Modeling load parameters of ball mill using frequency spectral features based on Hilbert vibration decomposition

Load parameters inside the ball mill is one of the key factors that affect grinding production ratio and production quantity of the grinding process directly. The ball mill produces soundly mechanical vibration and acoustical signals. Many methods have been applied to measure them. In this paper, a new frequency spectral feature of Hilbert vibration decomposition (HVD) based soft sensor approach is proposed. Sub-signals with different physical interpretation are obtained with HVD technology. Different frequency spectral features of these subsignals are selected using fast Fourier transform (FFT) and mutual information (MI), which are fed into kernel partial least squares (KPLS) for constructing soft sensor model of the mill load parameters. Experimental results on a laboratory ball mill show that the pulp density can be effective measured using the proposed method.