Pulsed GTAW is shown to have certain advantages which can make it competitive to other "temper-bead" procedures developed in the past for in situ weld repairs. Low pulse frequencies and low welding speeds are shown to be required for obtaining the necessary degree of microstructural refinement for such applications. This conclusion is drawn from a model based analysis experimental data from pulsed GTRAW weldments on a AT steel. Relating the model to the experimental results was possible only after the heat flow model was calibrated with the experimental data obtained from a metallograhic characterisation of the pulsed welds and partially from dilatometric measurements during thermal cycle simulation. This phenomenological or "inverse modelling" procedure of estimating the physical constants in the heat flow model was made easier and more "well-conditioned" because transient information about the thermal cycles and phase transformations could be inferred from the pulsed weldments and the weld simulation trials.