A Computational Temperature Analysis for Induction Heating of Welded Pipes

Recent approaches to controlling stress corrosion cracking in welded 304 stainless steel pipes have been based on various types of controlled heating procedures. When applied properly, the heating procedure introduces high compressive stresses in region of observed cracking. The compressive stresses are believed to be effective in deterring stress corrosion cracking. One procedure for applying controlled heating to the pipe employs induction heating and is called Induction Heating for Stress Improvement or IHSI. The effective utilization of induction heating requires an understanding of how the induction heating parameters are related to the resulting residual stresses. This paper describes the development of a computational model directed at evaluating the heat densities and temperature distributions for Induction Heating for Stress Improvement (IHSI). The basic mechanism of inducting differs from that of a welding arc in that induction heating produces a distribution of heat sources within the pipe wall while in weld arc heating, the heat source is confined to the surface. Thus the computational model requires two parts. The first part evaluates the induced electrical current and determines the density of heat sources in the pipe wall. The second part of the model uses these heating densities to evaluate the temperature distribution. Temperature dependent properties were found to be important in representing the induction heating phenomenon. However, including temperature dependent properties in the model leads to nonlinear equations which require iterative solution methods for each part of the model. The nonlinear characteristics of the equations also require iterations between the two parts of the model. The model includes the important parameters of the induction heating process and has shown good agreement with temperature data for two different pipe sizes. Because of the inherent nonlinearities in the model and the iterative methods required for general solutions, extensions of the model to improve the algorithimic efficiency are discussed.