Abstract


A unified approach for designing and evaluating human-machine interface system is proposed which will be configured as a structured simulation framework by the combination of Multilevel Flow Model (MFM) and the various knowledge-based processing methods. By the word ¡Èhuman-machine interface system¡É, it means various operator supports such as signal validation, alarm analysis, fault detection and diagnosis, trend prediction, operation procedure presentation, automatic operation, etc., to meet with the different levels of persons and objectives in the plant, that is, to see the whole nuclear power plant as the operator in the control room, roving operators and maintainers in various parts of plant system. And several words follow for the explanation of the proposed ¡ÈStructured Simulation Framework¡É. Since MFM is a model-based semiotic analysis method for describing complicated relations and mutual interactions among the various flow behaviors of mass, energy and information in the plant or component from the aspect of goals-functions, means-ends, part-whole relations, it can be useful for semiotic understanding of the targeted plant system with its relationships to the physical systems/components, when those information are assembled on the multi-layers of sheet with special graphical representation, while putting various knowledge processing by logical and numerical algorithms for inference, reasoning, prediction, etc., in various parts and levels of abstraction of graphical MFM appropriate for the relevant human-machine interface issues. The proposed system can be basically used as a computer-aided engineering tool for both the designers and the evaluators of human-machine interface from versatile aspects of human factors, but it can be also used as the software production tool for an effective work aid to both the operator and maintainers in a real nuclear power plant. In this paper, the above-stated approach is employed for the simulation of automatic diagnosis for various types of accidental event leading to the reactor trip of Pressurized Water Reactor (PWR). A sheet of MFM graphical sheet is constructed on a personal computer on which consequence analysis is conducted by the limited and simple reasoning rules to infer either the primary or potential causes of accident by utilizing the limited sensor signals of the plant with the preset alarm levels for individual sensors. During the simulation, the result of online consequence analysis is displayed both on the MFM chart and as the list of diagnosis calculation. In the simulation work thus far conducted, 19 cases of PWR accident simulation data by the RELAP5/MOD2 code were used for 4 loop PWR plant which is simplified as two loops, i.e., intact loop and broken loop with 21 plant parameters for monitoring. In concrete, types of accident simulation including Loss-Of-Coolant-Accident (LOCA), Steam Generator Tube Rupture (SGTR) and Main Steam Line Break (MSLB), with different break size and break location were conducted to evaluate the proposed automatic diagnosis model. Although the diagnosis accuracy depends on the threshold values of sensor alarms, the result of parametric study showed that the proposed model has a good ability to detect accident well earlier in time before reactor trip. In all simulation cases, the proposed model gives right answers and the process speed on a PC is very fast to be used as a real time alarm analysis system.

Powered by JDBS Ver.2.1.0
(C) 2020 Hirotake Ishii