Research On Compression And Reconstruction Algorithms For Multi-source Monitoring Information Of Rotating Machinery Equipment

With the rapid development of science and technology,large-scale,high-speed rotating machine equipment is more and more widely applied in various fields,especially in the field of wireless sensing,resulting in massive equipment monitoring information generated,which brings great pressure to the traditional signal processing methods.In this paper,compressive sensing is introduced into the processing of rotating machine equipment signals,and the research on the compression and reconstruction of single source and multi-source monitoring information is carried out.Firstly,the compression sensing theory and the distributed compression sensing theory are introduced in detail,in view of the shortcomings of the processing of rotating machine equipment information based on traditional technology,the single source monitoring information compression and reconstruction is studied.Aiming at the problem that single source signal is not sparse in time domain and how to choose the appropriate sparse basis,a sparse representation of single source signal on the Fourier basis,wavelet basis and discrete cosine basis is proposed.Due to different measurement matrices and different reconstruction algorithms have different influences on the reconstruction error,it is proposed to compare the reconstructing performance of single-source monitoring signals under multiple measurement matrices and multiple reconstruction algorithms to select a suitable measurement matrix and reconstruction algorithm,which lays a solid foundation for the research of multi-source monitoring information.Secondly,in order to meet the increasing demand for multi-sensor signal acquisition at present,the research on compression and reconstruction of multi-source monitoring information for rotating machinery equipment is proposed.By analyzing the sparsity of multi-source monitoring information,for JSM-2 joint sparse model,several distributed joint reconstruction algorithms suitable for MMV model are studied and compared.Aiming at the problem that the SOMP joint reconstruction algorithm chooses one atom at a time when reconstructing the signal,which causes large number of iterations and increases computation amount,an improved joint reconstruction algorithm which suitable for multi-source monitoring information is proposed.The simulation results show that,compared with the non-joint reconstruction algorithm and the other joint reconstruction algorithms,the algorithm not only improves the reconstruction accuracy obviously,but also improves the operation efficiency.In the end of the thesis,because the traditional reconstruction algorithm is mainly used to reconstruct the signal,which only needs to consider the reconstruction error and sparsity,there are obvious limitations for applying it to classify of device state,a reconstruction algorithm based on non-minimum reconstruction error criterion is proposed,which makes reconstruct signal not only consider the reconstruction error and sparsity when reconstructing,but also consider the difference between signals in different states.The simulation results show that,compared with the traditional reconstruction algorithm,reconstruct signal by the proposed algorithm to classify the device status not only solves the unification of device state classification accuracy and computational complexity under low sampling rate,but also provides the possibility of the online diagnosis of the device state.