Model-based Fault Diagnosis Research Articles

Fuzzy and Neural Network Approaches to Wind Turbine Fault Diagnosis

The fault diagnosis of safety critical systems such as wind turbine installations includes extremely challenging aspects that motivate the research issues considered in this paper. Therefore, this work investigates two fault diagnosis solutions that exploit the direct estimation of the faults by means of data-driven approaches. In this way, the diagnostic residuals are represented by the reconstructed faults affecting the monitored process. The proposed methodologies are based on fuzzy systems and neural networks used to estimate the nonlinear dynamic relations between the input and output measurements of the considered process and the faults. To this end, the considered prototypes are integrated with auto-regressive with exogenous input descriptions, thus making them able to approximate unknown nonlinear dynamic functions with arbitrary degree of accuracy. These residual generators are estimated from the input and output measurements acquired from a high-fidelity benchmark that simulates the healthy and the faulty behaviour of a wind turbine system. The robustness and the reliability features of the developed solutions are validated in the presence of model-reality mismatch and modelling error effects featured by the wind turbine simulator. Moreover, a hardware-in-the-loop tool is implemented for testing and comparing the performance of the developed fault diagnosis strategies in a more realistic environment and with respect to different fault diagnosis approaches. The achieved results have demonstrated the effectiveness of the developed schemes also with respect to more complex model-based and data-driven fault diagnosis methodologies.

Probabilistic model-based fault diagnosis for the cavities of the European XFEL

Abstract The European X-ray Free Electron Laser (EuXFEL) is a complex system with many interconnected components and sensor measurements. We use factor graphs to systematically design a probabilistic fault diagnosis method for its cavity system. This approach is expandable to further subsystems and considers uncertainties from measurements and modeling. After representing a model of the cavity system in the factor graph framework, we infer marginal distributions, e. g., of the fault classes using tabulated message-passing definitions. The emerging fault diagnosis method consists of an unscented Kalman filter-based residual generator and an evaluation of the residuals using a Gaussian mixture model. We include message-passing definitions for the training of the Gaussian Mixture Model from noisy data using the expectation-maximization algorithm.

Planetary gearbox fault diagnosis using bidirectional-convolutional LSTM networks

Gearbox fault diagnosis is expected to significantly improve the reliability, safety and efficiency of power transmission systems. However, planetary gearbox fault diagnosis remains a challenge due to complex responses caused by multiple planetary gears. Model-based gearbox fault diagnosis techniques extract hand-crafted features from sensor data based on underlying physics and statistical analysis, which are not effective in extracting spatial and temporal features automatically. While deep learning methods such as convolutional neural network (CNN) enable automatic feature extraction from multiple sensor sources, they are not capable of extracting spatial and temporal features simultaneously without losing critical feature information. To address this issue, we introduce a novel deep neural network based on bidirectional-convolutional long short-term memory (BiConvLSTM) networks to determine the type, location, and direction of planetary gearbox faults by extracting spatial and temporal features from both vibration and rotational speed measurements automatically and simultaneously. In particular, a CNN determines spatial correlations between two measurements within one time step automatically by combining signals collected from three accelerometers and one tachometer. Long short-term memory (LSTM) networks identify temporal dependencies between two adjacent time steps. By replacing input-to-state and state-to-state operations in the LSTM cell with convolutional operations, the BiConvLSTM can learn spatial correlations and temporal dependencies without losing critical features. Experimental results have shown that the BiConvLSTM network can detect the type, location, and direction of gearbox faults with higher accuracy than conventional deep learning approaches such as CNN, LSTM, and CNN-LSTM.

