Boundary feature study on engine wear based on wavelet analysis

Due to poor data of oil monitoring, the wear boundaries obtained by applying traditional three-line method (TTM) are not accurate enough and very likely to lead to higher false alarm rate. To mine wear regularity feature and singular characteristic information, Wavelet transformation was used to oil spectrometric and direct reading ferrography data denoised to acquire approximation coefficient and detail coefficient. Based on these coefficients, new three-line method (NTM) was constructed, and accordingly the normal boundary, warning boundary and dangerous boundary were extracted. To compare the boundaries obtained separately with NTM and TTM, the index, the ratio η of abnormal range and normal range, is put forward according to the curve of the effective lifetime vs failure rate of the mechanical equipment. The analysis of wear boundary feature for engine shows that the wear state determined using the new method is more consistent with the actual situation than TTM, and the index of new method is less than that of TTM, which proves higher reliability of boundaries determined using NTM. Moreover, the statistical results on engine wear state show new method is of higher accuracy rate and abnormalities discrimination rate.