Machine learning in calculating speeds in a railway sorting yard

Abstract

The main step in the process of disbanding railway rolling stock at the marshalling yard is the regulation of the speed of cars that slide freely from the hump, which in most cases is carried out thanks to the brake positions installed on the hump yard in several groups. The paper considers the solution of the problem of determining the optimal exit speed of a group of cars (trailers) target braking positions on the tracks of the sorting yard, since the efficiency of the sorting process in terms of optimal accumulation on the cars on the tracks of the sorting yard and its safety depends on the correctness of the calculation of this speed in terms of compliance with the permissible speed of collision with the cars on the tracks. In the proposed study, an analogy is drawn between the calculation of the optimal exit speed from the brake position and the compilation of nonlinear multiple regression. The classic and modern machine learning algorithms to build regression models were analyzed. The most suitable algorithm was identified within the study. In conclusion, the results of the introduction of machine learning tools at a real automation facility for sorting processes are presented.