Finite element analysis of thermal fields during repair welding of discrete rail defects

Other - Master Thesis, 2017

Discrete defects in a rail head may form due to aggressive wheel–rail contact in terms of thermal and/or mechanical loads, or due to indentations from foreign objects trapped in the contact. If large, such defects need to be repaired or the rail section removed. These are costly operations that cause operational disturbances. To decrease mitigation costs, discrete defect repair (DDR) procedures that include repair welding have been developed. These operations typically require high preheat temperature (350 °C) and long working process times.

This MSc-thesis work investigates a novel DDR rail welding procedure through numerical simulations. The new technique employs significantly lower preheat temperature (60–80 °C) and equipment that can easily be carried to the working place. However, the low preheating temperature introduces high temperature differences between the molten filler material and the surrounding rail steel. This may lead to the formation of defects, welding related cracks or martensitic areas.

The aim of the work is to simulate the DDR procedure and thereby be able to analyse the thermal history in the rail during the welding process. In this manner, cooling curves for critical locations in the rail head can be evaluated and the risk of weld related defects and metallurgical transformations to hard microstructures can be assessed. To achieve these ends, numerical models of a milled rail head were created in ABAQUS/CAE. The repair welding procedure was then simulated and the results compared to experimental data from the literature.

The results show temperature trends that are in line with temperature measurements from trials carried out some years ago. The simulations show the sensitivity to parameters such as the temperature of the molten filler and cooling times. There is thus a high potential in simulating operational procedures and thereby be able to e.g. investigate effects of various process parameters. However, to this end more highquality test data are required. In particular the simulations show how sensitive a calibration is to the exact position of thermocouples. On the other hand, the simulations performed in the thesis have shown that small variations in the geometry of the numerical model of the repair process do not have a significant influence on the predicted cooling curves.