Unsupervised Fault Detection Research Articles

Unsupervised fault detection and recovery for intelligent robotic rollators

Fault diagnosis is a key safety component in robotic assistive technologies. Although conventional model-based methods for sensor fault diagnosis in mobile robots have been well established, they face challenges due to model parameter changes and uncertainties. On the other hand, data-driven approaches becomes more appealing in order to take advantage from available historical data in the era of Big Data. To provide a new generic unsupervised solution to the fault detection and recovery, we explicitly include kinematic relations and temporal finite differences from measured sensor signals into training a multi-task deep neural network. To evaluate the proposed fault diagnosis and recovery framework, experiments have been conducted on a robotic rollator platform. Experiments under several conditions confirm that the proposed approach, which leverages machine learning-enhanced algorithms, exhibits reliable performance. Outperforming other baselines and state-of-the-art diagnosis algorithms, the framework presents a promising solution to sensor fault recovery challenges in assistive devices.

Generative Adversarial Neural Network for Unsupervised Bearing Fault Detection

Detecting bearing faults in advance is critical for mechanical and electrical systems to prevent economic loss and safety hazards. As part of the recent interest in artificial intelligence, deep learning (DL)-based principles have gained much attention in intelligent fault diagnostics and have mainly been developed in a supervised manner. While these works have shown promising results, several technical setbacks are inherent in a supervised learning setting. Data imbalance is a critical problem as faulty data is scarce in many cases, data labeling is tedious, and unseen cases of faults cannot be detected in a supervised framework. Herein, a generative adversarial network (GAN) is proposed to achieve unsupervised bearing fault diagnostics by utilizing only the normal data. The proposed method first adopts the short-time Fourier transform (STFT) to convert the 1-D vibration signals into 2-D time-frequency representations to use as the input to our (DL) framework. Subsequently, a GAN-based latent mapping is constructed using only the normal data, and faulty signals are detected using an anomaly metric comprised of a discriminator error and an image reconstruction error. The performance of our method is verified using a classic rotating machinery dataset (Case Western Reserve bearing dataset), and the experimental results demonstrate that our method can not only detect the faults but can also cluster the faults in the latent space with high accuracy.

An explainable artificial intelligence approach for unsupervised fault detection and diagnosis in rotating machinery

The monitoring of rotating machinery is an essential task in today’s production processes. Currently, several machine learning and deep learning-based modules have achieved excellent results in fault detection and diagnosis. Nevertheless, to further increase user adoption and diffusion of such technologies, users and human experts must be provided with explanations and insights by the modules. Another issue is related, in most cases, with the unavailability of labeled historical data that makes the use of supervised models unfeasible. Therefore, a new approach for fault detection and diagnosis in rotating machinery is here proposed. The methodology consists of three parts: feature extraction, fault detection and fault diagnosis. In the first part, the vibration features in the time and frequency domains are extracted. Secondly, in the fault detection, the presence of fault is verified in an unsupervised manner based on anomaly detection algorithms. The modularity of the methodology allows different algorithms to be implemented. Finally, in fault diagnosis, Shapley Additive Explanations (SHAP), a technique to interpret black-box models, is used. Through the feature importance ranking obtained by the model explainability, the fault diagnosis is performed. Two tools for diagnosis are proposed, namely: unsupervised classification and root cause analysis. The effectiveness of the proposed approach is shown on three datasets containing different mechanical faults in rotating machinery. The study also presents a comparison between models used in machine learning explainability: SHAP and Local Depth-based Feature Importance for the Isolation Forest (Local-DIFFI). Lastly, an analysis of several state-of-art anomaly detection algorithms in rotating machinery is included.

End-to-end unsupervised fault detection using a flow-based model

Fault detection has been extensively studied in both academia and industry. The rareness of faulty samples in the real world restricts the use of many supervised models, and the reliance on domain expertise for feature engineering raises other barriers. For this reason, this paper proposes an unsupervised, end-to-end approach to fault detection based on a flow-based model, the Nonlinear Independent Components Estimation (NICE) model. A NICE model models a target distribution via a sequence of invertible transformations to a prior distribution in the latent space. We prove that, under certain conditions, the L2-norm of normal samples’ latent codes in a trained NICE model is Chi-distributed. This facilitates the use of hypothesis testing for fault detection purpose. Concretely, we first apply Zero-phase Component Analysis to decorrelate the data of normal states. The whitened data are fed to a NICE model for training, in a maximum likelihood sense. At the testing stage, samples whose L2-norm of latent codes fail in the hypothesis testing are suspected of being generated by different mechanisms and hence regarded as potential faults. The proposed approach was validated on two datasets of vibration signals; it proved superior to several alternatives. We also show the use of NICE, a type of generative model, can produce real-like vibration signals because of the model's bijective nature.

