Author 's accepted manuscript, published in Control Engineering Practice 16 (2008)
AbstractFault diagnosis methods based on process history data have been studied widely in recent years, and several successful industrial applications have been reported. Improved data validation has resulted in more stable processes and better quality of the products. In this paper, an on-line fault detection and isolation system consisting of a combination of principal component analysis (PCA) and two neural networks (NNs), radial basis function network (RBFN) and self-organizing map (SOM), is presented. The system detects and isolates faulty operation of the analyzers in an ethylene cracking furnace. The test results with real-time process data are presented and discussed.
This paper presents a novel dynamic causal digraph reasoning method for fault diagnosis and its application to the short circulation process of a paper machine. In order to improve the fault detection ability of the original causal digraph method, a residual modification approach that takes into account the direction of different fault effects is presented. An improvement of the isolation capability of the original method, an inference mechanism between the arcs of the graph, is also proposed to locate process faults on the arcs. The results from the application show that the proposed method, compared to the conventional method, is able to detect the correct nodes and to identify the responsible arcs when the system is affected by a process fault.
The work presented in this paper addresses the issues of fault detection and isolation properties of the partial PCA method and the isolation-enhanced PCA method. In order to increase the sensitivity of the residuals with respect to various faults, the structured residuals generated from both partial PCA and isolation-enhanced PCA are optimized. For the residual evaluation, the bootstrap technique is combined with the CUSUM method to achieve fast and robust detection. Three sensor faults and three actuator faults were studied using simulations employing a rigorous, first principles based, paper machine simulator. All the faults were correctly detected and isolated with both studied methods, and the results are compared with the classical T 2 and SPE contribution plot methods.
Abstract:The aim of the work presented in this paper is to evaluate the ability of the causal digraph method to detect and isolate faults on a simulated paper machine process. A causal digraph model for the short circulation process of the paper machine was constructed, identified and used to detect and isolate artificial faults in the simulation environment. The fault of headbox slice opening was studied and diagnosed.
The aim of the work presented in this paper is to assess the ability of support vector machines (SVM) for detecting measurement faults. Two different support vector machine approaches for detecting faults are tested and compared to neural networks. The first method is based on a SVM regression model together with an analysis of the residuals whereas the second method is based on a SVM classifier. The methods were applied to a rigorous first principles based dynamic simulator of a dearomatization process.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.