内燃机优化VMD-CWD时频表征与BSNMF编码识别诊断方法

针对内燃机振动响应信号强耦合、弱故障特征的问题,提出一种基于参数优化VMD-CWD内燃机振动时频表征与BSNMF分块编码识别的故障诊断方法.利用变分模态分解(VMD)将内燃机振动信号分解成一组本征模态函数(IMF),并叠加IMF分量信号的Choi-Williams分布(CWD)获得时频聚集性良好,无交叉项干扰的振动谱图像.针对VMD分解过程中的参数选取问题,引入功率谱熵作为目标函数,对VMD的分解参数进行网格寻优,提高了VMD分解的自适应性.为了实现内燃机振动谱图像的自动识别诊断,在稀疏非负矩阵分解(SNMF)的基础上提出一种更容易收敛的分块稀疏非负矩阵分解算法(BSNMF),用来对内燃机振动谱图进行特征提取,并采用支持向量机对提取的特征参数直接进行模式识别.将本文方法应用于内燃机故障诊断实例中,结果表明:该方法能有效提取内燃机振动信号中的微弱故障特征,实现内燃机气门机构故障的自动诊断.

Parameter optimized VMD-CWD time-frequency representation and BSNMF identification diagnosis method of internal combustion engine

Aiming at the problem of vibration response signal of internal combustion engine featuring the strong coupling and weak fault,a fault diagnosis method based on vibaration time-frequency feature of parameter optimization VMD-CWD internal combustion engine and BSNMF block coding recognition is proposed.Variational Mode Decomposition (VMD) is used to decompose the vibration signal of internal combustion engine into a set of Intrinsic Modal Function (IMF),and the Choi-Williams Distribution (CWD) of the IMF component signal is superimposed in order to obtain the vibration spectrum image with better time-frequency concentration and without cross term interference.In allusion to the parameter selection in the process of VMD decomposition,power spectral entropy is introduced as the objective function and the successive grid optimization is achieved for decomposition parameter of VMD,which improves the adaptability of VMD decomposition.In order to realize the automatic recognition and diagnosis of the vibration spectrum image of the internal combustion engine,a more easily convergent Block Sparse Nonnegative Matrix Factorization(BSNMF) is proposed based on the Sparse Nonnegative Matrix Factorization(SNMF),which is used to extract features of vibration spectrum of internal combustion engine and the support vector machine is adopted to directly conduct the pattern identification of extracted feature parameters.The method is applied to the fault diagnosis of internal combustion engine.The results show that this method can effectively extract the weak fault characteristics of the vibration signal of internal combustion engine and realize the automatic diagnosis of the valve mechanism failure of internal combustion engine.