Because of the seed separation process complexity in the combine cleaning system and of the multitude factors influencing it, the paper approaches the possibility to describe the separation process along the sieve using a statistic mathematical model described by the multiple logistic regression model consisting from the two parameters logistic function P-x = 1+exp(alpha+beta x)/exp(alpha+beta x) where P-x - percentage of separated seeds along sieve length x; alpha and beta - logistic constants, expressed through multiple logistic regression functions. The values of the alpha and beta constants are significantly influenced by the sieve oscillation frequency (f), the specific alimentation flow with material of the sieves (q) and the straw parts content (pp/s) of the processed material. For the logistic coefficients alpha and beta was proposed the multiple linear regression function of Y = A + Bx(1) + Cx(2) + Dx(3) were x(1) = n, x(2)= q and x(3) = pp/s. By testing the multiple logistic regression model with experimental data, obtained from 7 experimentations done in laboratory conditions, was found a = - 10.381 - 0.012f - 0.661q + 36.917 pp/s, beta = 52.943 - 0.050f + 6.084q - 121.206 pp/s, for a correlation coefficient R-2= 0.803, respectively R-2= 0.904. The conformity between estimated seed loses and those determined in experimental measurements is illustrated by a correlation coefficient R-2 = 0.904. Using the proposed model it call be anticipated the seed lose probability in the cleaning system (=1-P-1) if we impose the working conditions through the values of f, q, pp/s and sieve length L, in experimental values range used at experimentations. These data are useful both in design activity, as well in efficient use of the cleaning system from the classic combines, which call be added to the field data library.