Model-free fault diagnosis for autonomous underwater vehicles using sequence Convolutional Neural Network

The AUV must be capable of fault diagnosis if it is to perform tasks in complex environments without human assistance. However, the current fault diagnosis methods for AUV lack of manual experience and accuracy, leading to the lack of fault handling capacity. Different from the traditional model-based fault diagnosis, we propose a new model-free fault diagnosis method characterized by a deep learning-based algorithm, which is a new Sequence Convolutional Neural Network (SeqCNN) that learns the patterns between state data and fault type. More specifically, the proposed SeqCNN aims to extract global feature and local feature from state data and classify the extracted information into different fault types, and can convert two-stage diagnosis mode into a single-stage one. Compared to the traditional model-based diagnosis, it can significantly reduce the time-consuming burden, simplify the diagnosis procedure and improve the efficiency. The effectiveness of SeqCNN was validated by a practical experiment on a small quadrotor AUV ‘Haizhe’. The results indicate that the proposed SeqCNN can solve the problem of fault detection and fault isolation in single-stage diagnosis mode and that its accuracy is far superior to that of other deep learning diagnosis algorithms.

Fault Diagnosis via Neural Ordinary Differential Equations

Implementation of model-based fault diagnosis systems can be a difficult task due to the complex dynamics of most systems, an appealing alternative to avoiding modeling is to use machine learning-based techniques for which the implementation is more affordable nowadays. However, the latter approach often requires extensive data processing. In this paper, a hybrid approach using recent developments in neural ordinary differential equations is proposed. This approach enables us to combine a natural deep learning technique with an estimated model of the system, making the training simpler and more efficient. For evaluation of this methodology, a nonlinear benchmark system is used by simulation of faults in actuators, sensors, and process. Simulation results show that the proposed methodology requires less processing for the training in comparison with conventional machine learning approaches since the data-set is directly taken from the measurements and inputs. Furthermore, since the model used in the essay is only a structural approximation of the plant; no advanced modeling is required. This approach can also alleviate some pitfalls of training data-series, such as complicated data augmentation methodologies and the necessity for big amounts of data.

Fault diagnosis and prognosis of steer-by-wire system based on finite state machine and extreme learning machine

In this paper, an integrated condition monitoring method combining model-based fault diagnosis and data-driven prognosis is proposed for steer-by-wire (SBW) system. First, the SBW system is modeled by bond graph (BG) technique and a two-degree-of-freedom (2-DOF) state-space model of the vehicle is built. Based on the 2-DOF model, the estimated self-aligning torque is used for the control of feedback motor. The fault detection is carried out by evaluating the analytical redundancy relations derived from the BG model. Since the fault isolation performance is essential to subsequent fault estimation process, a new fault isolation method based on finite state machine is developed to improve the isolation ability by combining the dependent and independent analytical redundancy relations, where the number of potential faults could be decreased. In order to refine the possible fault set to determine the true fault, a cuckoo search (CS)–particle filter is developed for fault estimation. Based on the estimated true fault, prognosis can be implemented which is important to achieve failure prevention and prolong system lifespan. To this end, an optimized extreme learning machine (OELM) is proposed where the input weights and hidden layer biases are optimized by CS. Based on data representing fault values obtained from the fault identification, the OELM model is trained for remaining useful life prediction of failing component. Finally, the proposed methodologies are validated by simulations.

Fault Isolation and Identification of a Four-Single-Gimbal Control Moment Gyro On-board a 3-axis Stabilized Satellite

Control moment gyros are known for their applications in attitude stabilization. These actuators are susceptible to malfunction, which results in faults and failures. Therefore, diagnosing the faults can improve the reliability of completing a mission while reducing maintenance costs. Thus, a model-based fault diagnosis method is proposed here. The intended algorithm is an enhanced version of previous work by the author. The enhancement employs a condensed approach to alleviate the delay caused by the filter’s confidence in its estimations. A case-study on a closed-loop controlled satellite is provided along with an extensive Monte Carlo simulation to evaluate the proposed method’s performance. The results show that the enhanced method can achieve superior performance while requiring less computational resources by eliminating extra grid search loops.

