Computer modeling - An emerging tool for welding engineers to achieve structurally sound and reliable welds

Welded joints contain spatial variations of composition, structure and properties resulting from heat transfer, flow of molten metal in the weld pool and solid state transformations during welding. Computational models, based on scientific principles, help in quantitative understanding of complex welding processes and welded materials and provide solutions for a variety of welding problems. Numerical models of heat transfer, fluid flow and thermal cycles can now be relied upon to provide fusion zone geometry, temperature fields and thermal cycles with a fair degree of confidence. Significant progress has also been made in building a quantitative understanding of various structural features such as phase composition, grain structure and inclusion structure, mostly for several relatively simple alloys. As progress is made through the application of existing models, advances in computational hardware and software are now providing unprecedented opportunities to build more ambitious, computationally more intensive and more reliable models. An important recent development has been the construction of heat transfer and fluid flow models that can determine uncertain welding variables from a limited volume of experimental data using multivariable optimization. Predictions from these models are more reliable because they learn from experimental data. Since the computed results reveal detailed insight about the welding processes and welded materials that cannot be obtained otherwise, the modeling efforts are already contributing to the growing quantitative knowledge base in fusion welding. The expanded knowledge base is necessary to transform welding to a mainstream engineering branch.