Artificial Intelligence-Based SHM Approach for Fatigue Damage Assessment

Abstract

Periodic inspections are common to find flaws in aircraft structures. Continuous monitoring of the structural integrity using Structural Health Monitoring (SHM) is desirable, as undetectable damage can occur at any time during a structure’s lifespan. Numerous SHM technologies have been developed to identify and to evaluate defects in structures. Using data-driven analysis methods, such as artificial intelligence (AI), information about the condition of structures can be extracted from large amounts of complex measured data. However, current AI-based damage assessment methods are often limited to simple, well-defined structures. This paper presents an AI-based approach for damage assessment on a full-scale fuselage panel with a complex substructure. An aluminum panel tested by the Federal Aviation Administration (FAA) at the William J Hughes Technical Center is used to explore new applications of artificial neural networks (ANN) by different input scenarios and classification concepts, for damage detection and damage level prediction. Initial damage is created by a sawcut. The panel is fatigued through cyclic loading, causing the crack to propagate, providing comprehensive sensor data responses due to the crack growth. The experimental setup involves Acellent Technologies piezoelectric (PZT) sensors symmetrically placed on opposite sides of the sawcut, capturing pitch-catch signals. The ANN model utilizes damage indices derived from the sensor signals to detect and assess the damage. This research highlights the potential of AI-based approaches in enhancing the capabilities of SHM systems, paving the way for more reliable and efficient structural monitoring.