An Intelligent Prediction for Detecting Bearing Vibration Characteristics Using a Machine Learning Model

Abstract Bearing failure in the heavy rotating machines results in shut down of many other machines and affects the overall cost and quality of the product. Condition monitoring of bearing systems acts as a preventive and corrective measure as it avoids breakdown and saves maintenance time and cost. This research paper proposes advanced strategies for early detection and analysis of taper rolling bearings. In view of this, mathematical model-based fault diagnosis and support vector machining (SVM) are proposed in this work. A mathematical model using dimension analysis by the matrix method (Dimension Analysis Method (DAMM)) and SVM is developed that can be used to predict the vibration characteristic of the rotor-bearing system. Types of defects are created using electrical discharge machining (EDM) and analyzed, and correlation is established between dependent and independent parameters. Experiments were performed to evaluate the rotor dynamic characteristic of healthy and unhealthy bearings. Experimental results are used to validate the model obtained by the DAMM and SVM. Experimental results showed that the vibration characteristic could be evaluated by using a theoretical model and SVM. Efforts have been made to extend the service life of the machines and the assembly lines and to improve their efficiency, so as to reduce bearing failure; what provides novelty to these efforts is the use of four machine learning techniques. Thus, an automatic online diagnosis of bearing faults has been made possible with the developed model based on DAMM and SVM.

A Model-Based Sensor Fault Diagnosis Scheme for Batteries in Electric Vehicles

The implementation of each function of a battery management system (BMS) depends on sensor data. Efficient sensor fault diagnosis is essential to the durability and safety of battery systems. In this paper, a model-based sensor fault diagnosis scheme and fault-tolerant control strategy for a voltage sensor and a current sensor are proposed with recursive least-square (RLS) and unscented Kalman filter (UKF) algorithms. The fault diagnosis scheme uses an open-circuit voltage residual generator and a capacity residual generator to generate multiple residuals. In view of the different applicable state of charge (SOC) intervals of each residual, different residuals need to be selected according to the different SOC intervals to evaluate whether a sensor fault occurs during residual evaluation. The fault values of the voltage and current sensors are derived in detail based on the open-circuit voltage residual and the capacity residual, respectively, and applied to the fault-tolerant control of battery parameters and state estimations. The performance of the proposed approaches is demonstrated and evaluated by simulations with MATLAB and experimental studies with a commercial lithium-ion battery cell.

Sliding Mode Control with Application to Fault-Tolerant Control: Assessment and Open Problems

This paper is prepared within a collaboration between the Instituto Politécnico Nacional, which is a Mexican research institute that manages research on sliding-mode control theory, and the ARIA research team of the Intégration du Matériau au Système Lab., a French research group that engages research on model-based fault diagnosis and fault-tolerant control theories. The paper reviews the application of sliding mode control techniques to fault tolerant control and provides perspectives leading to posing some open problems. Operating principles, definitions of the basic concepts are recalled along with the control objectives and design procedures. The evolution of the sliding mode control technique through five generations (as classified by Fridman, Moreno and co-workers) is reviewed. Their respective design procedures, limitations, and robustness properties are also highlighted. The application of the five generations of sliding-mode controllers to fault-tolerant control is discussed. The focus is on some open problems that are judged to commonly be overlooked. Some applications in real-world systems are also presented.

Fault diagnosis of a nonlinear hybrid system using adaptive unscented Kalman filter bank

In this paper, a model-based fault diagnosis scheme of a nonlinear hybrid system using an adaptive unscented Kalman filter (AUKF) bank is proposed. The hybrid system is an amalgamation of discrete dynamics and continuous states. Fault diagnosis for simultaneous occurrences of multiple faults such as leakage fault, clogging fault, sensor fault, and actuator fault on a benchmark three-tank system are simulated. The residual signal based output generates some discrete modes that guarantee the uniqueness of the concerning fault. The efficacy of the proposed scheme is compared with that of the adaptive extended Kalman filter (AEKF) bank on the same system to prove its better response over AEKF.

Dynamic System Fault Diagnosis Under Sparseness Assumption

Dynamic system fault diagnosis is often faced with a large number of possible faults. The purpose of this paper is to propose an efficient method for such situations. To avoid intractable combinatorial problems, sparse estimation techniques appear to be a powerful tool for isolating faults, under the assumption that only a small number of possible faults can be simultaneously active. However, sparse estimation is often studied in the framework of linear algebraic equations , whereas model-based fault diagnosis is usually investigated for dynamic systems modeled with state equations involving internal states . The main contribution of this paper is a link between these two formalisms through efficient and reliable algorithms, mainly relying on advanced analyses of residuals generated with the Kalman and Kitanidis filters. Based on these results, it becomes straightforward to solve fault diagnosis problems by applying well known sparse estimation techniques, in the framework of general time varying state-space systems involving unknown inputs.

