Abstract:

Aiming at the problem of spectrum aliasing of the cutting force caused by unreasonable setting of sampling parameters and filtering steepness of anti-aliasing filtering in the high-speed machining condition monitoring system,a spectrum sensing method based on the principle of frequency domain approximate sparseness is proposed.The non-linearity of the machining system and sampling process makes the output waveform of the monitoring system contain high order harmonics,which shows obvious approximate sparsity under the Fourier matrix (with the amplitude of most elements in the Fourier coefficients approximately zero,and the energy mainly concentrated in several frequency points).Several spectrum subsets are the result of spectrum approximation by retaining few spectral lines with a large amplitude only.Using the principle of subsampling mixing,the true frequency range of each spectrum subset is calculated,and the true frequency spectrum of the cutting force is corrected.According to the time-domain waveform characteristics of the subset of frequency bands,a general Linear Amplitude Modulation Sinusoidal Wave (LAMSW) model is constructed.The effectiveness of the proposed method is verified by LAMSW simulation analysis and high-speed milling aluminum alloy experiments.The results show that the proposed method is effective enough to recover the true waveform of the milling force signal,and that the relative envelope error between the recovery time-domain wave and the test signal is less than 4%.The research results provide a certain engineering and technical support for applying the sparse theory to the analysis of subsampling signals.

Key words: high-speed milling, condition monitoring, subsampling signal, spare representation, fast fourier transform