Adaptive Moving Finite-Element Simulation for the Temperature Distribution in Arc-Welding Process.

Abstract

The purpose of this study is to develop a generalized finite element simulation for the transient heat flow in welded thin plates. The moving finite element method is employed to obtain consistent accuracy throughout the entire period of simulation with a reasonable computer time. A rezoning scheme is employed for transforming data from grid nodal points at any time step to the corresponding grid nodes at the subsequent step. To improve the accuracy of the simulation, the finite element grid is adaptively optimized using the R-method nodal redistribution technique. To obtain the temperature field at each time step, the set of finite element ordinary differential equations is solved using the Crank-Nicholson time integration schemes. In the time stepping scheme, the phase change effect is analyzed by an algorithm which includes 3D fusion pool geometrical analysis, and incorporated with the heat-sink or heat-source model to account for the latent heat absorption-liberation process. For an arc power of 10.7 KW and welding speed of 125 cm/sec, a temperature field in the range of 250-900(DEGREES)C was predicted for points located within 5-20 mm of the weld centerline, with the peak temperature at the centerline being about 1400(DEGREES)C. For identical welding conditions and locations, except for a welding speed of 1 cm/sec, the temperature field was in the range of 275-1000(DEGREES)C, with a peak temperature of 21000(DEGREES)C. In both cases, good agreement was found with the middle plane temperature obtained experimentally. Continuous temperature measurements were made using thermocouples to experimentally verify the prediction obtained from the simulation. To minimize the variation between simulated and actual conditions, thin low carbon plates (.25 in thickness) were welded by precisely computer controlled welding variables. Except for locations directly affected by the heat source, the assumed 2-dimensional heat flow model was found to be reasonable. While the analysis is presently designed for 2-dimensional heat flow simulation, the same approach can be extended for the 3-dimensional case. The close agreement between simulated and recorded data suggests that an extension of the study to include thermal stress analysis would be feasible. (Abstract shortened with permission of author.)