Estimations of rail irregularities

Abstract

Inspection cars that have been used to measure rail irregularities are costly and need operators. This thesis proposes a method to estimate rail irregularities by measuring accelerations of a passenger car instead of direct measurement by using a conventional inspection car. Using a System Identification technique and data obtained by simulations based on a three-dimensional rail vehicle model with actual rail irregularities, the proposed method identifies an inverse system where inputs are accelerations of a vehicle and outputs are rail irregularities. The resulting model is assessed through comparing the estimated irregularity with the actual irregularity. Validation results show that the estimate agree well with the actual irregularity for the Vertical Irregularity. Though the estimation error for the Lateral Irregularity is larger than that for the Vertical Irregularity, the error is acceptable form a practical point of view. The quality of the estimation is evaluated quantitatively by using the Mean Square Error. In addition, resolution of the estimation is presented in order to demonstrate the accuracy of the estimation. Model uncertainties are assessed for a practical implementation. The effects due to two major uncertainties, mass variations and speed variations, are evaluated by using the Singular Value Decomposition in order to present the limitations of the estimation using a nominal model. In addition, this thesis proposes a compensation method for mass variations and speed variations.