Clarke, Jo Ann Marie2006-07-282006-07-2819981998http://hdl.handle.net/10012/228A columnar-to-equiaxed grain transition (CET) in gas tungsten arc (GTA) welds may reduce susceptibility to solidification cracking and brittle fracture and improve weld properties, such as toughness, ductility, and strength. A CET may be promoted in the weld pool by providing both (i) a supply of embryos from which equiaxed grains may develop and (ii) thermal conditions which favour the nucleation and growth of these embryos into equiaxed grains. However, the relationship between solidification and thermal conditions in the weld pool and process parameters which are controllable during welding is not well understood. Since it is experimentally difficult to control and characterize the thermal conditions in a moving weld pool, numerical simulations of the macroscopic heat and fluid flow behaviour in the GTA weld pool were made using a detailed thermofluids model. This finite element model was formulated for three-dimensional, steady-state, moving GTA welds. The model incorporated temperature-dependent material properties, latent heat of fusion, Gaussian-distributed arc heat and current inputs, and heat losses due to convection and radiation. Turbulent fluid flow driven by buoyancy, electromagnetic forces, and surface tension temperature gradient forces was also modelled. The numerical thermofluids model was used to simulate move GTA welds on thin sheets of aluminum-copper alloy using an experimentally realistic range of welding currents and steady-state welding speeds. Results from the simulations were post-processed to obtain estimates of weld pool thermal and solidification parameters, such as thermal gradient, interface growth rate, and growth undercooling, and thus predict the influence of welding process parameters on weld pool thermal conditions in Al-Cu alloys. An analytical model was then evoked to predict the CET based on the thermal conditions in the weld pool as predicted by the numerical thermofluids model. The analytical CET model was based on Hunt's expression for the CET [1], modified for welding by incorporating the marginal stability dendrite growth model developed by Kurz et al. [2, 3]. The coupled thermofluids model and analytical model predicted that, given an adequate supply of nucleants, the welding conditions which favour a CET are (i) for a given current, decreasing welding speed, or (ii) a high current and welding speed combination. It was also predicted that, although a large population density of heterogeneous nucleants having low nucleation undercooling will promote a CET for any of the welding conditions, particularly at higher solute contents, the welding conditions become increasingly important when the population density of heterogeneous nucleants is small of their efficacy decreases. A complementary experimental program was performed to examine the model's predictions. Experiments were executed using autogenous, alternating-current GTA welding for a range of welding currents and steady-state welding speeds. Full-penetration welds were produced on thin plates to ensure two-dimensional heat flow and thus facilitate subsequent microstructural interpretation. Binary aluminum-copper alloys were selected as a model solidification system since this alloy system is well-characterized with respect to material properties and is also particularly susceptible to solidification cracking. For the purpose of model validation, a series of experiments was performed in which thermocouples were embedded in Al-Cu plates prior to welding. Comparisons in GTA welds in Al-Cu alloys of experimental measurements of peak temperatures, temperature distributions, and weld widths, with predictions from the thermofluids model demonstrated excellent accord. To study the CET, experiments were performed on thin plates of aluminum alloys containing 2 and 4% copper. These alloys were inoculated with varying amounts of TiB2 particles which acted as nucleants for equiaxed grains. Measurements were made of % equiaxed grains in the weld bead and compared to the CET predictions made using the coupled thermofluids and analytical CET models. In comparison with experiment, the CET model was able to correctly predict whether or not a CET would occur and the general trends, as follows: (i) a fully columnar grain structure is produced at low current-welding speed combinations when TiB2 levels are low, (ii) the % equiaxed grains in the weld bead increases with increasing TiB2 content, (iii) the % equiaxed in the weld bead is greater for the alloy containing 4% Cu compared to that containing 2% Cu; i.e., the % equiaxed increases with increasing Cu content, (iv) the % equiaxed is greater for welds performed with higher current-welding speed combinations, and (v) for a given current, the % equiaxed is greater with lower welding speed. The numerical comparison of the predicted values of % equiaxed with the experimental measurements accorded moderately. The discrepancies between the observed and predicted % equiaxed were suggested to be due to two factors: (i) the assumption of a constant value for TN, (ii) the assumption that the latent heat of fusion evolved by the growing equiaxed grains was conducted away and did not influence the thermal conditions in the weld pool. The study revealed that the technique of coupling the predictions of a macroscopic thermofluids model with a microstructural model for the CET is an effective technique for the predictions of qualitative trends with respect to the CET as a function of GTA weld process parameters, alloy composition, and heterogeneous nucleant population density. Good correlation between experimental measurements of weld widths, peak temperatures, and temperature profiles and the numerical predictions was only possible when the effects of turbulent fluid flow were incorporated in the numerical model. The study demonstrated that a CET is favoured in GTA welds in Al-Cu alloys by (i) for a given current, decreasing welding speed, (ii) a high current and welding speed combination, (iii) increasing copper content, and (iv) increasing the wt% of the nucleating agent for equiaxed grains. Although a CET will be produced at all welding conditions in the presence of a large population density of heterogeneous nucleants, if the nucleants are inefficient or their numbers are low the welding conditions become increasingly important for CET promotion. Better agreement between the measured and predicted % equiaxed values would require (i) a more sophisticated CET model which is capable of accounting for a distributed nucleation undercooling, and (ii) an iterative coupling between the macroscopic thermofluids model and the CET model to account for the synergistic relationship between the latent heat of fusion released by the equiaxed grains ahead of the columnar grain interface and the growth of these equiaxed grains as well as the thermal gradients in the weld pool.application/pdf6951135 bytesapplication/pdfenCopyright: 1998, Clarke, Jo Ann Marie. All rights reserved.Harvested from Collections CanadaDoctoral Thesis