信号共振稀疏分解方法及其在机械故障诊断中的应用研究

Resonance-based sparse signal decomposition method is a novel nonliner signal decompositon mehtod recently proposed by Ivan W.Selesnick. In this method, the suitable Q-factor is selected according to the analyzed signal,the base function dictionary is acquired by wavelet transform with tunable Q-factor, and the target function is built by morphological component analysis；then, the optimal result is got by the split augmented Lagrangian shrinkage algorithm. By using of the resonance-based sparse signal decomposition method,a complex signal can be decomposed into high resonance component with oscillating component and low resonance component with impluse component. When a damage occurs in a mechanical system, its vibration signal is always composed of transient impluse signal,oscllating signal and noise. This project intends to introduce the resonance-based sparse signal decomposition method into mechanical fault diagnosis, and use it to separate the impluse signal and oscillating signal from mechanical fault vibration signals.This project will build complete and systematic mechanical fault diagnosis mehods based on resnonace-based sparse signal decomposition. Since the impluse component contains the fault information of a equipment related to impluse, and the oscillating component can reflect the working condition of a equipment and contains the fault infromation non-related to impluse, therefore, the reserch results in this project can be used to extract fault characteristics and enhance the accuracy of mechanical fault diagnosis effectively.

信号共振稀疏分解方法是最近由Ivan W. Selesnick提出的一种新的非线性信号分解方法。该方法首先根据待分析信号选择合适的品质因子，通过品质因子可调小波获得相应的基函数库，然后利用形态学分量分析方法建立信号稀疏表示的目标函数，最后通过优化求解，将一个复杂信号分解成由振荡成份组成的高共振分量和由冲击成份组成的低共振分量及随机噪声。当机械设备发生故障时，其振动信号通常由瞬态冲击信号、谐振信号及噪声构成。本项目拟将信号共振稀疏分解思想引入机械故障诊断，研究用信号共振稀疏分解方法从各类机械故障信号中有效分离出冲击信号与谐振信号，在此基础上建立完整、系统的基于信号共振稀疏分解的机械故障诊断方法。由于信号冲击成份包含了设备与冲击相关的故障信息，而谐振信号则反映了设备的运行工况和与冲击无关的故障信息，因此，本项目研究成果能有效用于机械故障振动信号的特征提取与分析，提高机械故障诊断的准确性。

本项目完成了信号共振稀疏分解（Resonance-based sparse signal decomposition, RSSD） 方法的理论研究和RSSD方法在机械故障诊断中的应用研究，在《Mechanical Systems and Signal Processing》、《机械工程学报》等国内外学术刊物上发表学术论文20余篇；培养博士生2名，硕士生5名。主要研究工作包括：（1）提出了基于RSSD的机械故障信号分解方法，并应用于转子系统振动信号分析；（2）提出了基于RSSD的包络解调方法与重分配小波尺度谱方法，并分别应用于转子系统、齿轮和滚动轴承故障诊断；（3）提出了基于信号共振稀疏分解的阶比分析方法与品质因子可调小波重构方法，并分别应用于齿轮与滚动轴承振动信号分析；（4）提出了最优共振稀疏分解方法与最优信号共振分量能量算子解调方法，并应用于轴承故障诊断与齿轮箱复合故障诊断；（5）提出了信号共振稀疏分解与分数阶微积分算子相结合的油液颗粒物监测研究方法。在上述研究工作基础上，初步建立起系统的基于信号共振稀疏分解的机械故障诊断方法，为机械故障诊断提供了新的途径与手段。