Extraction of time-frequency ridge line based on automatic peak search and curve fitting

Abstract

Abstract Instantaneous rotational frequency (IRF) plays a crucial role in diagnosing bearing faults under variable speed conditions. However, accurately extracting the rotational frequency becomes challenging due to significant noise and strong harmonic interference. To address this issue, this paper proposes an approach for extracting Time-frequency ridge line through automatic peak search and curve fitting with a minimal number of parameters. The approach eliminates disturbance from harmonics and noise by locating peak coordinates within specified areas. Moreover, these peak coordinates were fitted as interpolated nodes to prevent sudden changes in the curve due to noise interference. It also enhances the accuracy of the extracted ridge. The efficacy of the proposed technique is confirmed through simulations and specially designed test bench data. Results show that IRF-estimation errors of our method fall by about 20% compared with those of any traditional method, approaching TKEO and Phase Demodulation without using too many parameters.