Welding Residual Stress Modeling Benchmark Comparison Study

Abstract

Welding residual stress modeling is currently performed by researchers around the world using a wide variety of modeling methods to predict the final stress state of a completed weld. Among the key modeling assumptions used to perform a residual stress simulation are: • The geometric setup of the model, including boundary condition and assumptions. • The thermal and structural model simulating the welding process including lumping of weld passes. • The strain hardening input properties and hardening law. Researchers from Dominion Engineering, Inc. (DEI) and Westinghouse Electric Co. (WEC) have performed a benchmark comparison studying these key modeling assumptions and their results on the predicted welding residual stress distributions. Researchers from DEI and WEC have completed independent studies to validate their respective methods for calculating residual weld stress. In addition to the comparative evaluation, brief descriptions of the individual validations will be included in this paper. The weldment selected for evaluation is a typical reactor pressure vessel (RPV) outlet nozzle dissimilar metal safe end weld in a pressurized water reactor plant. This weld joins a low alloy steel nozzle to a stainless steel safe end using Alloy 182 weld material; this weld is completed in the manufacturing shop. The safe end is then field welded to the stainless steel reactor coolant loop piping. The residual stress distributions in the dissimilar metal welds, like the one selected, are important in predicting stress corrosion crack growth in Reactor Coolant System (RCS) components. The fabrication drawings for the selected RPV outlet nozzle were provided to both organizations, and independent residual stress simulations were performed using the best effort modeling techniques from each organization. This paper investigates the impact of the key modeling assumptions described above on the differences in the predicted welding residual stress distributions between the two simulation techniques. The results from the modeling comparison are provided in this paper.