AUGMENTING GENERIC FATIGUE CRACK GROWTH MODELS USING 3D FINITE ELEMENT SIMULATIONS AND GAUSSIAN PROCESS MODELING

Assessing the crack propagation life of components is a critical aspect in evaluating the overall structural integrity of a mechanical structure that poses a risk of failure. Engineers often rely on industry standards and fatigue crack growth tools such as NASGRO [1] and AFGROW [2] to perform life assessment for different structural components. A good understanding of material damage tolerant capabilities, and the component's loading mission during service conditions are required along with the availability of generic fracture mechanics models implemented in the lifing tools. Three-dimensional (3D) linear elastic fracture mechanics (LEFM) finite element modeling (FEM) is also a viable alternative to simulate crack propagation in a component. This method allows capturing detailed geometry of the component and representative loading conditions which can be crucial to accurately simulate the three dimensionality of the propagating crack shape and further determine the associated loading cycles. In comparison to a generic model, the disadvantage of the 3D FEM is the extended runtime. One feasible way to benefit from 3D modeling is to employ it to understand the crack front evolution and growth path for the representative loading condition. Mode I stress intensity factors (K1) along the predetermined crack growth path can be generated for use in fatigue crack growth tools such as NASGRO. In the current study, such a 3D FEM lifing process is presented using a classical bolt-nut assembly, components that are commonly used in engineering design. First, K1 solutions for a fixed crack aspect ratio a/c=1 are benchmarked against a similar solution available in NASGRO. Next, a predefined set of crack shapes and sizes are simulated using 3D FEA. A machine learning model Gaussian Process (GP) was trained to predict the K1 solutions of the 3D model, which in turn was used in the crack propagation simulation to accelerate the life assessment process. Verification of the implemented procedure is done by correlating the crack growth curves predicted from GP to the results obtained directly from 3D FE crack propagation method.