Modelling of mill load for wet ball mill via GA and SVM based on spectral feature

Abstract

The load of wet ball mill is a key parameter for grinding process, which affects the productivity, quality and energy consumption. A new soft sensor approach based on the mill shell vibration signal is proposed in this paper. As the frequency domain signal contains more evidently information than time domain, the power spectral density (PSD) of the vibration signal was obtained via fast Fourier transform (FFT). And then the mass and the central frequency of the small peaks of the spectrum are extracted as the spectral features. At last the support vector machines (SVM) is used to build the soft model. The parameters of SVM, the input variables including the mass and the central frequency of the peaks are selected by Genetic algorithm (GA). Experimental results show that proposed soft sensor model has higher accuracy and better predictive performance than the other normal approaches.