Fault Isolation Approach Research Articles

Deep transfer learning strategy in intelligent fault diagnosis of gas turbines based on the Koopman operator

The gas turbine engine is a predominant prime mover in the transport and energy sectors, and ensuring its reliable operation holds paramount significance. While intelligent fault diagnosis (FD) approaches have seen successful advancements within the gas turbine FD landscape, many existing methods operate under the assumption of identical health states during both data collection and the FD process. Moreover, most previous studies have overlooked the diagnosis of both sensors and actuators. Another critical challenge lies in isolating simultaneous and multiple faults and providing compromising FD performance, especially in the face of continued system performance degradation. Aiming at these problems, this study develops a novel unsupervised data-driven FD strategy based on leveraging the potential of transfer learning and the Koopman operator. A deep neural network-based transfer learning framework is proposed for realizing a precise adaptive linear model called the deep transfer linear (DTL) model enabling reliable prediction of the system’s behavior in various situations and designing structured fault residuals. To this end, a deep neural network framework is used to obtain a precise Koopman model realization using the richly collected data in the source domain. Subsequently, the realized model is fine-tuned for the target domains associated with the degraded system, mitigating the adverse effects of domain shift and addressing the rich data scarcity problem in the target domain. In addition, the dedicated and generalized residual sets are designed and generated employing the geometric approach for fault isolation and a decision-making analysis is developed to diagnose simultaneous faults. The reliability of the proposed strategy is demonstrated through various experiments in the presence of noise and performance degradation, and a comparative performance analysis is conducted between the proposed strategy and another data-driven method showcasing the superiority of the proposed approach.

Actuator and sensor fault isolation in a class of nonlinear dynamical systems

Fault isolation in dynamical systems is a challenging task due to modeling uncertainty and measurement noise, interactive effects of multiple faults and fault propagation. This paper proposes a unified approach for isolation of multiple actuator or sensor faults in a class of nonlinear uncertain dynamical systems. Actuator and sensor fault isolation are accomplished in two independent modules, that monitor the system and are able to isolate the potential faulty actuator(s) or sensor(s). For the sensor fault isolation (SFI) case, a module is designed which monitors the system and utilizes an adaptive isolation threshold on the output residuals computed via a nonlinear estimation scheme that allows the isolation of single/multiple faulty sensor(s). For the actuator fault isolation (AFI) case, a second module is designed, which utilizes a learning-based scheme for adaptive approximation of faulty actuator(s) and, based on a reasoning decision logic and suitably designed AFI thresholds, the faulty actuator(s) set can be determined. The effectiveness of the proposed fault isolation approach developed in this paper is demonstrated through a simulation example.

A new approach for fault detection and isolation of Cardiovascular system

This article analyzes the possibility of using observer-based approaches for the detection of cardiovascular anomalies through a representation of the model written in a quadratic observability normal form. This particular representation allows to design an observer that is sensitive to changes in the cardiovascular mitral and aortic valves. The idea of the present result is based on the design of a residual generator that allows analyzing when a failure occurs in the heart valves. In addition, numerical simulations are performed in order to show the performance and capacity of the proposed strategy.

Modeling of fault detection and isolation in pitch control of wind energy conversion systems via digital twins

There is a growing trend in renewable energies for wind harvesting and power generation. As new offers for the energy sector continue to rise up, the operation and maintenance team plays a vital role in increasing the availability and decreasing downtime of wind turbines subassemblies, based on conditional monitoring of their assets. This paper developed an approach for fault detection and isolation of the pitch control system on a direct drive permanent magnet synchronous generator using dynamic modeling and digital twins. The work presents two models of wind turbines, whose conditions were replicated in a multi-domain simulation software and the analyses were conducted under rotor constant speed. One concept represents a wind turbine model with a fault in pitch system, while the other is a healthy replica or a virtual model. Through the difference between three observable variables, namely: active power, angular speed and pitch angle, it was possible to isolate and classify each type of fault in the pitch system in wind turbines.

