基于稀疏编码和禁忌优化的故障信号抽取方法

Fault signal Extraction Method Based on Sparse Encoding and Tabu Optimization Algorithm

摘要:

为了克服经典正交匹配算法获取原子集时遍历冗余字典具有较大时间开销的缺点,提出了一种基于压缩感知理论和禁忌优化算法的的稀疏故障信号特征提取方法;首先引入了压缩感知模型并描述了基于信号稀疏表示的故障诊断原理,设计了满足RIP准则以最小化l₁范数为目标的稀疏信号解的求解方法,然后定义了一种基于正交匹配算法的稀疏信号重构算法,并以最小化余量为目标函数,采用改进的禁忌搜索算法在原子空间中搜索满足目标函数的最优原子集,最后,给出了基于稀疏编码和禁忌优化混合模型的故障信号提取算法;在Matlab仿真环境下对滚动轴承故障信号进行试验,仿真结果表明:文章方法能有效地对具有强噪声的故障信号进行稀疏重构,不仅具有较高的信噪比,而且具有较小的余量误差和仿真时间,与其它方法相比,具有较大的优越性。 

Abstract:

In order to conquer the defects of the classic orthogonal matching algorithm obtaining the atom set having the much time consumption, a feature extraction method based on compressed sensing theory and Tabu Optimization Algorithm is proposed. Firstly, the compressed sensing theory is introduced and the fault diagnosis principle is described, the sparse signal solving method using the l1 norm as the goal and satisfying the RIP rule. Then a sparse signal reconstruction method based on orthogonal matching is defined, and using minimizing remain value as the goal, the improved tabu optimizing algorithm is used to find the optimizing atom set in the atom space. Finally, the fault signal extraction algorithm is given by combining sparse encoding and tabu optimizing. The simulation experiment of rolling bearing fault signal is simulated in the Matlab simulation environment, and the result shows the method in this paper can effectively realize the sparse reconstruction for fault signal with strong noise, the method in this paper not only has higher signal noise rate, and also has the less remain error and simulation, and compared with the other methods, it has big priority.