Robust detection and isolation of process faults using neural networks

The present paper suggests neural approaches to observer-based schemes, in order to perform a robust diagnosis of process faults. The symptoms are generated by using dynamic neural networks with mixed structure. The residuals are then classified by means of static artificial nets. Application to a laboratory process is included. It refers to component and instrument fault detection and isolation in a three-tank system.

Process fault detection and diagnosis using neural networks—I. steady-state processes

This paper introduces a new nonlinear filter and deep learning-based Fault Detection and Isolation (FDI) approach for gas turbines. The system is first identified using a nonlinear feedforward network, including Finite Impulse Response (FIR) filters and Long Short-Term Memory (LSTM) networks. Following that, adaptive thresholds are devoted to generating residuals, and a Convolutional Neural Network (CNN) is employed for fault isolation. Utilizing two deep neural network frameworks ensures robust and accurate system identification under uncertainties and significantly improves fault detection and isolation processes compared to conventional methods. The proposed approach is validated on a heavy-duty gas turbine model. Test results show enhanced fault detection accuracy with reduced false alarms and improved fault isolation efficiency by correctly distinguishing fault effects in the system.

This work presents a framework for fault detection and diagnosis of multimode industrial processes. It is shown that the proposed method to generate features for the faults provides a generalization that can allow the fault diagnosis in modes for which no training was performed. The proposed method is applied to a sophisticated simulator of a continuous stirred tank reactor (CSTR). Experimental results suggest the usefulness of this approach for the detection of faults under operating conditions not considered during training phase.

Fault detection and diagnosis for non-linear processes empowered by dynamic neural networks

In this paper, a new robust fault detection and isolation (FDI) methodology for an unmanned aerial vehicle (UAV) is proposed. The fault diagnosis scheme is constructed based on observer-based techniques according to fault models corresponding to each component (actuator, sensor, and structure). The proposed fault diagnosis method takes advantage of the structural perturbation of the UAV model due to the icing (the main structural fault in aircraft), sensor, and actuator faults to reduce the error of observers that are used in the FDI module in addition to distinguishing among faults in different components. Moreover, the accuracy of the FDI module is increased by considering the structural perturbation of the UAV linear model due to wind disturbances which is the major environmental disturbance affecting an aircraft. Our envisaged FDI strategy is capable of diagnosing recurrent faults through properly designed residuals with different responses to different types of faults. Simulation results are provided to illustrate and demonstrate the effectiveness of our proposed FDI approach due to faults in sensors, actuators, and structural components of unmanned aerial vehicles.

Deep model integrated with data correlation analysis for multiple intermittent faults diagnosis

In this paper a neural network based fault detection and isolation concept for supervision of technical processes is introduced. Three different possibilities to employ neural networks for fault diagnosis are discussed. In most existing schemes the steps of residual generation and residual evaluation are performed on the basis of analytical process knowledge. Here neural networks are proposed for these tasks. First of all a neural network can be used instead of a mathematical model for residual generation, secondly another neural network can be trained to perform the classification task for residual evaluation and therefore fault isolation. A third possibility is a one-step diagnosis (OSD), where one neural network is trained to directly detect and isolate possible faults from the available measurements without the need for prior generation of intermediate signals as residuals. Results from the application of a restricted Coulomb energy neural network (RCE) to the residual evaluation and alternatively to the one-step diagnosis at an industrial actuator benchmark problem are presented.

An independent radial basis function neural network (RBFNN) is developed and employed here for an online diagnosis of actuator and sensor faults. In this research, a robust fault detection and isolation scheme is developed for an open-loop exothermic semi-batch polymerization reactor described by Chylla–Haase. The independent RBFNN is employed here for online diagnosis of faults when the system is subjected to system uncertainties and disturbances. Two different techniques to employ RBFNNs are investigated. Firstly, an independent neural network (NN) is used to model the reactor dynamics and generate residuals. Secondly, an additional RBFNN is developed as a classifier to isolate faults from the generated residuals. Three sensor faults and one actuator fault are simulated on the reactor. Moreover, many practical disturbances and system uncertainties, such as monomer feed rate, fouling factor, impurity factor, ambient temperature and measurement noise, are modelled. The simulation results are presented to illustrate the effectiveness and robustness of the proposed method.