Small Fault Diagnosis With Gap Metric

This article proposes a novel data-driven gap metric fault detection and isolation (FDI) approach for small multiplicative fault. First, the scheme of model-based fault classification and gradation is developed by means of the gap metric. Subsequently, the data-driven gap metric is utilized to detect a small fault via the mechanism model. Furthermore, fault detectability criterion is derived with the help of the developed fault detectability indicator. The relationship between fault detectability indicator and fault detection index is then investigated to analyze fault detection performance. To enhance fault isolability, a solution of appropriate fault cluster center model and radius is provided under the condition of fault isolation. Third, a gap metric fault-tolerant control strategy is exploited to guarantee system stability when a large fault is diagnosed by the developed FDI approach. The speed regulation of dc-motor and dc–dc converter are used for simulation and experiment verifications. Moreover, the comparison results and Monte Carlo simulation demonstrate the superiority and reliability of the proposed method.

A sensor fault detection, isolation, and estimation method for intelligent vehicles

Accurate and reliable sensor information is crucial for intelligent vehicles to ensure safety and reliability. It is significant to devise effective solutions to diagnose possible sensor faults. In this paper, a sensor fault detection, isolation, and estimation approach is developed for intelligent vehicle-integrated motion systems. The Luenberger observer is designed and embedded into the fault diagnosis framework to real-time estimate the system states. Interval observers are constructed for vehicle subsystems to estimate the interval bounds of the corresponding observation errors. The estimation errors enter into the corresponding intervals generated by the adaptive laws when the sensor fault is free. In the event of a sensor fault, it can be detected efficiently and the faulty sensor can be isolated. Moreover, occurred sensor fault can be estimated by modifying the interval observer structure. The simulation results in the standard J-Turn test scenario are provided to evaluate and verify the effectiveness of the proposed method.

A data-driven approach for fault diagnosis in multi-zone HVAC systems: Deep neural bilinear Koopman parity

Sensor faults in heating, ventilation, and air conditioning (HVAC) systems are inevitable and result in significant energy waste. This paper presents an innovative data-driven approach for sensor fault detection and isolation in multi-zone HVAC systems. The proposed solution integrates bilinear Koopman model realization, deep learning, and bilinear parity-space. A deep neural network realizes a bilinear model, enabling bilinear parity-space sensor fault detection and isolation. This yields a reliable, accurate, and interpretable data-driven framework. The method requires no prior HVAC dynamics knowledge, relying solely on normal operation data. It diagnoses additive, multiplicative, and complete failure sensor faults while minimizing false alarms, even with severe faults. A four-zone HVAC system is simulated in TRNSYS as a case study to demonstrate the performance and efficacy of the proposed approach. The proposed bilinear deep Koopman model realization is utilized to develop a bilinear model for the four-zone HVAC system. The bilinear model is then used for designing the bilinear parity-space. Further, considering various failure scenarios, the proposed sensor fault detection and isolation framework demonstrates promising diagnosis performance. Finally, a comparison is conducted to showcase the advantages of the proposed method over earlier works based on PCA and neural networks.

Simultaneous fault detection and isolation based on multi-task long short-term memory neural networks

Fault detection and isolation are two significant problems in process monitoring. However, due to the complicated relationships between the process variables, it becomes a challenging problem to build a model to capture the complicated relationship between process variables and perform fault detection and isolation based on the model. This paper proposes an advanced model, called multi-task long short-term memory (MTLSTM) neural networks, that can capture the most complicated relationships between process variables and a simultaneous fault detection and isolation approach based on MTLSTM neural networks. The proposed simultaneous fault detection and isolation approach takes a sparse autoencoder (SAE) to extract features from the measurements of process variables. Then, a long short-term memory neural network is trained on the extracted features with a multi-task learning strategy. Compared with the traditional multivariate statistical models (MVSM) such as principal component analysis and independent component analysis models, the MTLSTM neural networks can capture the nonlinear autocorrelations of process variables and correlations between process variables and perform fault detection and isolation for an industrial process simultaneously. To demonstrate the advantages and superiority of the proposed simultaneous fault detection and isolation approach, a case study on fault detection and isolation for the penicillin fermentation process is carried out. Case study results show that the proposed approach significantly outperforms existing fault detection and isolation approaches.