Complementary-Integrated Approach to Model-Based and Data-Driven Prognostics and Fault Diagnosis for Reusable Rocket Engine Systems

The cost reduction of engine maintenance is an important issue for a future reusable launch vehicle. Toward reliable operation of a reusable rocket, a novel health monitoring method using the phase plane trajectory of feature extracted sensor data by principal component analysis (PCA) is verified. Because there are few failure data to estimate new method, a simple physical model is constructed to produce two kinds of failure data duct trouble and reservoir leakage, and examine the effectiveness of this new health monitoring method. Results indicate that presented phase plane trajectory method has good visibility to make the change clear and make it easy to detect and identify failures. This new approach will contribute to the improvement of health monitoring technologies based on sensor data not only for the reusable rocket engine systems but also for general systems.

A novel matrix-vector-based framework for modeling and simulation of electric vehicle battery packs

Abstract Facing the growing electrification of the transport sector, there is an increasing need for models capturing the entire vehicle battery. The performance of the model is crucial for the development of energy storage systems but also for real-time applications. The challenge of these so-called control-oriented models is to precisely map the individual cell behavior and the interactions with the neighboring cells while keeping the computational effort low. In this work, a novel framework for the electro-thermal modeling of large-scale battery packs is presented. Both the individual cell behavior and the interactions between neighboring cells as well as cyclic aging are taken into account. For the parameterization, two cell types are characterized over a temperature range from 15∘C to 45∘C and over the entire state of charge range. The model is validated based on three fault cases in the battery system, which are mapped correctly with the model framework. From the results, fault-specific findings are derived for model-based and date-driven fault diagnosis. A battery system consisting of 500 cells is simulated to evaluate the model performance. The simulation of this battery pack is real-time capable up to a frequency of 100Hz.

Active Model-Based Fault Diagnosis in Reconfigurable Battery Systems

With the increasing demand for electric vehicles, the interest in battery systems is growing. In order to enable safe operation of these complex energy storage systems, methods of fault diagnosis are needed. Particularly, reconfigurable battery systems (RBSs) with switches are promising on the way to fault tolerance as they allow the system to be reconfigured in the event of a fault. In this article, a model-based fault diagnosis algorithm is developed and validated that uses the switches of an RBS to improve the fault isolability. Since the algorithm changes the structure of the system in order to differentiate between nonisolable faults, it is classified as an active fault diagnosis algorithm. The deviations between sensor measurements and model, called residuals, are stochastically analyzed. For fault isolation, a fuzzy clustering approach is used. A constrained sigma-point Kalman filter minimizes model uncertainties and therefore increases the sensitivity and robustness of the fault diagnosis approach. Furthermore, the filter allows estimating the fault amplitude in case of a fault. Based on active sequential hypothesis testing, a policy to calculate the next switch position is proposed and investigated. It is shown simulatively and experimentally that additional faults are isolated by the presented active approach.

Fault Diagnosis and Prognosis of a Brushless DC motor using a Model-based Approach

This paper proposes a model-based fault diagnosis and prognosis approach applied to brushless DC motors (BLDC). The objective is an early detection of mechanical and electrical faults in BLDC motors operating under a variety of operating conditions. The proposed model-based method is based on the evaluation of a set of residuals that are computed taking into account analytical redundancy relations. Fault diagnosis consist of two steps: First, checking if at least one of the residuals is inconsistent with the normal operation of the system. And, second, evaluating the set of the residuals that are inconsistent to determine which fault is present in the system. Fault prognosis consists of the same two steps but instead of considering current inconsistencies evaluates drift deviations from nominal operation to predict futures residual inconsistencies and therefore predict future fault detections and diagnosis. A description of various kinds of mechanical and electrical faults that can occur in a BLDC motor is presented. The performance of the proposed method is illustrated through simulation experiments.

