Research On Compressed Sensing Noise Reduction And Abnormal Fast Marking Technology For CNC Machine Tool Electronic Data

Nowadays,intelligent manufacturing is thriving,it is of great significance to bring the outstanding achievements of information technology into the manufacturing industry.It is difficult to apply the general data transmission and processing technology in industrial big data in machine tool.It is necessary to make some adjustment according to the source characteristics of the data,combining the practical application characteristics such as data acquisition,conversion and regulation.The instruction-Domain has laid a solid foundation for the analysis of numerical control machining data and actual machining situation.The current and electric power signal data in the electronic control data can reflect the various states of the machine tool.The current and electric power data collected from the servo motors of each axis is the keypoint of this paper.In the process of signal acquisition and transmission,due to measurement error,random error,conversion error,etc.,it is inevitable that it will be interfered by a large number of invalid signals.The research on sparseness of electronic control data of machine tools has laid a good foundation for data compression and denoising,sparsity is part of compressed sensing,and the sparsity of data is also the premise of compressed sensing.The collected signals are mapped to low-dimensional space.The target electrical signals can be reconstructed with sparseness in the sparse domain,while the noise can not be reconstructed without sparseness in the sparse domain.Finally,the reconstruction algorithm is used to reconstruct the current signal to achieve noise reduction.Due to the magnanimity of industrial big data,it is difficult to find abnormal data that is meaningful for analysis.This thesis proposes a calibration method for the abnormal data of parallel feed motor current and spindle power under the conditions of ring cutting and drilling.In the standard operation,the load fluctuation represented by the electrical signal is stable and relatively stable with regular fluctuation.This dissertation put up with the method of wavelet decomposition,filtering and reconstruction to mark according to the abnormal fluctuation of the abnormal data that does not meet the data trend.It is proposed to simultaneously process the X,Y two-axis torque current signal and the spindle electric power signal,which makes the abnormal situation more sensitive and the calibration moreeffective.Finally,experiments are used to demonstrate the effectiveness of denoising and fast labeling,which shows the effectiveness of the denoising method and the rationality of the abnormal fast marking method.