Hybrid deep fault detection and isolation: Combining deep neural networks and system performance models

With the increased availability of condition monitoring data on the one hand and the increased complexity of explicit system physics-based models on the other hand, the application of data-driven approaches for fault detection and isolation has recently grown. While detection accuracy of such approaches is generally very good, their performance on fault isolation often suffers from the fact that fault conditions affect a large portion of the measured signals thereby masking the fault source. To overcome this limitation, we propose a hybrid approach combining physical performance models with deep learning algorithms. Unobserved process variables are inferred with a physics-based performance model to enhance the input space of a data-driven diagnostics model. The resulting increased input space gains representation power enabling more accurate fault detection and isolation. To validate the effectiveness of the proposed method, we generate a condition monitoring dataset of an advanced gas turbine during flight conditions under healthy and four faulty operative conditions based on the Aero-Propulsion System Simulation (C-MAPSS) dynamical model. We evaluate the performance of the proposed hybrid methodology in combination with two different deep learning algorithms: deep feed forward neural networks and Variational Autoencoders, both of which demonstrate a significant improvement when applied within the hybrid fault detection and diagnostics framework. The proposed method is able to outperform pure data-driven solutions, particularly for systems with a high variability of operating conditions. It provides superior results both for fault detection as well as for fault isolation. For the fault isolation task, it overcomes the smearing effect that is commonly observed in pure data-driven approaches and enables a precise isolation of the affected signal. We also demonstrate that deep learning algorithms provide a better performance on the fault detection task compared to the traditional machine learning algorithms.

A data-driven fault isolation and estimation approach for unknown linear systems

This paper considers the data-driven fault isolation and estimation problem for linear time-invariant systems with unknown dynamic matrices and multiple actuator faults. In most of existing fault isolation methods, how to accurately identify the types of faults has not been solved well when the system matrices are unknown. To deal with this problem, a neural network-based fault isolation method is proposed by analyzing and extracting features of different fault models in terms of constructing sparse vectors and function libraries using the available input–output data. Then, a fault estimator is designed to estimate the fault signals within the data-driven framework, where its parameters are computed by the system’s Markov parameters and the identified types of faults. Finally, two examples are used to verify the advantages and effectiveness of the proposed fault isolation and estimation approach.

A Bayesian Approach for Active Fault Isolation With an Application to Leakage Localization in Water Distribution Networks

This article proposes an active fault isolation method for application to water distribution networks (WDNs) to localize leaks. The method relies on the classification of observed outputs to a discrete set of hypothetical faults. Due to parametric uncertainties, the outputs are random vectors that follow unknown probability distribution functions (PDFs). The output PDFs corresponding to the considered faults are approximated using smooth kernel density estimation (SKDE). They are used to calculate the posterior probability of each hypothesis, given the observed outputs, by applying Bayes’ rule. The difficulty to classify observed outputs to the right fault comes from the overlap between output PDFs. An active algorithm is introduced that proactively minimizes the joint overlap between the output PDFs by designing optimal control inputs. Due to physical limitations on control inputs and depending on the intensity of uncertainties, full separation, and hence fault isolation, cannot be guaranteed for a single observed output. Therefore, subsequent observations are used in an iterative framework, where the posterior probabilities of the previous time step serve as the prior probabilities for the next time step. The method is applied to locate leaks in a benchmark WDN for different levels of uncertainty in customer water demand and leakage magnitude. Improvements in the performance are observed in comparison to the best considered passive method from literature.

Signal-Model-Based Fault Diagnosis in Windings of Synchronous Generator

Any reduction in the cross-sectional area of wires (RCSAW) in the stator or rotor (field or damper) windings of synchronous generators (SGs) is considered as a fault in SGs. In this article, a signal-model-based fault detection, isolation, and estimation approach based on the eigenvalues estimation of SGs is presented to diagnose RCSAW fault. The subspace identification method is employed to estimate the state matrix eigenvalues. A form of the characteristic equation in the faulty condition is introduced for fault diagnosis of SGs in the presence of linear and nonlinear loads. The stator and field currents and voltages, and the rotor speed are used for fault diagnosis. The effectiveness of the introduced fault diagnosis method is shown using simulation and experimental results on a salient-pole SG in Iran-Safir (ER24) diesel-electric locomotives.

