基于CEEMDAN- EFICA去噪的风电齿轮箱故障诊断

针对风电齿轮箱实验样本较少,以及振动信号具有非平稳、非线性的特点,提出了基于完备集合经验模态分解(CEEMDAN)-高效快速独立分量分析(EFICA)的去噪方法。首先应用CEEMDAN与峭度-相关系数准则完成信号重构,对重构信号和原信号进行EFICA分离来获得去噪信号;然后提取去噪信号的时域特征、频域特征构建特征向量,使用核主分量分析(KPCA)对向量降维处理实现特征信息融合;最后采用复合神经网络对信号特征集进行分类完成故障诊断。通过实验数据对比,证明了该方法消噪效果更好且复合神经网络的诊断准确率最高,所提方法具有可行性和优越性。

Fault Diagnosis of Wind turbine Gearbox Based on CEEMDAN-EFICA Denoising

Aiming at the low experimental samples of wind turbine gearbox and the non-stationary and nonlinear characteristics of vibration signals,a denoising method based on complete set empirical mode decomposition(CEEMDAN)-efficient fast independent component analysis(EFICA)is proposed.Firstly,CEEMDAN and kurtosis-correlation coefficient criteria are used to reconstruct the signal,and the reconstructed signal and the original signal are separated by EFICA to obtain the denoised signal.Then the time domain and the frequency domain features of the denoised signal are extracted to construct the feature vector,using Kernel Principal Component Analysis(KPCA)to reduce vector dimension for implementing feature fusion.Finally,composite neural network is used to classify signal feature sets to complete fault diagnosis.The experimental data comparison shows that the method has better denoising effect and the diagnostic accuracy of composite neural network is the highest,the proposed method is feasible and superior.