In a chiller plant, primary or critical sensors are used to control the system operation while secondary sensors are installed to monitor the performance/health of individual equipment. Current sensor fault detection and diagnosis (SFDD) approaches are not applicable to secondary sensors which usually are not involved in the system control. Consequently, a hybrid multiple sensor fault detection, diagnosis and reconstruction (HMSFDDR) algorithm for chiller plants was developed. Machine learning and pattern recognition were used to predict the primary sensor faults through the comparison of the weekly performance curves. With the primary sensor signals reconstructed, the secondary sensor faults were estimated based on mass and energy balance. By applying the algorithm with various logged plant data and comparison with site checking results, a maximum of 75% effectiveness could be achieved. The merits of the present approach were further justified through off-site sensor testing which reinforced the usefulness of proposed HMSFDDR algorithm.

Fault diagnosis of rotating machines is an important task to prevent machinery downtime, and provide verifiable support for condition-based maintenance (CBM) decision-making. Deep learning-enabled fault diagnosis operations have become increasingly popular because features are extracted and selected automatically. However, it is challenging for these models to give superior results with rotating machine components of different scales, single and multiple faults across different rotating components, diverse operating speeds, and diverse load conditions. To address these challenges, this paper proposes a comprehensive learning approach with optimized signal processing transforms for single as well as multiple faults diagnosis across dissimilar rotating machine components: gearbox, bearing, and shaft. The optimized bicoherence, spectral kurtosis and cyclic spectral coherence feature spaces, and deep blending ensemble learning are explored for multiple faults diagnosis of these components. The performance analysis of the proposed approach has been demonstrated through a single joint training of the entire framework on a compound dataset containing multiple faults derived from three public repositories. A comparison with the state-of-the-art approaches that used these datasets, shows that our method gives improved results with different components and faults with nominal retraining.

Aiming at solving the multiple fault diagnosis problem as well as the sequence of all the potential multiple faults simultaneously, a new multiple fault diagnosis method based on the dependency model method as well as the knowledge in test results and the prior probability of each fault type is proposed. Firstly, the dependency model of the system can be built and used to formulate the fault-test dependency matrix. Then, the dependency matrix is simplified according to the knowledge in the test results of the system. After that, the logic ‘OR’ operation is performed on the feature vectors of the fault status in the simplified dependency matrix to formulate the multiple fault dependency matrix. Finally, fault diagnosis is based on the multiple fault dependency matrix and the ranking of each fault type calculated basing on the prior probability of each fault status. An illustrative numerical example and a case study are presented to verify the effectiveness and superiority of the proposed method.

Neural network observer based on fuzzy auxiliary sliding-mode-control for nonlinear systems

The Sixth International Symposium on Neural Networks (ISNN 2009)

This article addresses a robust fault detection and isolation procedure which is applicable for a hybrid system undergoing discrete mode changes. Global analytical redundancy relations, which may undergo change in their structure depending on the switched mode, are derived from the diagnostic hybrid bond graph model of the system. For fault detection and isolation analysis, the residuals, evaluated at every instant from the global analytical redundancy relations, are tested against adaptive thresholds which are changed with mode shift and variation of states, but not changed due to any fault. The parametric uncertainties are included in the diagnostic hybrid bond graph through the linear fractional transformation approach and the adaptive thresholds are obtained from the modified diagnostic hybrid bond graph model. The residuals are made robust in a sense that these would always remain within the upper and lower threshold bound for any perturbation except fault occurrence. For application of the robust fault detection and isolation analysis, a hybrid thermo-fluid system is selected whose behavioural evolution combines both discrete as well as continuous changes of states. It is shown through simulation that with parameter uncertainties in a hybrid system subject to frequent mode shift, robust fault detection and isolation is successfully achieved using the passive approach. The novelty of this work lies in the integration of the model-based diagnosis principle for hybrid systems with the approach for robust residual threshold generation for systems having uncertain parameters.