基于小样本数据挖掘的空气处理机组传感器隐形故障诊断方法  

Invisible Fault Diagnosis Method of Air Handling Unit Sensor Based on Small Sample Data Mining

作　　者：李彤洋 赵何超 胡真齐 杜伊帆 刘光宇 陈羽飞 Li Tongyang;Zhao Hechao;Hu Zhenqi;Du Yifan;Liu Guangyu;Chen Yufei(China Railway l2th Bureau Group Co.,Ltd.,TaiYuan 030000,China;School of Building Services Science and Engineering,Xi'an University of Architecture and Technology,Xi'an 710055,China;College of Information&Control Engineering.Xi'an University of Architecture and Technology,Xi'an 710055,China)

机构地区：[1]中铁十二局集团第二工程有限公司,太原030000 [2]西安建筑科技大学建筑设备科学与工程学院,西安710055 [3]西安建筑科技大学信息与控制工程学院,西安710055

出　　处：《绿色建造与智能建筑》2023年第7期53-58,共6页GREEN CONSTRUCTION AND INTELLIGENT BUILDING

基　　金：陕西省自然科学基金,2021JQ-516。

摘　　要：针对空气处理机组系统内传感器隐形故障诊断准确率不足及虚警率较高等问题,本文提出了基于集成经验模态分解软阈值去噪法与RBF神经网络(EEMDSTD-RBF)相结合的故障诊断方法,通过集成经验模态分解软阐值去噪法对采集的原始数据进行数据去噪,将去噪后的正常运行数据以及隐形故障数据输入RBF神经网络进行训练,得到EEMDSTD-RBF故障诊断模型,从而提高RBF神经网络对于空气处理机组传感器隐形故障的诊断准确率,同时,实验证明,该方法有效降低了虚警率。在关于空气处理机组六个传感器点位在三种隐形故障等级下的偏置故障的实验中,结果表明,与RBF及EEMDSTD-BP方法相比较,该方法在不同故障状态下,EEMDSTD-RBF的诊断准确率提高了8.7%~64.7%,虚警率降低了12.62%。In this paper,we propose a fault diagnosis method based on the combination of the integrated empirical mode decomposition soft threshold denoising method and RBF neural network(EEMDSTD-RBF),which denoises the collected raw data by the integrated empirical mode decomposition soft threshold denoising method,and the denoised normal operation data and invisible fault data are input into the RBF neural network for training.and invisible fault data into the RBF neural network for training to obtain the EEMDSTD-RBF fault diagnosis model,so as to improve the diagnosis accuracy of the RBF neural network for invisible faults of air handling unit sensors,and at the same time,the experiment proves that the method effectively reduces the false alarm rate.In the experiments about the bias faults of six sensor points of air handling units under three invisible fault levels,the results show that compared with theRBF andEEMDSTD-BPmethods,the method improves thediagnosis accuracy of EEMDSTD-RBF by 8.7%~64.7%and reduces the false alarm rateby12.62%under different fault states.

关 键 词：空气处理机组 传感器 隐形故障诊断 RBF神经网络 集成经验模态分解软阈值去噪法 

分 类 号：TB657.2[一般工业技术—制冷工程]