Values of several parameters that affect calculation of heat transfer during fusion welding such as the arc efficiency, arc radius and the effective thermal conductivity of the liquid metal are not always readily available. Following an inverse approach, a smart model that embodies an iterative procedure for the optimization of multiple unknown variables within the framework of phenomenological laws that govern heat transfer and fluid flow in the weld pool is developed. The optimization scheme considers the sensitivity of computed weld geometry with the unknown parameters. The weld penetration was found to be sensitive to all the unknown variables considered. The weld width was influenced mainly by the arc efficiency and the arc radius. The initial values of the unknown variables did not affect their optimum values but affected the number of iterations necessary for convergence. The model could correctly learn values of multiple unknown parameters from only a few measurements of weld penetration and width and, based on the knowledge of these parameters, provided realistic predictions of heat transfer, fluid flow and weld geometry.
All Science Journal Classification (ASJC) codes
- Electronic, Optical and Magnetic Materials
- Condensed Matter Physics
- Acoustics and Ultrasonics
- Surfaces, Coatings and Films