Abstract: Eliminate the crosstalk of wheel-rail lateral force and vertical force is an important technology of high precision wheel-rail force measurement. An improved fast independent component analysis(Fast ICA) method is presented for wheel-rail force decoupling. This method judges remaining information and finds out the computation end point after one independent component is figured out though add a noise estimation threshold, and consequently, the improvement avoid useless compute and enhance the separating property and stability of this algorithm. Based on vehicle-track coupling dynamics theory, the dynamics model is established and the wheel-rail dynamic responses are calculated, and high precision wheel-rail force based on instrumented rail is obtained by means of improved Fast ICA decouple wheel-rail force. Finally, the algorithm is applied to field test, use genetic algorithm optimized metabolic GM(1,1) model to process discrete signals of rail force gained though field measurement to get the measure results of continuous wheel-rail force of each wheel, improved Fast ICA algorithm is employed to decouple it and acquire good effect.

Key words: Decouple, Independent component analysis, Wheel-rail force