As well known, the mechanical (friction or pneumatic) brake system decelerates trains with the friction force between brake shoes and wheel tread. It seems to be a very simple mechanism. However, the friction changes in a severe non-linear and irregular form. Such an undesirable property makes brake control difficult. Friction cannot be measured easily on running trains, either. If the changes of the friction coefficient can be estimated on a real time basis by using some other signals such as wheel speed which can be obtained easily from the onboard existing hardware system, it must be useful for brake control. In our approach, the method for this purpose has been treated as a non-linear state estimation problem. Thus, we have been trying to apply the extended Kalman filter as a real time estimation method for the friction coefficient. In this paper, a mathematical model of the friction coefficient is made up by using the stochastic system concept, and the extended Kalman filter algorithm is applied for that purpose. Finally, the efficiency of proposed estimation method is proved by computer simulation and trial estimation using the measured data obtained from our dynamo brake test stand.

Copyright (c) 2005 Railway Technical Research Institute