Fault Detection In Large-scale Systems Research Articles

Distributed Fault Detection for Large-Scale Systems: A Subspace Intersection-Based Scheme to Accurately Monitor the Impact of Fault Propagation

Distributed Fault Detection for Large-Scale Systems: A Subspace Intersection-Based Scheme to Accurately Monitor the Impact of Fault Propagation

Distributed data-driven optimal fault detection for large-scale systems

This paper is concentrated on two new distributed data-driven optimal fault detection approaches in large-scale systems using a group of sensor blocks, each of which accesses part of the process variables. Towards this end, an optimal fault detection problem is first formulated and solved, which lays a foundation for further distributed studies. Based on it, the first distributed data-driven optimal fault detection scheme, consisting of offline distributed learning and online distributed detection, is developed using the average consensus algorithm. To further reduce communication and computation efforts, the second average consensus based fault detection is investigated. Considering that the iteration computations for average consensus algorithm can lead to fault detection delay, a variation of the average consensus based fault detection scheme is proposed with iterative estimation of the covariance matrices of random variables and implementation of the distributed test statistic during the consensus iteration. A numerical example and a case study on the PRONTO heterogeneous benchmark dataset are used to demonstrate the proposed approaches.

Multiple Hotelling’s T2 tests for distributed fault detection of large-scale systems

The distributed configuration is an effective approach for implementing Hotelling's T2 test to detect faults in large-scale systems. Multiple T2 tests are employed in the distributed configuration in which each test uses measurements from a different set of sensors. A decomposition (i.e. an allocation of the sensors among different sets called subsystems) must first be selected to implement the multiple T2 tests. The detection performance of the multiple T2 tests and their ability to satisfy certain operational constraints depend on the decomposition. We propose two methods that find a feasible decomposition of a large-scale system for which the detection performance of the multiple T2 tests is near optimal. The methods first use greedy algorithms to generate many feasible candidate subsystems and decompositions. The methods then quickly evaluate the detection performance of these feasible candidates using easy to compute analytic expressions and finally select a feasible decomposition with the best performance.

Distributed fault detection in large-scale systems using hybrid extended information filter

A fault monitoring system plays an important role to ensure and improve the reliability of an industrial plant to operate safely and efficiently. For large-scale systems, their high-dimensional nature impels the designer to develop a lower-order algorithm to overcome the practical limitations. To reduce the order of computations and the amount of communication, the system should be decomposed into low-dimensional sub-systems. For this purpose, a distributed fault detection algorithm was presented based on hybrid extended information filter. This algorithm has the advantages of two existing approaches, the hybrid extended Kalman filter and the information filter. In the proposed method, several local fault detectors were employed to monitor the local sub-system instead of a single centralized monitoring node. For each detector, a limited number of measurements were accessible based on the generated local state and residual vectors using a local filter. The local detector extracted the faults’ information content of the residual signals via CUSUM statistical hypothesis testing. Moreover, it could communicate the processed information of its neighbors if needed. The algorithm was implemented on alkylation of benzene process plant. The results obtained from the distributed algorithm showed a considerable decrease in computational burden and communication bandwidth.

A Distributed Networked Approach for Fault Detection of Large-Scale Systems

Networked systems present some key new challenges in the development of fault-diagnosis architectures. This paper proposes a novel distributed networked fault detection methodology for large-scale interconnected systems. The proposed formulation incorporates a synchronization methodology with a filtering approach in order to reduce the effect of measurement noise and time delays on the fault detection performance. The proposed approach allows the monitoring of multirate systems, where asynchronous and delayed measurements are available. This is achieved through the development of a virtual sensor scheme with a model-based resynchronization algorithm and a delay compensation strategy for distributed fault-diagnostic units. The monitoring architecture exploits an adaptive approximator with learning capabilities for handling uncertainties in the interconnection dynamics. A consensus-based estimator with time-varying weights is introduced, for improving fault detectability in the case of variables shared among more than one subsystem. Furthermore, time-varying threshold functions are designed to prevent false-positive alarms. Analytical fault detectability sufficient conditions are derived, and extensive simulation results are presented to illustrate the effectiveness of the distributed fault detection technique.

Optimal Topology for Distributed Fault Detection of Large-scale Systems

The paper deals with the problem of defining the optimal topology for a distributed fault detection architecture for non-linear large-scale systems. A stochastic modelbased framework for diagnosis is formulated. The system structural graph is decomposed into subsystems and each subsystem is monitored by one local diagnoser. It is shown that overlapping of subsystems allows to improve the detectability properties of the monitoring architecture. Based on this theoretical result, an optimal decomposition design method is proposed, able to define the minimum number of detection units needed to guarantee the detectability of certain faults while minimizing the communication costs subject to some computation cost constraints. An algorithmic procedure is presented to solve the proposed optimal decomposition problem. Preliminary simulation results show the potential of the proposed approach.

Decentralised fault detection of large-scale systems with limited network communications

An optimal decentralised fault detection (FD) system is proposed for a class of discrete-time large-scale systems, where limited information exchange among subsystems is enabled with communication networks. Such an FD system consists of two parts in each subsystem: a residual generator (RG) and a residual evaluator. By taking the advantages of network communications, the RGs are designed to match a proper reference model in a stochastic sense such that a best trade-off between the robustness against disturbances and the sensitivity to faults can be achieved. In each residual evaluator, a norm-based evaluation function is selected for evaluating residual signals and a corresponding threshold is calculated for detecting the occurrences of faults. All the solutions are presented in the form of linear matrix inequalities. Finally, a numerical example is given to illustrate the results and the improvement achieved by applying network communications.

Sensor Location Strategy in Large-Scale Systems for Fault Detection Applications

Fault detection in large-scale systems is conducted by the use of sensors, thus the sensor location influences the performances of fault detection directly. As the scale of systems increases, traditional input-output models may not work well or may even not be applicable. The Signed Directed Graph (SDG) model is used to describe large-scale complex systems and the cause-effect relationships among variables. However, SDG cannot express the dynamic propagation properties when describing the fault propagation phenomena. In this paper, time parameters are taken into account within the branches of an SDG, in order to approximately denote the propagation time of the variable changes in the systems. An SDG constructed this way is called a dynamic SDG. As to sensor location, because of the economic and technical limitations, the number of sensors should be limited while meeting the demands of fault detection. We analyzed the main criteria of the fundamental demand such as reachability, detectability and identifiability of faults in the framework of dynamic SDG, and presented an algorithm to describe the fault propagation process using forward inference and obtained a way to locate sensors. These indices guarantee that the faults can be detected in time and identified from each other. The criteria change when the faults propagate; they are much closer to the reality than those in the framework of a traditional SDG. Some general results, which extend the results obtained in the framework of a traditional SDG, were presented. An algorithm was presented to describe the fault propagation process using forward inference and a way to locate the sensors was obtained. Finally, two examples were used to illustrate the proposed method and results.