基于小波神经网络的多联式空调机组传感器故障检测和诊断  

Fault detection and diagnosis of sensors in multi-connected air-conditioning unit based on wavelet neural network

作　　者：黄荣庚 陈焕新[1] 李冠男[1] 李绍斌 胡文举[3] 

机构地区：[1]华中科技大学 [2]珠海格力电器有限公司 [3]北京建筑大学,供热供燃气通风及空调工程北京市重点实验室

出　　处：《制冷与空调》2018年第2期77-82,94,共7页Refrigeration and Air-Conditioning

基　　金：国家自然科学基金项目(51576074;51328602);供热供燃气通风及空调工程北京市重点实验室研究基金资助课题(NR2013K02)

摘　　要：针对多联式空调机组的结构复杂性以及传感器系统的非线性等特点,将小波神经网络用于对多联式空调机组传感器的故障检测与诊断。从实际生产环境中收集多联式空调机组运行数据,构建合适的小波神经网络,并针对常见的4种主要软故障进行诊断与预测,以验证其故障诊断能力,同时比较小波神经网络对于不同故障的敏感性。结果表明,小波神经网络对于多联式空调机组传感器故障检测和诊断具有良好的精度,对于偏差、漂移和失效的检测效果尤为明显。According to the structural complexity of multi-connected air-conditioning unit and the nonlinearity of sensor system,the wavelet neural network(WNN)is proposed for the fault detection and diagnosis(FDD)of the sensors.Based on the data collected from practical processing,an appropriate WNN is constructed,and employed to detect and diagnose four common and typical soft fault of sensors,so as to verify its property for FDD.In the meanwhile,its sensitivities to different faults are compared.The results show that the WNN shows a well accuracy on FDD of the sensors for the multi-connected air-conditioning unit,especially on the fault of biases,drifting and complete failure.

关 键 词：多联式空调机组 小波神经网络 小波变换 传感器故障 故障检测与诊断 

分 类 号：TB657.2[一般工业技术—制冷工程] TP183[自动化与计算机技术—控制理论与控制工程]