基于多维特征评价的风机齿轮箱早期故障诊断

Incipient fault diagnosis for wind turbine gearbox based on multidimensional feature evaluation

为了及时有效地诊断风机齿轮箱早期微弱故障,针对齿轮箱微弱故障信号非线性、非平稳、低幅值、低信噪比的特点,提出一种基于多维特征评价的风机齿轮箱早期故障诊断方法.首先,利用变分模态分解将原始振动信号分解为多个固有模态分量,并构建“信息熵-峭度-包络谱峭度”多维特征评价模型,结合熵权法筛选关键特征分量以重构信号;其次,运用改进的小波阈值法降低噪声干扰对重构信号的影响,得到显著的故障冲击特征;再者使用宽度学习系统进行状态识别,并利用L_(21)正则化技术进一步提高其网络结构的稀疏性;最后,通过对风机齿轮箱实测数据进行特征分析并与传统方法进行对比实验,表明所提出方法的有效性和优越性.

In order to timely and effectively diagnose the incipient weak faults of a wind turbine gearbox,a fault diagnosis method for a wind turbine gearbox is proposed,which deals with the nonlinear,nonstationary,low amplitude and low SNR vibration signals.Firstly,the original vibration signal is decomposed into multiple intrinsic mode functions by using the optimal variational mode decomposition.Meanwhile,an“information entropy-kurtosis-envelop spectrum kurtosis”multidimensional feature evaluation model is constructed,which is combined with the entropy weight method to screen key intrinsic mode functions to reconstruct the signal.Then an improved wavelet threshold method is designed to perform secondary noise reduction,and the obvious fault shock characteristics are obtained.The broad learning system is used for fault classification,and the L21 regularization technology is used to improve the sparsity of the network structure.By analyzing the measured data of the wind turbine gearbox and comparing with traditional methods,it is shown that the proposed method is effective and has good performance on incipient fault diagnosis.