Industries such as shipbuilding, aerospace, defense, offshore energy, automotive, and nuclear energy rely heavily on welding. For example, a single automobile can contain 5,000 to 10,000 welds. Yet many engineers consider the practice of welding more art than science. That’s because welding depends on the skill and competence of the individual welder or on the quality of robotic-welder programming.
Welding is also an art because the high temperatures, material-phase transformations, and deposition of materials make it difficult to produce a good weld every time. This challenge has pushed engineers to look for ways to study and predict how different welding techniques will affect the materials being fused. Such knowledge can help avoid design guesswork, speed product development, and make finished products with high quality.
Finite element analysis (FEA) software can help quantify and verify the many parameters involved in welding. Rolls-Royce Marine, Houston, uses FEA packages in the welding of propulsion mechanisms and equipment in marine power plants. “Welds are a complex modeling problem requiring thermal and structural solutions,” says Rolls-Royce Stress Engineer, Primary Components, David Hodgson. “We are looking to predict the distortion of components during manufacture, the position and magnitude of peak residual stresses, and how welding affects different materials.”
Working with the Materials Science Center at the University of Manchester and Serco Assurance, both in the U. K., Rolls-Royce has been evaluating several FEA packages for years, including Sysweld, VFT, and Abaqus FEA from Simulia, Providence, R. I. The company added the Abaqus Welding Interface (AWI) tool to analyze welds in more detail. The software streamlines the generation of 2D welding simulations by providing a GUI for defining all aspects of the weld model, including weld beads, weld passes, film loads, and radiation loads.“The weld-simulation tool is helpful because it lets us further analyze thermal and structural models directly inside Abaqus FEA,” says Hodgson. “This helps us reduce data-translation issues, training time, and expense.”
Research engineers used the AWI tool to examine an autogenously welded (where heat is used without the addition of filler metal) plate, an eight-pass groove-welded plate, and a seven-pass ring-weld disk. The autogenously welded-plate model simulated a single plate of pressure-vessel steel melted by one or two passes of a welding torch. The more complex eight-pass and ring-weld steel models involved multiple torch passes and a ferritic-steel filler material. These are processes that might be used in offshore oil installations or nuclear reactors. Engineers used quadrilateral and triangular heat-transfer elements for the thermal models and a generalize plane-strain model for the structural analysis.
Applied heat and the metal’s structural response to it are inextricably intertwined during welding. But from the simulation point of view, each of these parameters must be defined separately to be used in a multiphysics analysis of their coupled interaction. Through the AWI, the temperature history calculated in the thermal model provides the temperature input (load) for the structural model. The structural model can then predict the thermal expansion and contraction of the metals being welded resulting from changes to materials mechanical properties.
To set up the model, engineers imported a basic meshed part — no boundary conditions, loads, or interactions, but materials and sections were defined. Engineers then created weld beads at the appropriate spots. The software defines weld passes based on the weld bead order and assigns surface film and radiation heat-transfer properties.
“Among the most time-consuming aspects of weld modeling is the surface definitions for heat-transfer coefficients,” says Hodgson. “These definitions have to be constantly updated while the weld build-up modeling progresses. The AWI tool automates updating, which helps speed model building.” To control the heat flux in the thermal model, engineers set a fusion boundary. Sensor nodes in the mesh, defined at specific depths, end the heating step when their average temperature reaches a predetermined limit — 1,500°C in this case.
To verify the models against real-world results, engineers welded small test pieces of the pressure-vessel steel with a mechanized TIG welding head and measured the residual stresses after several welding passes. Measurement took place with neutron diffraction, which involves placing a sample in a beam of neutrons and recording the diffraction intensity pattern resulting from changes in the structure of the crystalline solids in the steel. The pattern records the crystal-lattice spacing in the steel that can be compared with a stress-free spacing value to calculate a strain measurement. The strain measurement can then be used to infer the stress in the metal.
Engineers compared the autogenously welded FEA models with the neutron-diffraction test results. The one-pass results showed a close correlation. Work continues on refining the more-complex models for accurate simulation of the many intertwined factors in welding. “The 2D weld modeling interface is useful and we are looking forward to its evolution into a full 3D moving heat source with yet more-advanced weld material modeling capabilities,” says Hodgson.