基于自适应小波特征提取一体化神经网络的空调电机振动噪声识别  

INTEGRATED NEURAL NETWORK FUSED WITH ADAPTIVE WAVELET FEATURE EXTRATION TO IDENTIFY THE VIBRATION AND NOISE OF ELECTROMOTER IN AIR-CONDITIONER

作　　者：赵学智[1] 叶邦彦[1] 陈统坚[1] 

机构地区：[1]华南理工大学机械工程学院,广州510640

出　　处：《振动与冲击》2007年第12期160-165,共6页Journal of Vibration and Shock

基　　金：国家自然科学基金资助项目(50305005)

摘　　要：正确识别空调电机的噪声类型是改善其噪声效果的重要前提,采用一种集特征提取与识别于一体的神经网络来解决这种识别问题。此网络利用Mexican hat小波作为母小波,同时将基于小波变换的特征提取过程融入为神经网络的一部分,网络学习时可针对输入信号对小波尺度和平移参数进行自适应调整,以实现对信号特征信息的充分获取。给出了此网络的学习算法。利用这一网络对空调电机的三种噪声信号即电磁噪声、不平衡噪声、轴承噪声信号进行了学习和识别,结果表明,学习后的网络以很高的可靠性准确地识别出了电机的不同噪声类型。To correctly identify the noise type of electromotor in air-conditioner is an important premise for taking measures to decrease its noise. One kind of neural network in which feature extraction and identification process are integrated is proposed to finish this task. Mexican hat wavelet being used as mother wavelet, the process of signal's feature extraction based on wavelet transform is fused into one part of the neural network. Moreover, wavelet's scale and shift parameters can be adaptively adjusted to fit input signal during network's learning course so that signal's feature information could be fully extracted. The network's learning algorithm is given and the network is used to identify the three types of noise signals of electromotor in air-conditioner, namely electromagnetic noise, unbalanced rotor noise and injuring bearing noise. The results demonstrate that the trained neural network can discern the different noise signals successfully with high reliability.

关 键 词：电机噪声 自适应小波 一体化神经网络 噪声识别 

分 类 号：TP274.3[自动化与计算机技术—检测技术与自动化装置] TM30[自动化与计算机技术—控制科学与工程]