A NUMERICAL FRAMEWORK FOR LOAD IDENTIFICATION WITH APPLICATION TO WHEEL-RAIL CONTACT FORCES

Paper in proceeding, 2009

This contribution focuses on optimization based methods for inverse identification, i.e. the solution of nonlinear least-square problems, which is of interest in railway mechanics to determine the contact forces between the train wheel and the rail. A particular setup is considered here: The radial strain, , is recorded at eight locations of the wheel, Figure 1. The contact force, , is determined as the minimiser of the discrepancy between the recorded strain and the corresponding predicted strain, , extracted from a finite element simulation of the train wheel. Thus, minimises . Since errors arising from the FE-discretization (both spatial and temporal) carry over to the identified contact force it is vital to ensure that these errors are of acceptable magnitude for the reliability of the results. In addition, measurement noise carries over to the solution. In recently proposed formulation, cf. [1], the pertinent optimality conditions are expressed in weak forms which enables the choice of discretization on the state equations in space-time, the sampling instances of the measurements and the parameterization of the sought force are all decoupled, in contrast to traditional methods e.g. dynamic programming.