evaluating the health status of a gearbox. The difficulty of compound fault diagnosis is due to the entanglement of fault symptoms. In this work, diagnosis of single and compound uniform wear and tooth root crack via gear condition indicators have been studied analytically. A three-dimensional dynamic model that can include several design parameters and working conditions has been employed to model the dynamic of compound faults for the first time. The effect of fault severity, noise, working conditions, and design parameters on the diagnosis efficiency have been investigated. The qualitative and quantitative assessment of fault diagnosis has been accomplished via t-SNE and Generalized Discriminative Value (GDV) measures, respectively. The results show that the gear condition indicators are more suitable to assess the general health state of the gearbox. However, their ability to detect the type of fault and decouple compound faults are limited. Furthermore, design parameters of the gear transmission system, such as helix angle and gear width, can highly affect the fault diagnosis efficiency.

Design of a progressive fault diagnosis system for hydropower units considering unknown faults

To address the misidentification problem of signals containing unknown faults for hydropower units, a progressive fault diagnosis system is designed. Firstly, in view of the non-stationary and nonlinear vibration signals of hydropower units, the method of complementary ensemble empirical mode decomposition is used to process the normal and fault vibration signal samples, and the intrinsic mode function (IMF) and residual components with different frequencies are obtained. Then the IMF energy moment is calculated and used as the feature vector. Furthermore, a classifier (IMF-K1) is constructed based on the feature vector samples of the normal vibration signals of hydropower units, fault symptom indicators, and K-means algorithm to determine whether the hydropower unit is faulty; a classifier (IMF-K2) is constructed based on the feature vector samples of the fault vibration signals of hydropower units, fault symptom indicators, and K-means algorithm to determine whether the hydropower unit has the known fault; a classifier (IMF-bidirectional long short-term memory neural network (BiLSTMNN)) is constructed to distinguish the fault type of hydropower units by combining the eigenvector samples of known fault vibration signals, fault symptom indicators, and BiLSTMNN. Finally, a progressive fault diagnosis system for hydropower units is constructed using IMF-K1, IMF-K2, and IMF-BiLSTMNN, and comparative experiments are designed using the sample data from the rotor test bench and actual hydropower unit. The results show that the designed progressive fault diagnosis system has greater effectiveness in mining signal features and high fault diagnosis accuracy.

Rise of Distributed Deep Learning Training in the Big Model Era: From a Software Engineering Perspective

Deep learning (DL) has become a key component of modern software. In the “ big model ” era, the rich features of DL-based software (i.e., DL software) substantially rely on powerful DL models, e.g., BERT, GPT-3, and the recently emerging GPT-4, which are trained on the powerful cloud with large datasets. Hence, training effective DL models has become a vital stage in the whole software lifecycle. When training deep learning models, especially those big models, developers need to parallelize and distribute the computation and memory resources amongst multiple devices (e.g., a cluster of GPUs) in the training process, which is known as distributed deep learning training , or distributed training for short. However, the unique challenges that developers encounter in distributed training process have not been studied in the software engineering community. Given the increasingly heavy dependence of current DL-based software on distributed training, this paper aims to fill in the knowledge gap and presents the first comprehensive study on developers’ issues in distributed training. To this end, we focus on popular DL frameworks that support distributed training (including TensorFlow, PyTorch, Keras, and Horovod) and analyze 1,131 real-world developers’ issues about using these frameworks reported on Stack Overflow and GitHub. We construct a fine-grained taxonomy consisting of 30 categories regarding the fault symptoms and summarize common fix patterns for different symptoms. We find that: (1) many distributed-specific faults and non-distributed-specific faults inherently share the same fault symptoms, making it challenging to debug; (2) most of the fault symptoms have frequent fix patterns; (3) about half of the faults are related to system-level configurations. Based on the results, we suggest actionable implications on research avenues that can potentially facilitate the distributed training to develop DL-based software, such as focusing on the frequent and common fix patterns when designing testing or debugging tools, developing efficient testing and debugging techniques for communication configuration along with the synthesis of network configuration analysis, designing new multi-device checkpoint-and-replay techniques to help reproduction, and designing serverless APIs for cloud platforms.

