Session: MF-03-02 Welding Residual Stress and Distortion Simulation and Measurement - 2

Paper Number: 105200

105200 - Application of Proper Generalized Decomposition to Sensitivity Analysis of Residual Stress in Dissimilar Metal Weld of Nuclear Power Plant 

The residual stress of weldment is well-known as one of the main causes of stress corrosion cracking in the piping system of nuclear power plants. So, it is important to estimate the residual stress during the design and repair process. Even though there are various methods to measure residual stress, it is still challenging to decrease the variation between measured results. This is why we depend on finite element analysis (FEA) for the integrity of weldment by residual stress. Therefore, verifying the results of FEA with sensitivity analysis is necessary.

There are various parameters affecting residual stress in the welding process. Significant computational costs are involved for sensitivity study for these parameter effects in the welding process. Therefore, for effective sensitivity analysis of these parameters on residual stress, we have introduced the PGD method which is one of the dimensional reduction techniques. For this, the 2D model of a dissimilar metal weld of a pressurizer surge nozzle was selected, and which model was used for international round-robin analysis (NUREG-2228). The modeling was performed with respect to the requirement of guidelines in NUREG-2228, and the FEA software SYSWELD (ESI group) was used for analysis. The welding parameters used in PGD are heat energy, welding velocity, inter-pass temperature, yield strength, and hardening model. As a result, we produced the correlation and sensitivity results of the input variables in real time effectively through the application of the PGD model. These results can be used to validate the FEA analysis and applied to analytic model results.

Presenting Author: Jin-Gyum Kim Korea Institute of Nuclear Safety

Presenting Author Biography: Jin-Gyum Kim received M.S. degree in mechanical convergence engineering at Hanyang university, Korea in 2015. He is working now in the department of nuclear safety research at Korea Institute of Nuclear Safety. His interest includes finite element analysis and non-destructive evaluation for material degradation monitoring using machine learning.

Authors:

Jin-Gyum Kim Korea Institute of Nuclear Safety
Sung-Sik Kang Korea Institute of Nuclear Safety
Changsik Oh Korea Institute of Nuclear Safety
Jangho Ahn ESI Group