Structural integrity assessment of clad nuclear steel

Abstract

Many of the world's nuclear power plants are reaching the end of their design lives and will have to demonstrate adequate structural integrity to continue operating. Critical components, such as the reactor pressure vessel (RPV), are manufactured from a combination of materials so as to economically provide the required mechanical properties. Typical combinations include ferritic carbon steel and austenitic stainless steels. Generally, these multi-material components are manufactured by welding processes and referred to as dis-similar metal welds (DMWs) consisting of a base (ferrite, grade SA508) and a cladding (austenite, stainless steel grades 308 and 309). Current industry practise often neglects the benefits of the cladding resulting in overly-conservative assessments. The aim of this work is to improve current methods in assessing DMWs by demonstrating how the presence of cladding impacts strain distributions compared to homogeneous materials. The impact of incorporating DMWs into components was investigated by analysing the fracture performance of multi-material (inhomogeneous) samples compared to homogeneous material counterparts for multiple initial defect sizes. The influence of initial defect location with respect to the weld boundary (surface breaking and sub-surface defects) and, the impact of residual stresses were also considered. The original contribution to knowledge is the direct comparison of strain distributions, during fracture toughness testing, for homogeneous and inhomogeneous samples. Results clearly indicate that the distribution of strain in inhomogeneous samples differs significantly compared to homogeneous samples.

Due to the clad welding process the microstructure of the cladding was found to be complex with multiple HAZs resulting in the formation of undesirable microstructure (martensite) and significant variability in the direction of grain growth. Digital Image Correlation (DIC) was used to monitor the development of strain during unloading compliance fracture toughness testing and it was shown that the cladding material absorbs significantly more energy compared to base material samples. Hence the presence of cladding material undoubtedly impacts the structural integrity of a RPV but quantifying that impact is shown to be highly dependent upon initial defect size, type (i.e. surface breaking or sub-surface) and orientation with respect to the weld boundary. Local hotspot regions in the cladding were identified using DIC and indicate that the undesirable and complex microstructure may provide preferential locations for defect initiation and/or propagation due to increased strains in hotspot areas. The mechanical properties and microstructure of the DMW studied in this work were found to be in agreement with literature data.

Standard sample geometries were used to simulate different defect orientations: Sub-surface defects were analysed using Compact Tension (CT) samples while surface breaking defects were analysed using Single Edge Notch Bend (SENB) samples. Sub-surface defects were oriented parallel to the weld boundary and were found to propagate in the lower strength material. This resulted in deviation of the defect from the expected path (i.e. perpendicular to the loading direction) into the softer material. This is in agreement with literature data. Finite element (FE) analyses showed good agreement with experimental data when 2D plane strain conditions or 3D conditions were assumed. A parametric study investigating the impact of cladding thickness indicated that sub-surface defects are less likely to propagate when the thickness of the cladding is increased. It was additionally shown that the proximity of the notch to the weld boundary impacted the strain distribution with notches in closer proximity to the weld boundary being increasingly asymmetrical this was as expected. Surface breaking defects were found to propagate along the expected path (i.e. perpendicular to the loading direction) irrespective of sample type (i.e. defect deviation remained perpendicular to the loading direction for both inhomogeneous and homogeneous samples). The influence of notch proximity to the weld boundary was addressed by varying the initial notch depth. It was found that inhomogeneous samples notched through the thickness of the cladding (i.e. notch tip in the base material) acted in a similar manner to unclad samples. Samples notched in the cladding were found to have strain distributions significantly different to similarly notched homogeneous samples. Shallow notches in the cladding exhibited more influence from the cladding microstructure compared with notches located in the cladding close to the weld boundary. FE 2D plane strain analyses showed good agreement, in terms of strain distribution, with experimental data despite neglecting local microstructural effects. It was further shown that explicitly modelling the martensitic HAZ region significantly impacted the predicted strain distribution.

Residual stresses were measured experimentally and modelled using 3D FE analyses. Two types of residual stress modelling were investigated: block dump modelling (BDM) where the weld was assumed to be laid down in chunks and, a moving heat source model (MHS) where the weld torch was explicitly defined based on known weld procedure data. Neither BDM nor MHS could be shown to accurately describe the experimental data however, both models incorporated an initial assumption that materials used in the analyses were in a `stress free' condition. This assumption was later shown, using experimental results, to be inaccurate. Despite the inaccuracies in initial assumptions residual stresses in individual test samples were shown to be in good agreement with experimental data.