Numerical modelling and experimental study of the welding process in P91 martensitic steel

P91 is a modified 9Cr martensitic steel which is widely used in the power generation industry in components which experience high temperature. In welded P91 components, the steep temperature gradients in the weld and heat affected zone (HAZ) give rise to residual stress during cool-down. In this thesis P91 martensitic steel welds are studied using a combination of computational and experimental methods. Bead-on-plate experiments are carried out by deposition of a single bead on a plate of P91 martensitic steel with the manual metal arc welding process (MMA). 

The residual stress distribution in the welds is evaluated by neutron diffraction (ND). The ND measurements are carried out at three different instruments (Engin,X, E3 and SALSA). A comparison is made between the measurements at the three instruments showing a good agreement between the measured profiles. A further comparison is made for welds manufactured with different welding conditions. In the first case two welds manufactured with similar heat input and different welding parameters (welding speed and weld current) are compared regarding the stress distribution. The two welds present a similar distribution of stress. In the second case a comparison is made between the stress distributions in two welds manufactured with different welding speed, resulting in different heat input. In this case, the stress profiles are similar with major differences in the longitudinal stresses of the Fusion Zone (FZ). The results of ND measurements in P91 martensitic steel are also compared with respective measurements from literature carried out on 316L stainless steel. In both studies, the stress distribution and magnitude are similar. 

An uncoupled finite element analysis is carried out with the temperature field solved first, and the stress calculation follows. In the thermal analysis, a heat flux with ellipsoid distribution is used to model the heat source. The heat source parameters and the welding efficiency are determined through trial-and-error analyses, matching the predicted temperatures with the respective recorded experimental temperatures. The results show that the geometry of a bead-on-plate and thermal boundary conditions can affect the cooling rate of the weld. An accurate prediction of the thermal history at the most reliable thermocouple is achieved by adjusting the welding efficiency. The predicted temperatures at thermocouple positions close to the weld are similar to the temperature histories of thermocouples that were not exposed to the arc shine of the weld. 

The temperature field is also used to predict the average prior austenite grain size (PAG) in the heat affected zone. The results are compared with reconstructed PAGs, measured by EBSD and analysed using after the toolbox MTEX in MATLAB. A parameter study of the grain growth prediction shows that the model is sensitive to the grain growth activation energy and the heat input parameters. 

The temperature field is subsequently used to develop a mechanical model predicting the residual stress distribution in the welds. The effect of induced strains due to martensitic transformation is considered in the model using a user defined subroutine. The results of the mechanical analysis shows that the metallurgical strain introduced due to the martensitic transformation reduces the stress in the HAZ and the FZ and it is compensated by transformation plasticity. A parameter study is carried out to investigate the effect of the yield strength, thermal expansion coefficient, hardening modulus and hardening model on the residual stress. The stress prediction of an isotropic hardening model is similar to the ND measurements of the bead-on-plate weld. In general, the measured and predicted residual stress present a reasonable agreement within the observed experimental differences. 

A post weld heat treatment (PWHT) is carried out to reduce and redistribute the stress in the bead-on-plate weld. The post weld heat treated specimen is measured by neutron diffraction and the results are compared with respective measurements in the as-welded condition. A creep step is added to the mechanical analysis, to model the PWHT and investigate the effect of creep parameters on the prediction of stresses and the evolution of stress respective to time. The application of the PWHT reduced significantly the stress in the weld which is also predicted by the creep model. The effect of the PWHT in the analysis depends on the magnitude of the residual stresses in the as-welded condition according to Norton’s law. A parameter study showed that the creep model applied in the PWHT step of the analysis is sensitive to the creep constants.