Unsupervised Fault Detection for Refrigeration Showcase Systems with Kernel Principal Component Analysis based Multivariate Statistical Process Control using Feature Selection with Maximal Information Coefficient

Unsupervised Fault Detection for Refrigeration Showcase Systems with Kernel Principal Component Analysis based Multivariate Statistical Process Control using Feature Selection with Maximal Information Coefficient

Unsupervised Fault Detection on Unmanned Aerial Vehicles: Encoding and Thresholding Approach.

Unmanned Aerial Vehicles are expected to create enormous benefits to society, but there are safety concerns in recognizing faults at the vehicle’s control component. Prior studies proposed various fault detection approaches leveraging heuristics-based rules and supervised learning-based models, but there were several drawbacks. The rule-based approaches required an engineer to update the rules on every type of fault, and the supervised learning-based approaches necessitated the acquisition of a finely-labeled training dataset. Moreover, both prior approaches commonly include a limit that the detection model can identify the trained type of faults only, but fail to recognize the unseen type of faults. In pursuit of resolving the aforementioned drawbacks, we proposed a fault detection model utilizing a stacked autoencoder that lies under unsupervised learning. The autoencoder was trained with data from safe UAV states, and its reconstruction loss was examined to distinguish the safe states and faulty states. The key contributions of our study are, as follows. First, we presented a series of analyses to extract essential features from raw UAV flight logs. Second, we designed a fault detection model consisting of the stacked autoencoder and the classifier. Lastly, we validated our approach’s fault detection performance with two datasets consisting of different types of UAV faults.

A Method for Detecting Incipient Faults in Satellites Based on Dynamic Linear Discriminant Analysis.

Timely detection and treatment of possible incipient faults in satellites will effectively reduce the damage and harm they could cause. Although much work has been done concerning fault detection problems, the related questions about satellite incipient faults are little addressed. In this paper, a new satellite incipient fault detection method was proposed by combining the ideas of deviation in unsupervised fault detection methods and classification in supervised fault detection methods. First, the proposed method uses dynamic linear discriminant analysis (LDA) to find an optimal projection vector that separates the in-orbit data from the normal historical data as much as possible. Second, under the assumption that the parameters obey a multidimensional Gaussian distribution, it applies the normal historical data and the optimal projection vector to build a normal model. Finally, it employs the noncentral F-distribution to test whether a fault has occurred. The proposed method was validated using a numerical simulation case and a real satellite fault case. The results show that the method proposed in this paper is more effective at detecting incipient faults than traditional methods.

An Unsupervised Fault-Detection Method for Railway Turnouts

Railway turnouts require high-performance condition monitoring to prevent disastrous railway accidents. In industrial practice, turnouts’ monitoring is usually done by railway workers who visually inspect the operating current curves. This results in huge labor costs and prone to human mistakes. Thus, automating the process of turnouts’ monitoring via fault-detection algorithms is imperative. The available turnout field data bring three difficulties to fault detection: 1) large amounts of data do not have any labels; 2) data collected in normal condition have multiple unknown modes; and 3) there are only a small number of samples in some modes. To address these difficulties, this article develops a novel unsupervised fault-detection method by using deep autoencoders, which is composed of an unknown modes’ mining stage and a multimode fault-detection stage. First, unknown modes are identified through clustering and employing engineer expertise. Then, an ensemble monitoring model, consisting of local monitoring models developed with individual fault-free modes and a global monitoring model developed by merging the data in all fault-free modes, is proposed to improve the overall fault-detection performance. In addition, to construct local models for the modes with a small number of samples, a one-class transfer learning algorithm is presented. In online monitoring, the decision of a newly arrived sample exploits both local models and the global model. Using both the simulated turnout data and the field data collected from a high-speed railway in China, the efficacy and robustness of the proposed approach are demonstrated by comparisons with other methods.

Unsupervised Fault Detection and Prediction of Remaining Useful Life for Online Prognostic Health Management of Mechanical Systems

Predictive maintenance allows industries to keep their production systems available as much as possible. Reducing unforeseen shutdowns to a level that is close to zero has numerous advantages, including production cost savings, a high quality level of both products and processes, and a high safety level. Studies in this field have focused on a novel approach, prognostic health management (PHM), which relies on condition monitoring (CM) for predicting the remaining useful life (RUL) of a system. However, several issues remain in its application to real industrial contexts, e.g., the difficulties in conducting tests simulating each fault condition, the dynamic nature of industrial environments, and the need to handle large amounts of data collected from machinery. In this paper, a data-driven methodology for PHM implementation is proposed, which has the following characteristics: it is unsupervised, i.e., it does not require any prior knowledge regarding fault behaviors and it does not rely on pre-trained classification models, i.e., it can be applied “from scratch”; it can be applied online due to its low computational effort, which makes it suitable for edge computing; and, it includes all of the steps that are involved in a prognostic program, i.e., feature extraction, health indicator (HI) construction, health stage (HS) division, degradation modelling, and RUL prediction. Finally, the proposed methodology is applied in this study to a rotating component. The study results, in terms of the ability of the proposed approach to make a timely prediction of component fault conditions, are promising.

