A Reduced Order Based Metamodel for Parametric Computational Studies of Local Post Weld Heat Treatment
Please login to view abstract download link
Post-welding heat treatment (PWHT) is an important process for nuclear industries, especially for welded big components, aiming to relax residual stresses and stabilize microstructural effects. Controlling the parameters of PWHT is crucial to relax the residual stresses in the target zone while reducing the impact in other parts of the pieces. Finite element simulations of the PWHT can offer a suitable framework to understand the effect of the underlying parameters on the mechanical components response. However, Direct Numerical Simulations (DNS) is not suitable for such application because of the non-linearity (elasto-visco-plastic material behaviour) and the high dimensionality aspects for PWHT parametric studies. In this study, we propose to merge FE simulations with a posteriori Reduced Order Model (ROM) in order to build a metamodel able to reproduce the 4D space-time mechanical fields of the PWHT process. Real time simulations are then posible without any need to re-run DNS for parametric studies. Due to the high temperature character of the studied process, an elasto-visco-plastic material behavior1 for ferritic low alloy steels is used for the simulations. DNS simulation results stemming from an optimized design of experiment covering the dimensions of the physical parameter space are firstly generated. The metamodel is then trained using the sparse High Order Proper Generalized Decomposition2 (sHOPGD) builder which has shown good accuracy for parametric numerical simulations of welding processes. The metamodel is then applied as a parametric PWHT simulations tool in order to test the influence of the chosen parameters.
complas_sec@cimne.upc.edu/ Telf. + 34 - 93 405 46 94
Copyright © 2022 CIMNE, All Rights Reserved. Terms of service
