Thermal and mechanical analysis of welded structures

Abstract

The prediction of the residual state of stress and deformation in a welded structure is one of the most interesting, challenging, and complex problems in structural mechanics. The wide spectrum of physical phenomena provides the interest; economic and safety considerations provide the challenge; and the essential nonlinearity of the analytical models provides the complexity. Even if it is assumed that reasonable prediction of tje transient temperature field in the structure is possible, determining the residual mechanical state is both a difficult and an expensive task. The analysis inevitably involves temperature-dependent mechanical properties and, in addition, the severe thermal gradients and high temperatures generic to the welding process induce irrecoverable inelastic creep and plasticity in the structure. In spite of this, the stress analysis is now considered to be a straightforward application for general purpose, nonlinear finite element structural programs. A few special features of such analyses, however, will be discussed: (1) the legitimacy of time-dependent plasticity theories for treating the residual stress problem; (2) criteria for choosing plane stress, plane strain, generalized plane strain, or fully three-dimensional models; (3) methods for coping with possible floating solid regions during cooling; and (4) the use of linear constraints to treat weld metal deposition and intermittent contact. Since the most important parameters in the welding process that pertain to the stress analysis are the cooling rate and the welding torch efficiency, the heat transfer problem seems to require a critical look. The dominant features of this problem are: (1) source (torch) characterization; (2) radiation from surfaces that are heated to high temperatures; (3) latent heat effects; and (4) subsidiary considerations, such as enforced convection heat transfer modes that are designed to control the cooling rate. Motion of the welding torch, even at low speeds, is not usually a critical factor in determining the residual mechanical state. Again, finite element analysis is applicable, provided that the solution accuracy can be adequately estimated. Several alternative, two-step, implicit time integration schemes will be compared, especially with regard to accuracy and numerical stability for welding-type problems. The efficacy of ‘flux correction’ will also be discussed and the application of these ideas to typical industrial welding problems will then be outlined.