Graph-Based Method for Fault Detection in the Iron-Making Process

Since the iron-making process is performed in complicated environments and controlled by operators, observation labeling is difficult and time-consuming. Therefore, unsupervised fault detection methods are a promising research topic. Recently, an unsupervised graph-based change point detection method has been introduced, and the graph of observations is constructed by the minimum spanning tree. In this paper, a novel fault detection method based on the graph for an iron-making process is proposed, and a weight calculation method for constructing the minimum spanning tree is introduced. The Euclidean distance and Mahalanobis distance are combined to calculate the weights in the minimum spanning tree, which contain important relations of variables. The distance calculation method is determined by the correlation coefficients of variables. Each testing observation is set as a change point candidate, and a change point candidate divides the observations into two groups. The number of a special type of edge in the minimum spanning tree is used as a fault detection statistic. That special edge connects two observations from two different groups. The minimum number of that type of edge corresponding to the change point candidate is a true change point. Finally, numerical simulation is used to test the power of the proposed method, and a real iron-making process including low stock, cooling, and slip faults is implemented to illustrate the effectiveness of fault detection in industrial processes.

Unsupervised Fault Detection in Refrigeration Showcase with Single Class Data using Autoencoders

Unsupervised Fault Detection in Refrigeration Showcase with Single Class Data using Autoencoders

Data-Driven Mode Identification and Unsupervised Fault Detection for Nonlinear Multimode Processes

In modern plants, industrial processes typically operate under different states to meet the different requirements of high-quality products. Many monitoring models for industrial processes were constructed based on the prior knowledge (the mechanism's model or the process data characteristics) to monitor such processes (called multimode processes). However, obtaining this prior knowledge is difficult in practice. Efficiently monitoring nonlinear multimode processes without any prior knowledge is an open problem that demands further exploration. Since data from different modes follow different distributions while data from the same mode are considered to be sampled from the same distribution, the modes of multimode processes can be uncovered based on the characteristics of the process data. This article proposes using a Dirichlet process Gaussian mixed model to classify the modes of multimode processes based on historical data, and then, determine the mode types of the monitored data. A nonlinear monitoring strategy based on the t-distributed stochastic neighbor embedding is then proposed to achieve nonlinear dimensionality reduction and visualize the data. Finally, a monitoring index that is integrated with support vector data description is constructed for comprehensive monitoring. The proposed nonlinear multimode framework completely realizes data-driven mode identification and unsupervised fault detection without knowing any prior knowledge. The effectiveness and feasibility of the proposed model are demonstrated using data from a simulated wastewater treatment plant.

Unsupervised Fault Detection and Analysis for Large Photovoltaic Systems Using Drones and Machine Vision

One of the most important sources of clean energy in the future is expected to be solar energy which is considered a real time source. Research efforts to optimize solar energy utilization are mainly concentrated on the components of solar energy systems. Photovoltaic (PV) modules are considered the main components of solar energy systems and PVs’ operations typically occur without any supervisory mechanisms, which means many external and/or internal obstacles can occur and hinder a system’s efficiency. To avoid these problems, the paper presents a system to address and detect the faults in a PV system by providing a safer and more time efficient inspection system in real time. In this paper, we proposing a real time inspection and fault detection system for PV modules. The system has two cameras, a thermal and a Charge-Coupled Device CCD. They are mounted on a drone and they used to capture the scene of the PV modules simultaneously while the drone is flying over the solar garden. A mobile PV system has been constructed primarily to validate our real time proposed system and for the proposed method in the Digital Image and Signal Processing Laboratory (DISPLAY) at Western Michigan University (WMU). Defects have been detected accurately in the PV modules using the proposed real time system. As a result, the proposed drone mounted system is capable of analyzing thermal and CCD videos in order to detect different faults in PV systems, and give location information in terms of panel location by longitude and latitude.

