Numerical Simulation of a Rolled Joint Formation Process

Abstract

The installation of pressure tubes in nuclear reactors involves rolling the inside of the pressure tube to deform it plastically against an end fitting to create an interference fit. Such a connection necessarily involves residual stresses upon which operating loads are applied. An understanding of the rolling process, and quantitative knowledge of the residual stresses produced by it, are important factors in establishing the integrity of the rolled joint under operating conditions. The purpose of this work is to develop an accurate numerical model to simulate the rolling process, and obtain the residual stress distribution caused by rolling. Previous work demonstrates that while 2-D models are inadequate, 3-D finite element modelling of the rolling process can practically determine the residual stress distribution. The tube, hub and five rollers are all modelled as separate components, connected by sliding contact surfaces. Friction at the tube/hub interface is required to keep the tube in place during rolling. Also, with a five roller tool, and the use of periodic symmetry, only one fifth of the geometry needs to be modelled. Modifications are made to an inhouse analysis code to include friction in the contact algorithm, and a general periodic symmetry method is implemented to permit finer discretization of the geometry. The contact friction model, and the periodic symmetry method represent new advances in the finite element technique that significantly improve the accuracy and efficiency of simulating the rolling process. Results obtained by simulating the complete rolling process (rolling + ironing + retraction) indicate that primary features of rolling are modelled accurately. Predicted extrusion of the tube compares well with measurements, and the computed residual stress distribution, for radial and hoop stress components, is consistent with an interference fit. Quantitatively, predicted residual stress is inside the wide scatter band of measured values, but most measurements have been made on joints with a variety of different parameters. Hence, a definitive quantitative comparison of computed and measured residual stresses is not complete. In addition, the rolling process involves repeated load reversals which require a kinematic hardening material model. Future work will focus on better material models for high plastic strains, finer discretization, and the study of various geometries and friction coefficients.