The dynamics of a railway vehicle are driven by the geometry and conditions at the wheel-rail contact. Typically the condition and shape of the wheel and rail are monitored separately and off line. The work presented here is part of ongoing research into on-line model-based estimation of parameters in the wheel-rail contact dynamics. This paper outlines a practical approach to estimating a nonlinear function within a dynamic system by using a piecewise cubic functions. The parameters for the cubic functions are estimated with a least squared approach applied to the dynamic measurements taken from the system. A simplified plan-view wheelset and suspended mass model is introduced to use as an application of this technique. A contact geometry term, conicity, which is a nonlinear function of the relative lateral wheel-rail position, is included in the rail vehicle model. The conicity is successfully estimated using the least-squares method outlined in the paper.