Real-Time Neural Classifiers for Sensor and Actuator Faults in Three-Phase Induction Motors

The main steps involved in a fault-tolerant control (FTC) scheme are the detection of failures, isolation and reconfiguration of control. Fault detection and isolation (FDI) is a topic of interest due to its importance for the controller, since it provides the necessary information to adjust and mitigate the effects of the fault. Generally, the most common failures occur in the actuator or in sensors, so this article proposes a novel model-free scheme for the detection and isolation of sensor and actuator faults of induction motors (IM). The proposed methodology performs the task of detecting and isolating faults over data streams just after the occurrence of the failure of an induction motor (IM), by the occurrence of either disconnection, degradation, failure, or connection damage. Our approach proposes deep neural networks that do not need a nominal model or generate residuals for fault detection, which makes it a useful tool. In addition, the fault-isolation approach is carried out by classifiers that differentiate characteristics independently of the other classifiers. The long short-term memory (LSTM) neural network, bidirectional LSTM, multilayer perceptron and convolutional neural network are used for this task. The proposed sensors’ and actuator’s fault detection and isolation scheme is simple. It can be applied to various problems involving fault detection and isolation schemes. The results show that deep neural networks are a powerful and versatile tool for fault detection and isolation over data streams.

An Integrated Model-Based and Data-Driven Gap Metric Method for Fault Detection and Isolation.

This article proposes an integrated approach of model-based and data-driven gap metric fault detection and isolation in a stochastic framework. For actuator and sensor faults, an adaptive Kalman filter combining with the generalized likelihood ratio method is suggested. For component faults, especially incipient faults, the model-based scheme maybe not a good choice due to the existence of disturbances or noises. Hence, a novel data-driven gap metric strategy is presented. The design of the appropriate fault cluster center model and radius via the gap metric technique is put forward to enhance the isolability of the incipient faults. Numerical simulation results are given to demonstrate the effectiveness of the proposed fault detection and isolation algorithm.

A time‐varying process monitoring approach based on iterative‐updated semi‐supervised nonnegative matrix factorizations

AbstractThis paper proposes a new time‐varying process monitoring approach based on iterative‐updated semi‐supervised nonnegative matrix factorizations (ISNMFs). ISNMFs are a type of semi‐supervised model that constructs a semi‐nonnegative matrix factorization (SNMF) model of a process using both labelled and unlabelled samples. Compared with the existing nonnegative matrix factorizations (NMFs) where NMFs are referred to as matrix factorization algorithms that factorize a nonnegative matrix into two low‐rank nonnegative matrices whose product can well approximate the original nonnegative matrix, ISNMFs have advantages in terms of the model update and the use of labelled samples. The ISNMFs‐based process monitoring approach concerns fault detection and isolation and updates an SNMF model iteratively using the latest samples to capture the change of statistical property of time‐varying processes. Moreover, the proposed fault detection and isolation approach is supported by the k‐means algorithm in theory. At last, we demonstrate the superiority of ISNMFs over the existing NMFs in terms of fault detection and isolation through a case study on the penicillin fermentation process.

A new method of diagnostic row reasoning based on trivalent residuals

The genesis of the proposed fault isolation approach was the belief that the added value may be achieved through the synthesis of various approaches. In this paper, we propose a new method of fault isolation, which promises the achievement of a favorable effect of synergy. Principally, the proposed approach merges a fault isolation system (FIS) with a modified hitting set tree (HS) inferring algorithm. As a result, a new hybrid approach based on three-valued residuals is presented. It manifests two highly desirable features of diagnostic systems, namely efficient isolability of multiple faults which is inherent for the hitting set tree approach, and a high distinguishability of faults provided by the fault isolation system. The beneficial synergy effect was indicated and confirmed in the example of the diagnosis of a two-tank system. The performed comparative study shows a significant increase in the fault distinguishability obtained with the proposed approach compared with others. The advantageous features of the proposed approach in terms of fault isolability, flexibility of applied models, and the usage of expert knowledge predestinate it for industrial implementations.

