ABSTRACT
A comparative analysis of conventional and neural model for forecasting the wearing out of bearing meeting of stirrer bioreactor is made in the paper. It shows that the preciseness of the neural model is comparable with this of the conventional one and can be used for forecasting. An axial sliding with contact elements “sphere-plane” is investigated. The resource and parameters of pressure of the bearing are determined and the construction is optimized on this basis, thus increasing the reliability of the bioreactor.