基于GMM/ANN混合网络的碰摩声发射源识别  

Rubbing Acoustic Emission Recognition Based on GMM/ANN Hybrid Network

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1期164-171,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项目(ZKJ201202)

摘　　要：研究提出了一种基于GMM/ANN混合网络模型的风电机组碰摩声发射源识别检测方法。以人工神经网络理论与高斯混合识别模型为基础,充分利用高斯混合模型在识别中良好的数据分布能力和人工神经网络具有的较强的学习能力,在人工神经网络中嵌入高斯混合模型来进行声发射识别,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。To identify and detect the acoustic emission source of the wind turbine generator,A new algorithm based on GMM/ANN hybrid network is proposed.Based on the theory of artificial neural network and Gauss hybrid model,GMM(Gaussian Mixture Model)with good ability of data distribution in recognition is embedded in ANN(Artificial Neural Network)with strong ability of learning to recognize acoustic emission.The experimental results verifies data of the real-time performance and accuracy.And,the algorithm meets the requirements of the state information monitoring and fault diagnosis of wind turbine operation.

关 键 词：声发射 声源识别 高斯混合模型 人工神经网络 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]