| Abstract: | In this paper, three numerical algorithms for the identification of wheel–rail contact forces based on measured wheel disc strains on an instrumented railway wheelset are discussed and compared. The three algorithms include one approach resting on static calibration, one that is applying a Kalman filter and the third is exploiting an inverse identification scheme. To demonstrate and evaluate the alternative methods, two load cases including periodic excitation by sinusoidal wheel–rail irregularities and transient excitation by an insulated rail joint are considered. Based on a previously presented vehicle–track interaction model in the time domain, load scenarios are defined by taking the calculated vertical wheel–rail contact forces as the reference force to be re-identified by the proposed algorithms. The reference contact forces are applied on a finite element model of the wheel to generate synthetic observation data, that is, radial strains at the positions of the strain gauges, serving as input to the identification procedures. It is concluded that the inverse identification scheme leads to superior accuracy at higher computational cost. If on-line implementation and evaluation is required, the Kalman filter generates better accuracy than the static calibration approach. |
