Comparison Between Statistical and Observer-Based Approaches for Fault Detection and Isolation in a Chemical Process

A combined data‐driven and observer‐design methodology for fault detection and isolation (FDI) in hybrid process systems with switching operating modes is proposed. The main contribution is to construct a unified framework for FDI by integrating Gaussian mixture models (GMM), subspace model identification (SMI), and results from unknown input observer (UIO) theory. Initially, a GMM is built to identify and describe the multimodality of hybrid systems using the recorded input/output process data. A state‐space model is then obtained for each specific operating mode based on SMI if the system matrices are unknown. An UIO is designed to estimate the system states robustly, based on which the fault detection is laid out through a multivariate analysis of the residuals. Finally, by designing a set of unknown input matrices for specific fault scenarios, fault isolation is performed through the disturbance‐decoupling principle from the UIO theory. A significant benefit of the developed framework is to overcome some of the limitations associated with individual model‐based and data‐based approaches in dealing with the problem of FDI in hybrid systems. Finally, the validity and effectiveness of the proposed monitoring framework are demonstrated using a numerical example, a simulated continuous stirred tank heater process, and the Tennessee Eastman benchmark process. © 2014 American Institute of Chemical Engineers AIChE J, 60: 2805–2814, 2014

This paper presents a Fault Detection and Isolation (FDI) method for stochastic nonlinear dynamic systems. Our contribution consists of showing another method of tackling the problem of the physical origin diagnosis of faults by combining the technique based on the innovations and the technique using the multiple Kalman filters for a nonlinear dynamic system strongly non-stationary. The usefulness of this combination is the implementation of all the faults dynamics if the decision threshold on the standardized innovation exceeds a fixed value. In the other case, one filter is enough to estimate the process state. An algorithm is described and applied to a perfectly stirred chemical reactor functioning in a semi-batch mode. In this paper, the chemical reaction used is an oxido reduction one, the oxidation of sodium thiosulfate by hydrogen peroxide.

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.

ABSTRACTQuick and timely fault detection is of great importance in control systems reliability. Undetected faulty sensors could result in irreparable damages. Although fault detection and isolation (FDI) methods in control systems have received much attention in the last decade, these techniques have not been applied for some classes of nonlinear systems yet. This paper deals with the issues of sensor fault detection and isolation for a class of Lipschitz uncertain nonlinear system. By introducing a coordinate transformation matrix for states and output, the original system is first divided into two subsystems. The first subsystem is affected by uncertainty and disturbance. The second subsystem just has sensor faults. The nonlinear term is separated to linear and pure nonlinear parts. For fault detection, two sliding mode observers (SMO) are designed for the two subsystems. The stability condition is obtained based on the Lyapunov approach. The necessary matrices and parameters are obtained by solving the linear matrix inequality (LMI) problem. Furthermore, two sliding mode observers are designed for fault isolation. Finally, the effectiveness of the proposed approach is illustrated by simulation examples.

This paper deals with the fault detection, diagnosis and control problems of time delay systems by designing a disturbance observer (DOB). In proposed method, the DOB estimates the changes occurred as an input disturbance. In fact, the DOB is implemented as a residual generator for the system. Subsequently, the estimated changes can be used to compensate the effect of faults and plant alterations directly. In this method, the estimated disturbances can be used for the FDI purpose and the final system would be fault tolerant, without the need to have all other faults isolable and identifiable. It provides a simple, low cost and effective fault detection and isolation method which can be applied to more practical systems. The applicability and effectiveness of the proposed method is illustrated by the simulations provided for an industrial water pressure control system which is supposed to be subject to faults.

This work presents two fault detection and isolation (FDI) approaches for wind turbine systems (WTS). Firstly, a non-linear mathematical model for wind turbine (WT) dynamics is developed. Based on the developed WTS mathematical model, a robust fault detection observer is designed to estimate system faults, so as to generate residuals. The observer is designed to be robust to system disturbance and sensitive to system faults. A WT blade pitch system fault, a drive-train system gearbox fault and three sensor faults are simulated to the nominal system model, and the designed observer is then to detect these faults when the system is subjected to disturbance. The simulation results showed that the simulated faults are successfully detected. In addition, a neural network (NN) method is proposed for WTS fault detection and isolation. Two radial basis function (RBF) networks are employed in this method. The first NN is used to generate the residual from system input/output data. A second NN is used as a classifier to isolate the faults. The classifier is trained to achieve the following target: the output are all “0”s for no fault case; while the output is “1” if the corresponding fault occurs. The performance of the developed neural network FDI method was evaluated using the simulated three sensor faults. The simulation results demonstrated these faults are successfully detected and isolated by the NN classifier.

A robust Principal Component Analysis (PCA) model-based approach to Fault Detection and Isolation (FDI) was developed for the Helical Coil Steam Generator (HCSG) systems of the International Reactor Innovative and Secure (IRIS) nuclear reactor. The PCA model was developed using the data generated from a simulation of the system dynamics. Because all the operation modes can be excited through well-designed simulations, a PCA model developed from simulation data is more robust to changes in operation conditions than a PCA model developed from historical data collected during routine operation conditions. In order to deal with model uncertainty that exists in a simulation model, the mismatch between PCA model predictions and online plant measurements was analyzed to characterize the model uncertainty. Based on the assumption that model uncertainty has a structured property, a complete algorithm for the PCA-based FDI was derived such that the FDI results are robust to model uncertainty. When this new method was applied to the IRIS HCSG systems, the results showed that this approach could avoid false alarms and fault misdiagnosis due to changes in operation conditions and model certainty. The approach would overcome the limitations of traditional data-driven model based FDI when routine operational data do not contain sufficient information, and the limitations of physical model based FDI when these models are too complicated for a direct use in FDI design and contain model uncertainty.

