Low-Energy Impact Characterization Models and Damage Detection in Thin Monocoque Structures Using Artificial Intelligence

Abstract

The low damage tolerance of aircraft structures leads to an oversizing of their optimum weight as they are designed to withstand impacts on aerodynamic surfaces. Through the application of Structural Health Monitoring (SHM) techniques, which allow the characterization of impacts with a reduced number of sensors, further structural optimization is possible. This article proposes the use of Artificial Intelligence (AI) models for a complete impact characterization performed on an anisotropic Carbon Fibre Reinforced Plastic (CFRP) plate. The three principal objectives are impact location, characterization of the Impactor Object (IO), focusing on splitting the mass and velocity at equienergy impacts, and damage detection. The location task consists of the prediction of the impact coordinates (X-Y), while the energetic characterization of the IO predicts its mass, velocity, and energy when the impact occurs. These models are powered by low-cost piezoelectric (PZT) sensors, which acquire the acoustic wave generated by the impact, which allows the monitoring of large surfaces of complex geometry with a reduced number of sensors. These models have been trained with experimental data acquired with an Autonomous Impact Machine. This machine has performed more than 40,000 impacts on a coordinate grid that vary the IO mass, velocity, and impact energy, focusing on multiple equal energy combinations of the IO mass and impact velocity. Damage detection is performed by comparing the predictions of multiple location and characterization models.