Advancements in the data analysis methodology for creep crack growth tests on type 316H stainless steel

Abstract

Analysis and prediction of Creep Crack Growth (CCG) in alloys such as Type 316H stainless steel relies on correlating the crack growth rate, a ̇, exhibited in an experimental test with the fracture mechanics parameter C^*. Experimental determination of values for C^* relies on knowledge of the creep contribution to the Load Line Displacement (LLD) rate of a specimen, ∆ ̇_c. Established methods of determining ∆ ̇_c use displacement rate partitioning to calculate the elastic and plastic contributions to the LLD rate, ∆ ̇_e and ∆ ̇_p, and subtract them from the total LLD rate, ∆ ̇_T. These methods use expressions based on a Ramberg-Osgood fit to tensile data to determine ∆ ̇_p which were shown to produce significant overestimation because the expressions implicitly assume nonlinear elastic behaviour and do not account for strain history effects, and also because it is difficult to get an accurate power-law fit to tensile data for Type 316H. Elastic-plastic-creep FEA simulations using uniaxial tensile data were recommended as an improvement over the established method to determine separate contributions to the LLD, although ∆ ̇_p was still shown to be significant, indicating that creep was not the dominant deformation mechanism in the sample. These initial simulations assumed continuous crack growth. Further investigation into simulation of discontinuous cracks revealed that separate contributions to the LLD cannot be readily determined as both plastic and creep deformation occur while the crack is stationary, and that if ductile damage is included in such discontinuous simulations then the plastic contribution to the LLD is still dominant. This work also revealed the significant complexity associated with defining a ductile failure locus and highlighted that use of Bridgman expressions to determine triaxiality and strain values from tensile notched bar tests is inaccurate, with the error increasing as the notch radius decreases. Determining accurate values for ∆ ̇_c also requires reliable prediction of creep deformation behaviour. Creep strain rates are often characterised by Norton’s law which relates the strain rate to the applied stress. However, the data used to obtain the fitting parameters for Norton’s law often assume that uniaxial tensile creep tests occur at the engineering stress and that the stress does not change even under load control. A novel creep strain measurement method using image capture showed the importance of characterising strain rates as a function of true stress, and that significant stress variation occurs in a single load controlled creep rupture test for an alloy such as Type 316H. This casts doubt as to whether the creep deformation behaviour can reliably be predicted by Norton’s law. Displacement rate partitioning methods assume that plastic and creep deformation are independent of each other, and that the plastic behaviour is that of as received material even after creep deformation has occurred. Uniaxial creep tests were carried out and interrupted with subsequent quasistatic tensile testing undertaken to attempt to understand the effect of creep strain on the plastic behaviour of the material. Importantly, this showed both that it is not possible to characterise the material state by simple macroscopic parameters such as strain, and that the behaviour exhibited in a quasistatic tensile test post creep is not representative of actual material deformation under constant load at high temperature. It was therefore recommended that future work should focus on characterising the deformation behaviour as time dependent inelastic accounting for micromechanical parameters associated with the internal material state. These findings have significant implications for CCG analysis. Treating the material behaviour as time dependent inelastic eliminates the need for displacement rate partitioning with respect to plastic and creep contributions as they would not exist separately from each other. It could also mean that C^* cannot be determined for short term tests and instead a ̇ should be correlated with a parameter such as J ̇. But given how differently the material behaves under higher loads it is very strongly recommended that attempts are made to replicate plant conditions as closely as possible in experimental tests, even if this does require much longer term testing which will be practically more difficult to undertake.