Linguistic OWA and two time-windows based fault identification in wide plants

This chapter discusses the definition of a detection observer, its structural conditions and the solution of their solution. A method of designing a bank of detection observers to give maximise the fault isolation capability, is described. The design approach is guided by a fault isolation logic which is not necessarily associated with a particular fault detection method, but can be also be used with other fault detection methods. It is thought that this fault isolation logic may serve as a general guide for the design of fault detection schemes based upon quantitative models and not solely for the design of the type of fault detection observers under discussion.

SummaryThis paper proposes an integrated actuator and sensor active fault‐tolerant model predictive control scheme. In this scheme, fault detection is implemented by using a set‐valued observer, fault isolation (FI) is performed by set manipulations, and fault‐tolerant control is carried out through the design of a robust model predictive control law. In this paper, a set‐valued observer is used to passively complete the fault detection task, while FI is actively performed by making use of the constraint‐handling capability of robust model predictive control. The set‐valued observer is chosen to implement fault detection and isolation (FDI) because of its simple mathematical structure that is not affected by the type of faults such as sensor, actuator, and system‐structural faults. This means that only one set‐valued observer is needed to monitor all considered actuator and sensor statuses (health and fault) and to carry out the fault detection and isolation task instead of using a bank of observers (each observer matching a health/fault status). Furthermore, in the proposed scheme, the advantage of robust model predictive control is that it can effectively deal with system constraints, disturbances, and noises and allow to implement an active FI strategy, which can improve FI sensitivity when compared with the passive FI methods. Finally, a case study based on the well‐known two‐tank system is used to illustrate the effectiveness of the proposed fault‐tolerant model predictive control scheme. Copyright © 2016 John Wiley & Sons, Ltd.

In this paper, it is shown that the high automation level of the object-oriented modeling paradigm for physical systems can significantly rationalize the design procedure of fault detection and isolation (FDI) systems. Consequently, an object-oriented FDI method for complex engineering systems consisting of subsystems from different physical domains like mechatronic systems, commercial vehicles, and chemical process plants is developed. The mathematical composition of the objects corresponding to the subsystems results in a differential algebraic equation (DAE) that describes the overall system. This DAE is automatically analyzed and transferred into a set of residual generators that enable a two-stage FDI procedure for multiple fault modes.

In this paper, a diagnosis method for time-delay fractional systems is developed. The fault detection and isolation are ensured by the use of a bank of unknown input fractional observers. Theoretical development leads to the obtention of the observer parameters in order to engender structured residuals. A numerical example illustrating the validity of the method in fault detection and isolation is given.

Fault detection in continuous industrial processes is crucial to ensure production, quality, safety, environmental and economic aspects. Reliable fault detection systems have become even more necessary due to increasingly stringent regulations. The complexity of processes, on the one hand, and the availability of data, on the other, have spread the adoption of the data-driven approach for industrial fault detection. Deep Reinforcement Learning (RL) has recently been applied for this purpose, mainly in rotating machinery and rolling bearings. In chemical processes, most RL applications are in process control. The main contribution of this work lies in proposing a reinforcement learning (RL)-based framework for fault detection in continuous chemical processes. The challenges are related to the large number of variables, nonlinear relationships and changing operational dynamics, to name a few. The classic Tennessee Eastman Process (TEP) benchmark was used as a case study. The results were compared with Principal Component Analysis (PCA), commonly used for chemical process monitoring, dynamic PCA (DPCA), and neural networks. Compared to PCA, the RL approach showed considerably lower missed detection rate (MDR) and time to detection (TTD) while maintaining reasonable false alarm rate (FAR) for all hard-to-detect faults, contributing to early fault detectability. Similar results were obtained with regard to DPCA. Compared to the supervised approach, RL proved to be more flexible regarding model parameterization, which can be very useful for practical aspects of industrial fault detection. The results show that RL-based approaches can be very useful for fault detection in continuous, non-linear, and time-varying systems, such as chemical processes. For example, with regard to adaptive alarm thresholds and the detection of known and unknown faults. The latter represents a major challenge in industrial practice.

Inter-turn short-circuit (ITSC) faults on the induction machine has received much attention in the recent years. Early detection of such faults in wind turbine systems would allow to avoid fluctuation on wind power output and maintain the reliability level. In this paper, Sliding Mode Observers (SMO)-based fault detection and isolation method is developed for induction generator (IG)-based variable-speed grid-connected wind turbines. Firstly, the dynamic model of the wind turbine and IG was given and then, the control was made based on Maximum Power Point Tracking (MPPT) method. The IG closed-loop via Indirect Rotor Flux Oriented Control (IRFOC) scheme was also described. Hence, the performance of the wind turbine system and the stability of injected power to the grid were analyzed under the ITSC fault conditions. The control schemes were proved to be inherently unstable under the faulty conditions. Then, robust SMO were investigated to design an ITSC fault detection and isolation scheme. Finally, simulation results of ITSC detection and isolation in the variable-speed grid-connected wind turbine with affected IG confirm the theoretical development.

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.