Multi-source data fusion soft-sensor method for ball mill operating parameters based on transformation of radial basis function

Reliable measurement of the mill load is one of key factors to improve mill productivity, production quality and decrease energy consumption for the grinding process. A multi-source data fusion soft-sensor method is proposed to estimate the operating parameters which present the mill load. Fast Fourier transform (FFT) is used to estimate the power spectral density (PSD) of the vibration signals from mill shell and acoustic signals near mill because some important information of mill load is closely related to the relative amplitudes of different frequencies. Radial basis function (RBF) is used to map the spectral feature to nonlinear space after feature variables of the spectral data are extracted through principal component analysis (PCA). At last, three partial least squares (PLS) models are built to predict mill operating parameters, whose input variables are the linear and nonlinear spectral features and current signal of mill motor. A case study shows that the proposed soft-sensor method has higher accuracy and better predictive performance than other methods.