Generalization and Communication of Welding Simulations and Experiements Using Scaling Analysis

This paper proposes that the generalization and communication of welding simulations

and measurements can be greatly helped with the use of relatively simple expressions

based on the dominant factors and correction factors based on secondary factors.

A six-step framework is described to achieve the proposed results rigorously, yet

simultaneously being flexible. The final results are simple, general, and accurate

expressions valid in a well delimited range of process parameters. The accuracy of

expressions is of the order of scatter in the welding experiments or simulations. Examples

of the application will be presented for the complex case of coupled heat transfer and

plastic deformation in friction stir welding, micro electron beam welding, and other

simpler problems.

This approach to complex multicoupled, multiphysics systems is complementary with the

more popular pursuit of seamless integration of numerical models into comprehensive

simulations of a system at large. The need for the approach proposed is that even if

perfect simulations were available, it would be of limited help at the most influential

stage of design: the concept design stage of radically new welding systems.

At the conceptual stage, many competing alternatives are evaluated, and might (should!)

differ widely in their working principles, configuration, size, operating parameters,

materials used, and more. It is not possible to understand deeply each alternative and to

create predictive models for each case. Constraining the alternatives to existing models

can exclude desirable alternatives.

Welding processes involve many of the issues of thermofluids plus electromagnetic body

forces, chemical reactions, phase transformations, and complex free surface conditions.

Technologies based on so many physical phenomena are difficult to study, whether

numerically or experimentally. It is likely that in a relatively near timeframe appropriate

integrated simulations will exist, including elaborate sensitivity analyses based on neural

networks and genetic algorithms.