Neural network based estimation of friction coefficient of wheel and rail

The number of modern control theory applications in vehicle dynamics are emerging and have led to great progress in vehicle stability, handling and ride comfort. However, some of the parameters needed for control applications are difficult to measure online. Such examples are the wheel/rail contact forces, attack angles of wheelsets and the friction coefficient /spl mu/ between wheel and rail of railway vehicles. Other areas where the adequate knowledge of adhesion is vital are the electric drive and adhesion control systems of locomotive drive systems, since as the result of changing friction coefficient wheel spinning, slipping can occur, which can cause faulty operation and overloading of traction units. In order to cope with this problem, this paper presents different methods to estimate the friction coefficient /spl mu/, based on neural network estimation and a computational method.