Abstract:
This work details a computationally efficient methodology for conducting statistical fatigue and residual strength analysis of complex aircraft structures with multiple cracks, aiming to improve fleet management beyond current over-conservative approaches. The core of the method is a mathematical "splitting scheme" that decomposes a complex three-dimensional problem with multiple cracks into a series of more manageable subproblems: a single global, un-cracked analysis and numerous local analyses of single cracks. These subproblems are solved using STRIPE, a highly scalable program based on the hp-version of the Finite Element Method, which is run on high-performance computing clusters with thousands of processors. The solutions from the subproblems are then linearly combined to accurately and efficiently calculate thousands of stress intensity factor (K) solutions for various crack sizes and configurations with negligible additional cost per configuration. This approach makes large-scale Monte Carlo type statistical studies of 3D multiple-site fatigue crack growth problems computationally feasible, with the ultimate goal of creating the world's largest K-solution database for implementation in crack growth programs like AFGROW.
