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,提出一种并行感知器神经网络,通过模式分类来尝试联系系统故障状态与热力参数特征向量之间的映射关系.对故障模拟实验结果的应用表明,基于并行感知器网络的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 was selected and measured to form the characteristic vector that represents the malfunction state of a refrigeration system. A type of parallel perceptron network was introduced to accomplish the pattern classification task of “fault symptom-fault source” relationship mapping. The fault simulation experiments on a real air-source heat pump indicate that this technique is feasible and effective. This work can be used as the research base for the development of fault diagnosis system based on neural network framework for refrigeration system.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2005年第8期1233-1239,共7页
Journal of Shanghai Jiaotong University
基金
国家重点基础研究发展规划(973)项目(G2000026309)
关键词
制冷系统
故障诊断
感知器
模式分类
refrigeration system
fault diagnosis
perceptron
pattern classification
作者简介
刘相艳(1981-),女,山东临沂市人,硕士生,现主要从事翅片管换热器换热性能和压降、空调装置故障诊断研究。
谷波(联系人),男,教授,电话(Tel.):021-62933240;E-mail:gubo@sjtu.edu.cn。