Research On Noise Reduction And Compression Of Mechanical Vibration Signal Based On Compression Sensing

As the world’s industrial powers announced their fourth industrial revolution,informatization plays an important role in the process of industrial automation,but the increasing amount of data also poses a major challenge to the existing signal acquisition and processing technology.The signal acquisition based on the traditional Nyquist sampling theorem produces a lot of redundant information,which seriously wastes the physical storage space and brings great pressure to the data transmission.In recent years,the proposed compression sensing technology provides a new framework for signal acquisition.It can carry out signal acquisition and compression at the same time.It can contain all the required information with far less data than Nyquist sampling theorem,so it becomes a research hotspot in signal processing.In this paper,based on the theory of compression sensing,taking bearing vibration data as the experimental object,the application of signal denoising and compression is studied,and a new concept and improved algorithm are proposed.The main research contents include:1.In the aspect of signal sparse decomposition and its application of noise reduction,aiming at the problem of atom check in matching pursuit(MP)algorithm,a processing method of inner product addition is proposed,which can make full use of the dictionary atom compared with directly deleting the selected atom or covering the inner product;in the noise reduction experiment,on the one hand,the concept of "sparse extreme point" is proposed through simulation experiment;on the other hand,In order to solve the problem of introducing training data noise into dictionary learning,the method of "pre noise reduction" for training data is put forward,and a new method of sparse noise reduction is established.The noise reduction experiment of bearing vibration signal proves that the method has a significant improvement on noise reduction effect.2.In the aspect of the construction of measurement matrix,in order to improve the reconstruction accuracy of the deterministic matrix,this paper takes the toeplitz and circulant matrix as the research object,add the multiplication link in its cycle process,make full use of the randomness of the independent elements,and obtain the toeplitz and circulant multiplication matrix.Experimental results show that the proposed measurement matrix has the advantages of good reconstruction performance of random matrix and less independent elements of deterministic matrix which is easy to implement by hardware.3.In the aspect of signal reconstruction algorithm,in order to solve the problem that the setting of threshold parameter(ts)has a great influence on the algorithm of stagewise orthogonal matching pursuit(St OMP),the algorithm of adaptive stagewise orthogonal matching pursuit(ASt OMP)is proposed according to the number of atoms selected in each iteration,which has better robustness than St OMP and regularized orthogonal matching pursuit(ROMP)and has shorter reconstruction time compared with the sparse adaptive match tracking(SAMP)algorithm with high reconstruction accuracy.