1998

Talja, Heli

The topic of this thesis is computational aspects in the assessment of ductile failure in metals. The first part briefly describes the micromechanics of ductile crack growth and methods for assessing it. The "classic" approach to describe material behaviour using fracture mechanics is summarised. The limitations of the one parameter approach based on the stress intensity factor K or the J-integral are described. Attempts to extend the application field of fracture mechanics parameters by introducing triaxility or constraint parameters are also presented. Different local approach methodologies are summarised. Special attention is paid to the modified Gurson model, which is based on micro-mechanical studies of void initiation, growth and coalescence. The main part of the work consists of numerical analyses with the modified Gurson model. The parameters of the model are first determined by matching tensile test results by finite element analysis, and then applied to J-R curve prediction. This methodology is applied to several reactor pressure vessel materials: A533B, 20 MnMoNi 5 5 and austenitic VVER-440 cladding. As a result, the applicability of different specimen types for the parameter determination of the modified Gurson model has been evaluated. Because a combination of experimental and numerical work is needed, it proved to be most feasible to use specimens which can be simulated with two-dimensional or axisymmetric finite element models. Further, a practical way to treat anisotropic material behaviour using the modified Gurson model by using separate parameter sets for different orientations has been proposed and verified. The correspondence between the observed scatters in tensile and fracture mechanical test results has been examined. Best agreement was obtained fitting the scatter of tensile tests by varying the values of initial parameters. Reasons for apparently higher ductility measured from sub-sized than standard size tensile specimens were studied by detailed comparison of stress and strain states. The observed differences could partly be clarified. In the case of austenitic VVER-440 reactor pressure vessel cladding material, the transferability of the modified Gurson model parameters proved to be quite limited, which is interpreted to be partly due to different microcrack initiation and competing failure mechanisms in this kind of material. However, lack of some essential experimental data limits the possibility to draw precise conclusions on this. Finally, suggestions for future work are presented. One important application of the modified Gurson model would be to study the specimen size and geometry dependence of J-R curves.