Digital Twin for Fatigue Analysis

Abstract

As part of fatigue from finite elements (FE) analysis inputs (geometry, material, and loading), the stress level in a component is a key input for representativity of simulation. To illustrate, assuming a simplified power law between stress and life in excluding all plasticity, sequence effect and any other non-linearities, a small variation in the stress input value may propagate into a large variation in the output simulated life.To reduce these epistemic uncertainties, correlation with strain gages is a fundamental added value to any FE model, as it permits to validate mesh convergence, model, stiffness, and boundary conditions. The construction of a robust digital twin heavily benefits from this important tool for further usage and to improve predictability.The current work shows how to extract insights from strain gages measurements. First of all, improving FE/test correlation may be helped by virtual strain gage analysis. However, strain gage measurement accuracy highly relies on strain gage position and orientation. Consequently, the optimization of sensors positions and orientations, regarding representativity of each independent load case, will reinforce Test Engineer expertise in his test plan specification. Then, the core subject of this paper depicts the load reconstruction process, which means how to get load histories from both strain gage measurements and independent unitary FE load cases, in quasi-static and dynamic framework. As a result, obtaining input loads is a way to extrapolate stress level and life results from discrete points (strain gages) to full component, accessing life results in hot spots with large stress concentration.The theorical background will be described and the methodology on an application case illustrated.