神经元网络专家系统及其在核反应堆事故诊断中的应用

The Neural Network Expert Systems and Their Application in Nuclear Reactor Fault Diagnosis

神经元网络是一种不同于目前人工智能研究中逻辑模型的新的表达方式。我们在本文中将其应用到一个核反应堆事故诊断专家系统RODES中,提出了神经元网络启发式的概念。神经元网络启发式的并行、分布特征和可学习性为专家系统的知识表达和获取、不确定性推理提供了新的途径。它在核反应堆事故诊断专家系统中的实践表明了在人工智能中的潜在前景。

Neural network is a new kind of representation model which differs from the current logical model. In the paper, we apply it to RODES -an expert system for diagnosing nuclear reactor fault and propose the idea of neural net heuristics. The parallelism, distribution and capability of learning of neural net heuristics provide a new way for knowledge acquisition, knowledge representation and uncertainty reasoning in expert systems. The practice in nuclear reactor fault diagnosis shows its potential power in the research of artificial intelligence.