![]() |
|||
![]()
|
![]() |
![]() |
![]() |
3.5.2. OASYS (On-Line Operator Aid System) OASYS was developed as an on-line advisory system to nuclear plant operators (Jeong et al., 1994). OASYS is an integrated system with four modules (see Figure 3):
3.5.3. Modular Real-Time Operator Advisor Expert System Another expert system developed by Hajek et al. (1989) was the Operator Advisor Expert System. The knowledge engineers working on the project identified four tasks that nuclear power plant operators carry out:
FIGURE 3 Modules in OASYS. If a fault is detected the system evaluates the fault as either a Emergency Operating Procedure (EOP), Abnormal Operating Procedure (AOP) or an Alarm Response procedure (ARP). The system then gives the steps that are needed to rectify the fault in accordance to the industry accepted Emergency Procedure Guidelines (EPGs). The systems monitors the operators' progress as the emergency procedures are executed and provides backup steps should an emergency step fail. 3.5.4. Plant Operation and Guidance System Plant automation is widely used in Japan with nearly all of the generating components such as turbines, feedwater pumps, and generators being controlled by computer. As a result of the large level of automation it is important that the nuclear plant operator receives direct and unambiguous messages about the state of the components. Expert systems are used to ensure that plant functions are performed in a safe and reliable manner. Incorporation of expert systems into the nuclear plant's operation and guidance systems is difficult to accomplish with a traditional shell. To accomplish this task Goto et al. (1990) developed a domain-specific expert system shell with the specific aims of enhancing plant safety, increasing plant reliability, enhancement of operational flexibility and efficiency, and reduced operator workload. The system uses a knowledge representation structure that was specifically designed for this problem, called Plant Tables (PT). A PT consists of a section describing plant conditions with multiple inputs and one output, and a large number of operation inputs for several plant conditions; divided into timing conditions, pre-conditions and completion conditions. Using their new expert system shell Goto et al. developed a nuclear power plant operation and guidance system. The knowledge base was constructed by using the knowledge from multiple domain experts, including nuclear plant operators, startup test engineers, automatic control design engineers and plant characteristic analysis engineers. Knowledge was also gained from analysing plant manuals, operational data, design specifications, design drawings and plant dynamic characteristics. The inference engine uses both forward and backward chaining, and checks alarms and control limits against a Monitoring Control Table. The system provides various outputs to the operator including screen, voice announcement, warning lights and graphical displays. The system has already been supplied to five Japanese boiling water reactor type plants and automates the following systems: reactor pressure control, main turbo-generator control, reactor feedwater transfer control and reactor power control. 3.5.5. The French Experience -- Expert Training System (SEPIA) Expert systems have also proved to be useful for training; particularly when incorporated within a fully functional simulator. In 1987 the French built a combined expert system and simulator (International Power Generation, 1991), for training operators to react to plant malfunctions or breakdowns and deal with them quickly before a major catastrophe occurred. One of the first problems to be modeled with the simulator was steam generator tube failure. The simulator mimics the actions and responses of the nuclear plant's main components and variations in parameters can be simulated very realistically. The results from the simulator have been checked against Electiricité De France (EDF) reference codes.
FIGURE 4 Output of sepia. Normal staff training is a two stage process. The first stage involves the operator using the simulator and experiencing various different accident scenarios, with all data from the simulator, including the operator's responses being recorded. The second stage involves a highly skilled technician manually going through the simulation replay step by step and commenting on the operator's performance. The expert system, which is called SEPIA (a French acronym for Training system using Artificial Intelligence), and the simulator run on SUN workstations. SEPIA uses artificial intelligence, an investigation model and a model of the instructions and how they should be applied. All of these systems are controlled by an expert system with a knowledge base containing over 1,000 rules and over 2,000 objects. The SEPIA systems offers a range of powerful functions:
SEPIA takes the place of a highly skilled technician instructor, and exhaustively analyses an operator's performance during a simulated exercise. SEPIA goes through the simulation step by step and generates error messages, as shown in Figure 4. 4. RESEARCH ISSUES AND FUTURE TRENDSExpert systems have made some important advances in the area of complex machine diagnostics. However, large complex machines each have their own individuality and conventional expert system technology may not be flexible enough to model the entire problem. For this reason, researchers have in recent years been experimenting with other AI technologies. The most important of these are discussed in this section.
|
![]() |
|
Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details. |