Using surrogate models to predict nodal results for fatigue risk analysis

Abstract

Fatigue life analysis and FEA require high computational costs for aerospace models. Goodman diagrams are useful tools for determining the fatigue life of a part, and depend on structural and modal FEA simulations. Surrogate models have been used in fatigue analysis to predict the steady stresses and alternating stresses FEA model for every par node enabling a dynamically-updated Goodman diagram that responds to design parameter changes in real time. This research analyzes a jet engine compressor blade, and compares the accuracy of these nodal surrogate methods to previous surrogate methods where only a single value is predicted for failure analysis. Quantitative data summarizing the error between the single value and full field models are presented for the transonic Purdue blade.