Amplitude Normalization for Speed-Induced Modulation in Rotating Machinery Measurements

Rotating machinery under variable-speed conditions suffers from amplitude modulation (AM) effects induced by speed fluctuations, complicating accurate fault detection. To address this issue, an amplitude normalization method based on support vector regression (SVR) is proposed to estimate and remove the AM effects. The method employs a correlation-based feature selection strategy to construct feature vectors strongly associated with rotational speed, thereby enabling the accurate quantification of speed-induced AM effects. The robust nonlinear fitting capability of SVR is then utilized to model and remove these effects, enhancing fault signal clarity. The proposed method is validated through two case studies and compared with advanced amplitude normalization techniques, demonstrating its superior accuracy, robustness, and reliability. Experimental results demonstrate that the proposed method accurately estimates and eliminates speed-induced AM, significantly improving fault diagnosis accuracy by up to 34.7%.