Development of Techniques for Signal Validation in Nuclear Power Plants

信号验证是一种交叉检查不同传感器输出的技术，以获得每个关键重要物理参数的单一明确指示。该技术的关键问题是获取不同信号之间的因果关系知识，并将这些知识组织成智能知识库。以下基本技术组件已开发，以满足需求. 1.获取因果关系的技术。2.结构化的信号验证知识库.按需信号验证咨询系统。本文对传统的定性模拟方法进行了改进，引入了模糊逻辑，为满足1。这种方法使我们能够以更精确的方式表征信号之间的因果关系。正确使用这种方法，我们可以获得知识，用于交叉检查不同的信号。这样获得的知识应该与关于植物结构的知识相关，以进行有效的信号验证。为了以有组织的方式存储这两种类型的知识，开发了一个能够处理框架表示和产生式规则的知识库。知识库的内容既可以预先存储，也可以在发现预先存储的数据不足时生成。以知识库为主要组成部分，综合并实现了一个按需信号确认咨询系统原型。所开发的系统的实用性，通过模拟实验证明。

Signal Validation is a technique for cross-checking outputs of dissimilar sensors to obtain a single unambiguos indication for each physical parameter of critical importance. The Key issues in this technique are to acquire knowledge on cause-effect relationships between dissimilar signals, and to organize the knowledge in a form of intelligent knowledgebase. Following basic technical components have been developed to meet the need.1. Technique for acquiring causal relationships.2. Structurized knowledgebase for signal validation.3. Consultation system for on-demand signal validation. The method of conventional qualitative simulationw as modified by introducing the fuzzy logic to provide a technique for satisfying the need of 1. This method allows us to characterize the causal relationships between signals in more precise manner. With the proper use of this method, we can obtain knowledge to be employed for cross-checking of dissimilar signals. The knowledge thus acquired should be related to the knowledge about plant structure for efficient signal validation. To store the two types of knowledge in an organized manner, a knowledgebase capable of handling both of the frame-representation and production rules was developed. The contents of the knowledgebase can be either prestored or generated whenever the prestored data are found to be insufficient. By using the knowledgebase as the main component, a protype of consultation system for on-demand signal validation was synthesized and implemented on a personal comuter. The usefulness of the developed system was demonstrated through simulator experiments.