论文部分内容阅读
针对双谱分析在应用于机械设备故障诊断过程中面临的问题,提出了含有稀疏度约束的非负张量分解算法及基于此的二次故障特征提取方法.首先,改进已有的非负张量分解算法,加入稀疏度控制策略;其次,将机械振动信号的双谱图像堆叠为一个三阶张量;然后利用改进后的分解算法对该张量进行二次故障特征提取,得到代表局部特征的“基图像”;最后,通过计算得出基图像在构成原双谱图像中所占的权重,并将得到的权重向量用于故障分类.将该方法应用于齿轮箱故障诊断的结果表明,从齿轮箱振动信号的双谱中提取出来的二次特征不仅能够反映出系统中存在的一些非线性特征,而且二次特征与故障特征频率之间有直观的对应关系,从而为解释齿轮箱故障与对应双谱之间的关系提供了很大的方便.
Aiming at the problem of bispectral analysis applied in mechanical equipment fault diagnosis, a nonnegative tensor decomposition algorithm with sparsity constraint and the method of quadratic fault feature extraction are proposed.Firstly, the existing nonnegative Zhang Secondly, the bi-spectral images of the mechanical vibration signal are stacked into a third-order tensor. Then, the second-order fault feature of the tensor is extracted by using the improved decomposition algorithm to obtain the representative local features Finally, the weight of the base image in the original bispectrum image is calculated and the weight vector is used in the fault classification.The method is applied to the result of gearbox fault diagnosis It shows that the quadratic feature extracted from the bispectrum of the vibration signal of the gearbox can reflect not only some nonlinear features existing in the system but also an intuitive correspondence between the quadratic feature and the fault feature frequency, Box fault and the corresponding bispectrum provides a great deal of convenience.