Dynamic strain mapping and real-time damage-state estimation under random fatigue loading

This paper presents a passive sensing technique for real-time structural health monitoring. In the proposed technique, two reference or healthy state dynamic models are estimated using a nonlinear statistical pattern recognition technique known as the Gaussian process. The nonlinear dynamic models are estimated by mapping real-time dynamic strain measurements at two different locations with corresponding load histories. The strain measurements are performed using strain gauge rosettes placed on opposite sides of a probable damage path. Using the estimated reference model with the new loading information at any given instant of time, the corresponding dynamic strains are predicted. The predicted strains are compared with the actual strain measurements and used to estimate the corresponding damage states via correlation analysis. The state estimation approach is demonstrated on an aluminum-2024 cruciform specimen fatigued under random biaxial loading. The numerical study shows that the proposed approach can estimate real-time damage states caused by both stage II and stage III fatigue cracks.