Finite Element Modelling of 5- and 8-Pass Ferritic Steel Welds Using Phase Transformation Material Models

Abstract

The modelling of welds is desirable to predict the distortion of components during manufacture, the position and magnitude of peak residual stresses and to predict metallurgical effects in specific regions. Welds are a complex modelling problem requiring both thermal and structural solutions. This has lead to the development of several weld-specific simulation packages and codes for finite element (FE) analysis. This paper describes the application of phase transformation material models to ferritic groove weld test specimens. These specimens were manufactured from SA508 Grade 3 Class 1 pressure vessel steel plates 200×150×20 mm with SD3 1Ni 1/4Mo weld metal deposited in a groove 10 mm deep. The fifth weld pass in both specimens had two stop-starts introduced to investigate their effect on the residual stress field. The first stop linearly ramped the torch power down before backtracking and continuing the bead. The second stop had the torch abruptly switched off before restarting in the same location. The residual stresses in these specimens were measured using Neutron Diffraction (ND) which has been compared with the FE predictions. The FE modelling used a decoupled thermo-mechanical approach. The VFT-CTSP weld simulation package was used for the thermal analysis and Abaqus 6.8-3 for the mechanical analysis using the VFT UMAT-WELD user subroutine with phase transformation material properties. The thermal results appear to be consistent with the thermocouple traces recorded during manufacture of the plates. The simulated thermocouple temperature peaks are within 10% of manufacturing peaks. The simulated heating and cooling rates closely follow the manufacturing heating and cooling rates. The stresses calculated appear to be similar to the ND results measured on the specimen plates though some suspected errors have to be taken into account. The predicted stress field in the weld bead has a discontinuity as the material within the model changes from SA508 to SD3. This is to be expected due to the slightly different Young’s modulii of the two materials. This effect is present in the FE results due to the inability to model the metal mixing that occurs at the fusion boundary. The ND results were continuous across the fusion zone (FZ) and heat-affected zone (HAZ). The phases predicted appear to be similar to those expected for welds of this type. The martensite formation in the weld metal is consistent with the cooling rates experienced at the stop-start locations. The ramped stop-start had the lower cooling rate and therefore less martensite forms while the abrupt stop-start had a higher cooling rate which produces a larger amount of martensite. The subsequent remelting caused by passes six-eight removes the effects of the stop-start features in the eight-pass plate in the FE predictions.