A hybrid approach for improving unsupervised fault detection for robotic systems

The use of robots in our daily lives is increasing. As we rely more on robots, thus it becomes more important for us that the robots will continue on with their mission successfully. Unfortunately, these sophisticated, and sometimes very expensive, machines are susceptible to different kinds of faults. It becomes important to apply a Fault Detection (FD) mechanism which is suitable for the domain of robots. Two important requirements of such a mechanism are: high accuracy and low computational-load during operation (online). Supervised learning can potentially produce very accurate FD models, and if the learning takes place offline then the online computational-load can be reduced. Yet, the domain of robots is characterized with the absence of labeled data (e.g., “faulty”, “normal”) required by supervised approaches, and consequently, unsupervised approaches are being used. In this paper we propose a hybrid approach - an unsupervised approach can label a data set, with a low degree of inaccuracy, and then the labeled data set is used offline by a supervised approach to produce an online FD model. Now, we are faced with a choice – should we use the unsupervised or the hybrid fault detector? Seemingly, there is no way to validate the choice due to the absence of (a priori) labeled data. In this paper we give an insight to why, and a tool to predict when, the hybrid approach is more accurate. In particular, the main impacts of our work are (1) we theoretically analyze the conditions under which the hybrid approach is expected to be more accurate. (2) Our theoretical findings are backed with empirical analysis. We use data sets of three different robotic domains: a high fidelity flight simulator, a laboratory robot, and a commercial Unmanned Arial Vehicle (UAV). (3) We analyze how different unsupervised FD approaches are improved by the hybrid technique and (4) how well this improvement fits our prediction tool. The significance of the hybrid approach and the prediction tool is the potential benefit to expert and intelligent systems in which labeled data is absent or expensive to create.

Evaluating Unsupervised Fault Detection in Self-Healing Systems Using Stochastic Primitives

This research was partially supported by the Scottish Informatics and Computer Science Alliance (SICSA).

Fully unsupervised fault detection and identification based on recursive density estimation and self-evolving cloud-based classifier

In this paper, we propose a two-stage algorithm for real-time fault detection and identification of industrial plants. Our proposal is based on the analysis of selected features using recursive density estimation and a new evolving classifier algorithm. More specifically, the proposed approach for the detection stage is based on the concept of the density in the data space, which is not the same as the probability density function, but is a very useful measure for abnormality/outliers detection. This density can be expressed by a Cauchy function and can be calculated recursively, which makes it memory and computational power efficient and, therefore, applicable to on-line applications. The identification/diagnosis stage is based on a self-developing (evolving) fuzzy-rule-based classifier system proposed in this paper, called the AutoClass. An important property of AutoClass is that it can start learning “from scratch”. Not only do the fuzzy rules not need to be prespecified, but neither do the number of classes for AutoClass (the number may grow, with new class labels being added by the online learning process), in a fully unsupervised manner. In the event that an initial rule base exists, AutoClass can evolve/develop it further based on the newly arrived faulty state data. In order to validate our proposal, we present experimental results from a level control didactic process, where control and error signals are used as features for the fault detection and identification system, but the approach is generic and the number of features can be significant due to the computationally lean methodology, since covariance or more complex calculations, as well as storage of old data, are not required. The obtained results are significantly better than the traditional approaches.

A support vector clustering‐based probabilistic method for unsupervised fault detection and classification of complex chemical processes using unlabeled data

A new support vector clustering (SVC)‐based probabilistic approach is developed for unsupervised chemical process monitoring and fault classification in this article. The spherical centers and radii of different clusters corresponding to normal and various kinds of faulty operations are estimated in the kernel feature space. Then the geometric distance of the monitored samples to different cluster centers and boundary support vectors are computed so that the distance–ratio–based probabilistic‐like index can be further defined. Thus, the most probable clusters can be assigned to the monitored samples for fault detection and classification. The proposed SVC monitoring approach is applied to two test scenarios in the Tennessee Eastman Chemical process and its results are compared to those of the conventional K‐nearest neighbor Fisher discriminant analysis (KNN‐FDA) and K‐nearest neighbor support vector machine (KNN‐SVM) methods. The result comparison demonstrates the superiority of the SVC‐based probabilistic approach over the traditional KNN‐FDA and KNN‐SVM methods in terms of fault detection and classification accuracies. © 2012 American Institute of Chemical Engineers AIChE J, 59: 407–419, 2013

Standoff Bearing Fault Detection Using Directional Microphones and Unsupervised Neural Networks

A fault detection method is introduced that uses standoff directional microphones to minimize the number of sensors and an unsupervised neural network to cope with noise and fault signature variability in the microphone signals. In this method, a directional microphone located up to 25 feet is used to sense acoustic signals from a test bearing. These signals are then processed to extract features representing the bearing condition and are used as inputs to an unsupervised fault detection network (FDN) to identify the presence of faults. The main advantage of the FDN is that it does not require seeded-fault data for supervised training of its weights. The proposed fault detection method is tested using microphone data from a bearing with an inner race defect. The results indicate that the FDN provided a 100% detection rate for all microphone locations.