EMD-based time–frequency denoising algorithm for the self-sensing of vibration signals in ultrasonic-assisted grinding

Abstract

Abstract An empirical mode decomposition (EMD)-based time–frequency denoising algorithm is derived in this paper to accomplish the extraction of active ingredients for multi-frequency mixed signals, and then the self-sensing of vibration signals is realized without additional sensors for an ultrasonic-assisted grinding device. EMD is employed to accomplish the signal decomposition, and multiple intrinsic mode function components and the residual are obtained. Then, the weighted factors used for signal reconstruction are obtained based on innovation statistical distance, which is selected as a criterion to evaluate the time–frequency domain correlation between decomposition results and the original signal. Next, the maximum correntropy criterion-based similarity detection method is designed to adaptively modify the weighted factors, and the signal is reconstructed on the foundation of the modified weighted factors. Finally, numerical simulations and a self-sensing experiment are conducted to verify the denoising performance of the proposed algorithm. The self-sensing of the 4.59 dB vibration signal is realized for an ultrasonic-assisted grinding device.