基于HVD小波包降噪编码深度学习的风电机组智能诊断研究  

A study on wind turbine intelligent diagnosis based on HVD wavelet packet de noising coding deep learning

作　　者：时培明[1] 范雅斐 伊思颖 韩东颖[2] SHI Peiming;FAN Yafei;YI Siying;HAN Dongying(School of Electrical Engineering,Yanshan University,Qinhuangdao 066004 China;School of Vehicle and Energy,Yanshan University,Qinhuangdao 066004 China)

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004 [2]燕山大学车辆与能源学院,河北秦皇岛066004

出　　处：《振动与冲击》2022年第12期196-201,共6页Journal of Vibration and Shock

基　　金：国家自然科学基金(61973262);河北省自然科学基金(E2020203147);河北省高层次人才资助项目(A201803001);河北省引进留学人员资助项目(C20190371)。

摘　　要：针对风机齿轮箱轴承振动信号非线性非平稳性对故障诊断的干扰问题,提出一种基于降噪编码器深度特征学习和希尔伯特振动分解(hilbert vibration decomposition,HVD)的智能故障诊断方法。引入峭度评估指标,对HVD分量进行模态选择,并以小波包提取分量能量熵构造特征向量,实现数据预处理。构建层叠降噪编码器(stacked denoising autoencoder,SDAE)模型完成信号的特征学习和故障分类。采用两个轴承数据集进行算法验证,试验结果表明,提出的基于HVD小波包降噪编码方法(HWSDAE)能高效地识别故障信号,具有突出的故障诊断性能,单次最高诊断准确率高达100%,平均诊断准确率可达99.49%,相比未经预处理的轴承数据输入SDAE模型提高了13.52%的故障诊断精度。An intelligent detection method based on the deep feature learning and the Hilbert vibration decomposition(HVD)of the denoising encoder was proposed to address the problem of the disturbance caused by the nonlinear and non-stationarity effect of the vibration signals of a fan gearbox bearing.The kurtosis evaluation index was introduced to evaluate the HVD component modal,and the wavelet packet was used to extract the energy entropy of the component to construct the feature vector to realize data preprocessing.A stacked denoising encoder(SDAE)model was constructed to complete signal feature learning and fault classification.Two bearing datasets were used for algorithm verification.Experimental results show that the proposed HVD wavelet packet denoising encoding method(HWSDAE)can effectively identify fault signals and has outstanding diagnostic performance.Single diagnostic accuracy is up to 100%,the average diagnostic accuracy is 99.49%,improving the diagnostic accuracy by 13.52%compared to the unprepared bearing data input into the SDAE model.

关 键 词：故障诊断 希尔伯特振动分解(HVD) 小波包分解 降噪编码器 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置] TH165.3[自动化与计算机技术—控制科学与工程]