Abstract: To solve the problems of wear debris image segmentation of reflected ferrogram captured by full field-of-view online visual ferrograph (OLVF) under the condition of out-of-focus bubble interference, a low frequency nonlinear threshold function and a high frequency one are derivated respectively by using hyperbolic tangent function. On the basis of the novel threshold functions, a wear debris image segmentation algorithm based on wavelet modulation of strong bubble interference is proposed. In order to realize the effective segmentation of wear debris image of full field-of-view OLVF reflected ferrogram, the bubble impression and background noise in the reflected ferrogram need to be suppressed and the contrast and sharpness of its wear debris image are able to be enhanced by using this current one. Comparing with the segmentation effects of wear debris images processed by other algorithms, the results show that the complete segmentation of wear debris images in the reflected ferrogram can be achieved more accurately by modulating bubble impression and background noise with the proposed algorithm. Finally, an online monitoring experimental measurement of gear wear is carried out, and the index of debris concentration is extracted and obtained to characterize the fast-changing of gear wear in the full useful life. Results show that the effects of bubble impression on the vilidity of wear debris image segmentation of reflected ferrogram during the sampling periods of full field-of-view OLVF can be eliminated by this current algorithm, and the abnormal wear fault of gear failure is accurately detected. Accordingly, it has been further verified that this proposed algorithm for full field-of-view OLVF is applied to extracting the visual characteristics of wear debris in in-use lubricants.

Key words: full field-of-view OLVF, bubble impression, wavelet modulation, reflected ferrogram, wear debris segmentation