Experimental and numerical investigation on strain-based structural health monitoring of impacted composite stiffened panels

This study investigates a structural health monitoring method for impacted composite stiffened panels based on experimental data and the random forest (RF) algorithm. Various types of impact damage are introduced into the composite stiffened panels. An RF model is then developed, utilizing the strain data to accurately detect and identify the damage that traditional methods, such as visual inspection, struggle to recognize. A damage-equivalent methodology is employed to construct a finite element model (FEM). Based on the experimental results, the feasibility of using the FEM as a supplement to experimental data in training machine learning models has been substantiated. Furthermore, FEMs are developed for entirely new types of damage, and the trained machine learning model proficiently identifies the presence and specific type of damage. This research underscores the robustness and adaptability of strain-based SHM, providing a dependable and meticulous approach for assessing structural integrity.