基于GMM均值超向量稀疏分解的碰摩声发射源识别  

Rubbing Acoustic Emission Recognition Based on GMM Mean Supervector Sparse Decomposition

作　　者：殷智浩[1] 沈舷[2] YIN Zhihao;SHEN Xian(School of Information and Electrical Engineering,College of Industrial Technology,Xuzhou Jiangsu 221140,China;Changzhou Textile Garment Institute,Changzhou Jiangsu 213164,China)

机构地区：[1]徐州工业职业技术学院信息与电气工程学院,江苏徐州221140 [2]常州纺织服装职业技术学院机电工程系,江苏常州213002

出　　处：《电子器件》2018年第4期928-935,共8页Chinese Journal of Electron Devices

基　　金：国家自然科学基金项目(61301295);江苏省自然科学基金(ZKJ201202)。

摘　　要：研究提出了一种基于GMM均值超向量稀疏分解的风电机组碰摩声发射源识别检测方法。以压缩感知理论与GMM识别模型为基础,将高斯分量的均值向量连接组成GMM均值超向量作为压缩感知的稀疏基,并利用MP、OMP算法进行稀疏分解,根据稀疏系数识别风电机组碰摩声发射源,而不需要对信号进行重建,提高了识别性能和效率。通过实验观察上位机输出结果,验证了识别结果数据的实时性和准确性,达到了对风电机组运行的状态信息监测和故障诊断的要求。A new method for the identification and detection of the acoustic emission source of the wind turbine generator based on the sparse decomposition of the GMM mean is proposed.Based on compressed sensing theory and GMM model,the Gauss mean vectors are used to generate GMM mean supervectors,which are sparse basis of compressed sensing theory.Then,the sparse decomposition is done based on MP and OMP algorithm.According to the sparse coefficient,the wind turbine rubbing acoustic emission source is identified without the signal reconstruction.This method improves the recognition performance and efficiency.The experimental results verifies the real-time performance and accuracy.And,the algorithm can meet the requirements of the state information monitoring and fault diagnosis of wind turbine operation.

关 键 词：风电机组 声发射 声源识别 实时监测 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]