Real-time Model Based Diagnosis Of GasTurbines

Abstract

With the increasing complexity of problems and applications being tackled by automatic monitoring and diagnosis systems, it is becoming essential to tackle different aspects of the problem with different paradigms. Some applications are complex now, but the most optimum tool for each aspect of the problem must be well integrated with other optimum solutions for other aspects of the problem, in order to achieve more robust and complex systems. For an interesting class of problems a single monitoring and diagnosis paradigm is no longer adequate. This paper describes the work of integrating various reasoning paradigms being conducted in the ESPRIT project, TIGER, for gas turbine monitoring and diagnosis. In this work we combine real-time rulcbased diagnosis techniques with real-time situation assessment techniques and qualitative simulation model based diagnosis. These techniques are integrated using a task architecture. The work is being applied to an industrial gas turbine at Exxon Chemical in Scotland, and an Auxiliary Power Unit turbine at Dassault Aviation in France. INTRODUCTION A gas turbine is an essential part of the modem process and power generation system. Powerful gas turbines are often the primary driver of a power station or for major process plants. Because of their high cost they are normally not deployed in a redundant manner, and hence their availability is critical. In addition, with the increasing complexity of these systems and the gradual spread of their use, the gap between a typical engineer's understanding of the working of the turbine and its real performance is increasing (See Figure 1). A final important aspect is that maintenance costs of such critical items of equipment are very high. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517 266 Artificial Intelligence in Engineering Many companies would like to move from a calendar or running hour based maintenance strategy to a condition based system. In order to achieve this, powerful monitoring and diagnosis is needed to identify whether the turbine is working properly or problems are developing [9]. The TIGER project (ESPRIT III Project: 6862) is developing a real-time situation assessment system for gas turbines. This combines real-time artificial intelligence based diagnosis techniques using more traditional rulcbased approaches with qualitative models and simulation of the turbine behaviour. The role of each of these are described in turn (See Figure 2).