In this paper, various methods for estimating frequencies, amplitudes and phases on the basis of Maximum Entropy Spectrum are compared with each other. And Maple Maximum Entropy Spectral Line Estimation (MMESLE) is proposed which includes (1)parameter estimation for Maximum Entropy Spectrum by Marple's fast recursive method, (2)calculation of power spectrum and determination of frequencies by FFT, and (3)estimation for amplitudes and phases by Cyclic Descent. In comparison with the conventional FFT spectral analysis, it has a higher estimation accuracy and resolution, especially suitable for the spectral analysis of short record data. Then the MMESLE is applied to the spectral analysis of spiral undulation errors of a helical gear and the differential train errors of hobbing machine. It is shown that the main error sources can be exactly diagnosed. The results are satisfactory.
展开▼
