Development of methods for forecasting the technological indicators of the work of a vibrating screen with elastic sives

Abstract

Screening is one of the main operations in the processing of loose material. In the process of operation, the structures of the sieving surfaces must have high in-dicators of the efficiency of the sieving process depending on the classification conditions and keep the sizes of the sieving holes (slots) constant as long as possible in order to min-imize the possibility of coarsening or crushing of the classified material of the calculated size. Currently, sieves made of non-metallic materials, in particular rubber or polyure-thane, which have high wear resistance, have become widely used. A new one has been developed at the Department of Mechanical Engineering of USТ. the construction of an elastic sieving surface, which has high performance indica-tors. On a vibrating screen with a developed elastic sieve, experimental studies were car-ried out and dependences of the screening efficiency on the operating parameters of the vibrating screen were obtained. Mathematical and computer models have been developed, which will allow, using the obtained experimental data, to predict screening efficiency depending on the operat-ing parameters of the vibrating screen. The multiple regression equation for determining screening efficiency was obtained: Y = -51,6593 + 1,598X1-2,5501X2 + 12329X3 (where X1 is the rotation frequency of the motor-vibrator shaft (frequency of vibration of the screening box), Hz; X2 is the amplitude of vibrations of the screening box, mm; X3 - feeding productivity, t/h). Gaussian process regression models give the smallest error. In particular, regres-sion with an exponential kernel function shows the best result. The results obtained by the regression method and machine learning were com-pared. Comparing the results, it can be concluded that the models obtained by the machine learning method give a more accurate result (δmax=5,3%), and when applying the mul-tiple regression equation, the error is 16,2%.