DIAGNOSIS WITH THE CORRELATION INTEGRAL IN TIME DOMAIN

In this paper, a method to detect some kind of mechanical failures has been proposed. Based on the fact that any defect on the mechanical system would be inherently defined as statistical differences, we propose that the correlation exponents based on the Grassberger–Procaccia algorithm can be adopted as a diagnostic index. Chattering vibration during cutting process has been investigated as typical example to detect an abnormal vibration. Then the difference between the two stages of normal condition and chattering has been obtained based on the tendency of correlation exponents. Moreover, in order to reduce the noise, only principal components are used for re-embedding space and the modified version of Grassberger–Procaccia algorithm is applied to the re-embedding space. Then the correlation exponents estimated with the modified version of the Grassberger–Procaccia algorithm show more obvious distinction between normal and abnormal condition than those of original Grassberger–Procaccia algorithm.