Expert Knowledge Transfer from CAE Models to CNN Models Using Enhanced Adversarial Domain Adaptation

It is a common belief that convolutional neural networks (CNN) are incapable of acquiring knowledge from domain experts for fault detection and diagnosis. To address the challenge, this paper proposes a knowledge-transfer scheme from computer-aided engineering (CAE) models to CNN models. Domain experts build the CAE models that emulate the faulty behavior of rotating machines by incorporating fault symptom and controlling the degree of fault severity. Fault data are hardly acquired from rotating machines in the field, while a sufficient number of fault data can be generated using the CAE models. Then, a domain adaption model is trained using synthetic data (i.e., normal and fault data) from the CAE models and real data (i.e., normal data only) from rotating machines. To evaluate the validity of the proposed method, a small-scale testbed is regarded as the target system that does not have any fault data. This study contributes to resolve the dearth of fault data from most safety-related engineering assets such as power plant steam turbines, wind turbines, and urban air mobility.

A two-stage data quality improvement strategy for deep neural networks in fault severity estimation

As fault diagnosis based on deep neural network (DNN) has been developing from laboratory to industrial application, creating the appropriate data pipeline to train and evaluate the models has become a noticeable challenge. For mechanical fault diagnosis, high-quality data should be consistent with comprehensive fault information. In this paper, a two-stage data quality improvement strategy is proposed to sculpt the data for deep learning models in fault severity estimation. In the first stage, the spatial information reconstruction (SIR) approach is developed to transform the sensing data from time domain into spatial domain to establish the relationship between fault symptom and spatial position. In the second stage, spatial domain signals are organized based on fault characteristics and types of DNN models, which aims to convert data into useful information. The proposed strategy has been verified by experimental cases, and a comprehensive evaluation on the performance of DNNs trained by different input data has been applied. The experimental results proved that with thoughtfully engineered data, baseline DNN models can achieve high accuracy and robustness under different speed conditions. This paper provides a way for engineering practitioners to design industrial DNN applications with their domain knowledge.

A just-in-time manifold-based fault detection method for electrical drive systems of high-speed trains

Electrical drive systems of high-speed trains are typical complex industrial systems with dynamic nonlinearity. During the actual operation of high-speed trains, the operation state is switched to meet the operation requirements, which leads to the multi-mode characteristics of electrical drive systems. Inherent characteristics of electrical drive systems have brought great obstacles to common fault detection methods. Therefore, online detection of incipient faults in electrical drive systems is imperative. On the one hand, the symptoms of incipient faults are slight and easy to be covered by unknown noises and disturbances; On the other hand, incipient faults will corrupt the health state and system remaining life, and gradually evolve into destructive faults. With the help of the idea to solve global problems through local modeling, this paper constructs a just-in-time manifold model by integrating local manifold learning into the just-in-time learning framework. The proposed scheme avoids the loss of feature information in the global structure by extracting the feature information of each local structure. The model construction is based on the eigenstructure of local data, which reduces the computational complexity of modeling and improves the detection accuracy. Ultimately, the efficacy and superiority of the proposed scheme are illustrated via a series of experiments on a platform of electrical drive systems.

A Novel Framework of Cooperative Design: Bringing Active Fault Diagnosis Into Fault-Tolerant Control.

Fault-tolerant control (FTC) may conceal fault symptoms, thereby increasing the difficulty of fault diagnosis (FD). In this article, a novel framework for the cooperative design of active FD and FTC is proposed to optimize FD performance while maintaining fault-tolerance performance. The proposed framework consists of four steps: 1) controller design; 2) residual generation; 3) performance evaluation; and 4) gain tuning. First, a controller with undetermined gains is constructed, and a fault detection observer is designed to generate residuals that can indicate the fault. Then, the performance of fault detection is evaluated. Finally, suboptimal controller gains are obtained by solving an optimization problem. Within the framework of the collaborative design, the occurring faults can be detected faster and more accurately, and the performance of FTC can be guaranteed at the same time. A simulation study is provided to demonstrate the effectiveness of the developed framework.