Slip and Adhesion in a Railway Wheelset Simulink Model Proposed for Detection Driving Conditions Via Neural Networks

Abstract

Constantly enlarging operation of locomotives with a very high tractive power in modern railway transport has caused problems with optimal supplying torque from motor to wheel-sets. Losses emerging with inadequate torque values lead to wheel slipping connected with excessive wear and limited acceleration. In models simulating dynamics of torque transmission from the drive units to wheels, the most important are the submodel of the drive and the submodel of balance between traction forces and drive resistances. Some issues of this field studied within a PhD program and SGS (CTU Students Grant Competition) has been focused on increasing quality of these submodels. This contribution is aimed at an innovated part in the existing Simulink model utilizing new data sources and modeling techniques. This improvement supports application of operating point detection methods based on machine learning techniques. New control facilities provided with pulse-width modulated frequency control of the asynchronous motor will be used for automatic submission of optimal operating points. The idea of utilization of via simulation obtained data is an on-line training of polynomial neural unit as an approximation of current driving conditions.