基于概率神经网络的制冷空调系统故障诊断分析  

Analysis on Fault Diagnosis for Refrigeration and Air-Conditioning System Based on Probabilistic Neural Network

作　　者：卞荷洁[1] 谷波[1] 黎远光[1] 

机构地区：[1]上海交通大学制冷与低温工程研究所,上海200030

出　　处：《上海交通大学学报》2004年第10期1613-1616,1622,共5页Journal of Shanghai Jiaotong University

基　　金：国家重点基础研究发展规划(973)项目资助(G2000026309)

摘　　要：选择热力参数集组成反映制冷系统故障状态的特征向量,提出了利用概率神经网络通过模式分类来联系系统故障状态与热力参数特征向量之间的映射关系.对实际试验结果的应用尝试表明,该诊断方法可行且有效,为开发以人工神经网络为框架的制冷系统故障诊断系统提供了研究基础.A set of thermal parameters are selected and measured to form the characteristic vector that represents the malfunction state of refrigeration and air-conditioning system. Probabilistic neural network is introduced to accomplish the pattern recognition task of fault symptom-fault source relationship mapping. The fault simulation and diagnosis experiments on a real air-source heat pump indicate that this approach is feasible and effective. This work can be the research base for the development of fault diagnosis system based on neural network framework for refrigeration and air-conditioning system.

关 键 词：制冷空调系统 故障诊断 模式分类 概率神经网络 

分 类 号：TK14[动力工程及工程热物理—热能工程]