Failure Diagnosis of Modular and Decentralized Discrete Event Systems

点击翻译按钮获取中文摘要

The increasingly stringent requirements on performance and reliability of complex technological systems have necessitated the development of systematic methods for timely and accurate diagnosis of system faults. For several years, we have been developing a methodology for centralized failure diagnosis using a discrete event systems (DES) approach. This methodology has been successfully applied in industry.We propose to build on our previous results on centralized and decentralized failure diagnosis of DES in order to develop a comprehensive methodology for failure diagnosis of large-scale complex systems with modular and distributed architectures where intermittent failures may occur. Our primary focus is the on-line diagnosis of the system, i.e., while it is operating, as opposed to off-line testing of a system when it is not under operation. Our research will include:(Pl) diagnosis of intermittent failures in the context of centralized architectures;(P2) dealing with communication delays in the context of coordinated decentralized architectures;(P3) development of new techniques for decentralized failure diagnosis with communication among local diagnostic modules at different sites; and (P4) development of new techniques exploiting the modularity of the system structure in failure diagnosis. These problems are based on feedback received from industrial applications of our methodology and aim at extending its applicability. Problems (PI) and (P2) will be directly based on our prior work on centralized and coordinated decentralized architectures for failure diagnosis. Problems (P3) and (P4) aim at extending the scope of our methodology in new directions. In Problem (P3) we intend to exploit communication among sites to enhance diagnostic capabilities in systems with decentralized information. In Problem (P4) we intend to exploit system structure to achieve computational savings in failure diagnosis of modular systems.The insight gained from our generic formal approach to the above-mentioned problems will provide the guidelines for the design of ad hoc diagnostic protocols that will have reduced complexity and still perform well. Therefore the research described in this proposal will have a significant impact on the systematic design of failure diagnosis protocols for a wide variety of complex engineering and computing systems.