The Post-Fault Current Model of Voltage Source Converter and Its Application in Fault Diagnosis

The open-switch fault of voltage source converters (VSC) may change the switching state and distort the current, which is the basis for model-based fault diagnosis methods. Based on the analysis of the switching state, this letter presents a post-fault current model to estimate the current of each phase after one open-switch fault occurs. Combined with the current model-based method, the proposed post-fault model can be used not only to check the accuracy of the diagnosis result but also to detect multiple open-switch faults. The current, voltage, and driving signals used in this model are all from the VSC control system, therefore, this model can be easily implemented without additional sensors or hardware. Finally, the effectiveness of the proposed model is verified by the experimental results.

Fault Diagnosis of an Autonomous Vehicle With an Improved SVM Algorithm Subject to Unbalanced Datasets

Safety is one of the key requirements for automated vehicles and fault diagnosis is an effective technique to enhance the vehicle safety. The model-based fault diagnosis method models the fault into the system model and estimates the faults by observer. In this article, to avoid the complexity of designing observer, we investigate the problem of steering actuator fault diagnosis for automated vehicles based on the approach of model-based support vector machine (SVM) classification. The system model is utilized to generate the residual signal as the training data and the data-based algorithm of the SVM classification is employed to diagnose the fault. Due to the phenomena of data unbalance induced poor performance of the data-driven method, an undersampling procedure with the approach of linear discriminant analysis and a threshold adjustment using the algorithm of grey wolf optimizer are proposed to modify and improve the performance of classification and fault diagnosis. Various comparisons are carried out based on widely used datasets. The comparison results show that the proposed algorithm has superiority on the classification over existing methods. Experimental results and comparisons of an automated vehicle illustrate the effectiveness of the proposed algorithm on the steering actuator fault diagnosis.

Robust Fault Diagnosis for Adaptive Structures With Unknown Stochastic Disturbances

Adaptive structures can react to environmental impacts and can significantly improve the load-bearing behavior of, e.g., building structures. For building structures, reliability and safety are major concerns, and thus, a fault diagnosis system, detecting and isolating actuator and sensor faults, is necessary. Adaptive building structures are exposed and adapt to unknown, stochastic wind and surface loads that are difficult to model and can therefore not be considered in model-based fault diagnosis methods, which decreases their performance. For this reason, this article combines the model-based method of parity equation with the data-based method of principal component analysis (PCA) for the fault diagnosis in adaptive structures. The proposed approach characterizes the unknown disturbance in the residual data derived by parity equations using PCA. PCA determines orthogonal directions of decreasing variance in the residual which are decoupled to decrease the sensitivity of the residual to disturbances. Since decoupling of residual directions influences the diagnosability of faults, the number of decoupled principal components is determined such that the diagnosability is not significantly affected. The diagnosability is analyzed by the Kullback–Leibler divergence based on the system model and disturbance characterization. The resulting decoupled residual significantly improves the fault detection performance using a cumulative sum algorithm for change detection. Moreover, the fault input matrix of the system model and the knowledge of the fault time profile yield reliable fault isolation using a Bayes classifier as illustrated in a simulation study.

Aeroengine Control System Sensor Fault Diagnosis Based on CWT and CNN

The aeroengine control system is a piece of complex thermal machinery which works under high-speed, high-load, and high-temperature environmental conditions over lengthy periods of time; it must be designed for the utmost reliability and safety to function effectively. The consequences of sensor faults are often extremely serious. The inherent complexity of the engine structure creates difficulty in establishing accurate mathematical models for the model-based sensor fault diagnosis. This paper proposes an intelligent fault diagnosis method for aeroengine sensors combining a deep learning algorithm with time-frequency analysis wherein the signal recognition problem is transformed into an image recognition problem. The continuous wavelet transform (CWT) is first applied to seven common health condition signals in an engine control system sensor in order to generate scalograms that capture the characteristics of the signal. A convolutional neural network (CNN) model trained with preprocessed and labeled datasets is then used to extract the features of a time-frequency graph based on which faults can be identified and isolated. This method does not require modeling and design thresholds, so it has strong robustness and accuracy rate of over 97%. The trained model effectively reveals faults in sensor signals and allows for accurate identification of fault types.