A general defect modelling and simulation-assisted approach for fault isolation in failure analysis

Over the years, defect modelling has demonstrated its effectiveness in intracell defect or root cause estimation via scan test and scan diagnosis methods. This paper describes a novel application of defect modelling in fault isolation and defect estimation in failure analysis (FA). In this article, basic defect models are first created based on failure phenomena and principles of fault isolation tools. Then, considering the corresponding failure phenomena, defect models are inserted into targeted circuits to demonstrate defect impacts via analogue simulation tools. Through this method, unreasonable defect estimations are eliminated, while the remaining estimations can cause failure; thus, defect estimations become more accurate, and further analysis becomes more targeted. To facilitate methodology execution, the analysis flow is optimized and validated in real cases. As such, visible defects are rapidly identified in three real cases, which are suitable examples to reflect the effectiveness of this method in regard to fault isolation. Moreover, further investigation of the three cases proves that how defect modelling and simulation is one powerful approach able to significantly reduce the number of defect candidates and the size of the area of interest (AOI) compared to regular approaches. In addition, via defect modelling and with the assistance of analogue simulation tools, we have an opportunity to continue to isolate defects and enhance the defect estimation accuracy through the validation of each original defect candidate under the condition of insufficient fault isolation data, or we can even meet the resolution limitation of fault isolation tools.

Data‐driven sensor fault detection and isolation of nonlinear systems: Deep neural‐network Koopman operator

This paper proposes a data-driven sensor fault detection and isolation approach for the general class of nonlinear systems. The proposed method uses deep neural network architecture to obtain an invariant set of basis functions for the Koopman operator to form a linear predictor for a nonlinear system. Then, the obtained Koopman predictor has been used in a geometric framework for sensor fault detection and isolation purposes without relying on a priori knowledge about the underlying dynamics as well as requiring faulty data, leading to a data-driven sensor fault detection and isolation framework for nonlinear systems. Finally, the approach's efficacy is demonstrated using simulation case study on a two-degree of freedom robot arm.

A New Fault Diagnosis Approach for Heavy-Duty Gas Turbines

Effective fault detection, estimation, and isolation are essential for the safety and reliability of gas turbines. In this article, a hybrid fault detection and isolation (FDI) approach is presented for condition monitoring of heavy-duty gas turbines. First, nonlinear dynamical models are constructed using an orthonormal basis function and an adaptive neuro-fuzzy inference system through experimental data. Following that, a fuzzy inference system is employed to estimate the fault severity. For this aim, fuzzy rules are extracted from the fault patterns by defining appropriate features. Later, various faults are isolated using an ensemble decision tree classifier. The proposed nonlinear modeling compensates for disturbances and uncertainty in the system and leads to adaptive thresholds for fault detection, which reduces the false alarm rate. Moreover, the proposed FDI method brings high accuracy in fault estimation by properly modeling a bounded uncertainty using the adaptive threshold. Experimental data are applied to validate the gas turbine model. Test results indicate that the proposed hybrid FDI method via adaptive threshold overwhelms the other FDI methods, where the misclassified data are 5.6%.

Data-driven fault detection and isolation of nonlinear systems using deep learning for Koopman operator

This paper proposes a data-driven actuator fault detection and isolation approach for the general class of nonlinear systems. The proposed method uses a deep neural network architecture to obtain an invariant set of basis functions for the Koopman operator to form a linear Koopman predictor for a nonlinear system. Then, the obtained linear model is used for fault detection and isolation purposes without relying on prior knowledge about the underlying dynamics. Moreover, a recursive method is proposed for fault detection and isolation that is entirely data-driven with the key feature of global validity for the system’s whole operating region due to the Koopman operator’s global characteristic. Finally, the approach’s efficacy is demonstrated using two simulations on a coupled nonlinear system and a two-link manipulator benchmark.