基于小波神经网络的空调传感器故障诊断方法
A Fault Diagnosis Method for Air Conditioning Sensors Based on Wavelet Neural Network
曹华1
作者信息
摘要
当前空调传感器故障诊断节点设置一般为独立目标式,诊断的覆盖范围较小,导致传感器故障诊断单元耗时延长.为此提出对基于小波神经网络的空调传感器故障诊断方法的设计与验证分析.根据当前测定,先进行故障特征提取,采用多阶的方式,扩大诊断的覆盖范围,部署多阶的诊断节点,以此为基础,设计小波神经网络空调传感器故障诊断模型,采用局部重叠逼近的方式来实现故障诊断处理.测试结果表明:对比于传统基于t-SNE降维和KNN算法空调传感器故障诊断测试组、传统MMD迁移学习空调传感器故障诊断测试组,此次所设计的小波神经网络空调传感器故障诊断测试组最终得出的传感器故障诊断单元耗时均被控制在0.25s以下,说明在小波神经网络技术的辅助下,当前对于传感器故障诊断的效率明显提升,针对性增大,误差可控,具有实际的应用价值.
Abstract
Currently,the setting of fault diagnosis nodes for air conditioning sensors is generally independent target type,with a small diagnostic coverage,resulting in an extended time consumption for sensor fault diagnosis units.This paper proposes the design and validation analysis of a fault diagnosis method for air conditioning sensors based on wavelet neural networks.Based on the current measurement,fault feature extraction is carried out first,and a multi-stage approach is adopted to expand the coverage of diagnosis.Multi stage diagnostic nodes are deployed.Based on this,a wavelet neural network fault diagnosis model for air conditioning sensors is designed and designed,using local overlap approximation to achieve fault diagnosis processing.The test results show that compared to the traditional t-SNE dimensionality reduction KNN algorithm for air conditioning sensor fault diagnosis test group and the traditional MMD transfer learning air conditioning sensor fault diagnosis test group,the time consumption of the sensor fault diagnosis unit ultimately obtained by the designed wavelet neural network air conditioning sensor fault diagnosis test group is controlled below 0.25 seconds,indicating that with the assistance of wavelet neural network technology,the current efficiency of sensor fault diagnosis has significantly improved,with increased specificity and controllable errors,which has practical application value.
关键词
小波技术/神经网络/空调传感器/故障识别/故障诊断/诊断方法Key words
wavelet technology/neural network/air conditioning sensor/fault identification/fault diagnosis/diagnostic methods引用本文复制引用
出版年
2023
NETL
NSTL
维普
万方数据