Use of creep-plastic analysis methods to predict complex loading cycle crack growth and nucleation behaviour

Abstract

Cracks starting at corner features are the critical consideration for gas turbine manufacturers when demonstrating the damage tolerance of discs to satisfy regulatory requirements. Little has been published on this crack type, however, beyond the derivation of geometry correction factors. This work therefore studies closure, overload and high temperature effects in corner cracks using finite element (FE)analyses, representing the crack tip as a sharp notch and including the effects of plasticity and creep. Meshes for 2D edge and 3D corner cracks were generated using Microsoft Excel macros, enabling models to be built very quickly with precise control of element geometries. Geometry correction factor polynomials were derived for a wide range of corner crack test piece designs that account for the constraints imposed by their threaded ends. The FE work then studied closure, highlighting the role of crack flank plastic strains on the closure mechanics. Closure predictions were compared with experimental Potential Drop voltage-load measurements from selected loading cycles. The evolving stress-distance profiles as the crack grows following an overload were simulated by removing arcs of elements ahead of the crack tip. The predicted durations of the overload effects were in line with classical theory for 2D edge cracks, although in corner crack testing the retardation behaviour persisted for much longer than suggested by the models. At high temperatures, creep crack growth was simulated by again removing arcs of elements, this time when a critical fraction of the rupture life was exceeded. This technique was used to explore the conditions under which creep-dominated time dependent crack growth occurs. Finally, a method was developed for calculating growth rates based on the low cycle fatigue damage accumulated ahead of the moving crack tip. Predictions for different R-ratios agreed well with experimental data, and the application of the model to overloads was demonstrated.