Simplified contact filters in wheel/rail noise prediction

When predicting rolling noise due to wheel and rail roughness a “contact filter” is generally applied to account for the effect of the finite size of the wheel/rail contact. For time-domain analysis these calculations must be fast enough to get results in a reasonable time. Remington and Webb have devised a versatile three-dimensional ‘distributed point reacting spring’ (DPRS) contact model that is relatively quick, but if only one line of data is available along the contact it is unnecessarily complex, so a simpler two-dimensional version has been developed here. When this new model was checked against a Boussinesq analysis of the contact, the results in one-third octave bands were found to agree to within 3 dB. These results further suggest that the two-dimensional DPRS model might have an unexpectedly wide range of applicability, including large amplitude sinusoidal roughness and discrete features such as a rail joint. When implemented at each step in a time-domain wheel/rail interaction analysis, this model gave similar results to quasi-static roughness filtering with a constant load for moderate roughness, but dynamic effects became significant when the roughness amplitudes were large, particularly with dipped rail joints.