Online identification methods for low-order wheel polygons coupled with axle cracks

Wheel polygons pose a threat to the safety of train operations due to wheel fatigue, necessitating online identification. The online identification of wheel polygons predominantly relies on the axle box vertical nx (n-multiple integer of axle rotation frequency) signals generated by wheel polygons. The low-order vertical nx signals are easy coupling to other faults like axle cracks, leading to errors in wheel polygon identification. A signal decoupling method is crucial for removing the crack components from the vertical nx signal to obtain more accurate identification results for low-order wheel polygons. Employing principles of signal superposition for different faults, this article proposes a method to decouple the wheel polygon nx signal from the vertical nx signal by subtracting the crack-related portion. This method has been validated through an ABAQUS + SIMPACK online rigid-flexible coupling model, indicating that it significantly improves the accuracy of wheel polygon identification. The 1x amplitude of the decoupled signal decreased by 60.78%, and the 2x amplitude decreased by 55.79%.