基于知识矩阵的核动力装置运行故障诊断方法  

Fault diagnosis algorithm of nuclear power plant based on knowledge matrix

作　　者：王天舒 余刃[1] 毛伟 彭俏[1] WANG Tian-shu;YU Ren;MAO Wei;PENG Qiao(College of Nuclear Science and Technology,Naval Univ.of Engineering,Wuhan 430033,China)

机构地区：[1]海军工程大学核科学技术学院,武汉430033

出　　处：《海军工程大学学报》2022年第6期73-78,83,共7页Journal of Naval University of Engineering

摘　　要：针对核动力装置运行故障诊断知识规则数量庞大、推理过程复杂,使得采用传统方法构建的专家系统结构复杂、推理效率较低等问题,首先提出了故障诊断专家知识的元命题分解策略,设计了由状态向量、转移触发向量、输入变换矩阵和输出变换矩阵构成的知识矩阵,给出了各自的定义和生成规则,形成了标准化的专家知识数学建模方法体系;然后,设计了基于知识矩阵的并行推理算法,将复杂的逻辑推理过程转换为简单的矩阵运算,可有效提高专家系统的推理效率,满足诊断过程的实时性要求,使专家系统具有良好的实用性和可扩展性;最后,利用加压核动力装置模拟器运行数据作为运行故障诊断专家系统的输入,以一回路主冷却剂系统破损故障的实时诊断为例,对该方法的正确性和有效性进行了验证。In view of the large number of rules and complex reasoning process of nuclear power plant operation fault diagnosis knowledge,which makes the structure of expert system constructed by traditional method complex and with low reasoning efficiency,a meta-proposition decomposition strategy for expert knowledge of fault diagnosis was proposed,the knowledge matrix composed of state vector,transition trigger vector,input transformation matrix and output transformation matrix was designed,as well as their definitions and generation rules.By this way a standardized numerical mode-ling method system of expert knowledge was formed.The parallel reasoning algorithm based on knowledge matrix was designed to transform the complex logical reasoning process into simple matrix computation,which could effectively improve the reasoning efficiency of the expert system,meet the real-time requirement of the diagnosis process,and make the expert system have good practicability and expansibility.Using the operation data from a pressurized nuclear power plant simulator as the input of the designed operation fault real-time diagnosis expert system,the correctness and effectiveness of this method were verified by an example of primary coolant system breakage fault diagnosis.

关 键 词：核动力装置 故障诊断 知识矩阵 并行推理 

分 类 号：TL383[核科学技术—核技术及应用]