Loading...
1 results
Search Results
Now showing 1 - 1 of 1
- Application of Machine Learning Algorithms to the Study of Fatigue of Materials on Aerospace Structures: The case study of the Portuguese Air Force Epsilon TB-30 aircraftPublication . Barros, Tomás de Oliveira; Gamboa, Pedro Vieira; Infante, Virgínia Isabel Monteiro Nabais; Alexandre, Luís Filipe Barbosa de AlmeidaMilitary aircraft are frequently subjected to severe operating conditions due to demanding maneuvers under a wide range of load factor levels, allowing the growth of fatigue cracks that shall be duly monitored to prevent unexpected failures. In order to monitor the operation of the Portuguese Air Force Epsilon TB-30 fleet which performs basic and elementary piloting training, two systems were installed in these aircraft, allowing load factor data in the aircraft center of gravity and strain data in two critical components to be recorded, the C2 bulkhead beam and the main spar. Firstly, this study aimed to evaluate the growth of fatigue cracks on the C2 bulkhead beam, one of the aircraft critical locations, through experimental tests and numerical simulations. The methodology used in these two approaches comprised the application of real variable amplitude load sequences to a 2024-T351 aluminum specimen that is representative of this critical location. Results from these tests and simulations showed that NASGRO model is the one that comes closest to experimental data with maximum deviations of approximately 8% at the final failure. The recording of a large amount of load factor data in the Epsilon TB-30 fleet allied to the recent developments in data science allowed the application of machine learning algorithms, namely artificial neural networks, to the study of fatigue of materials. So, as the second part of this study, the application of these algorithms focused on aircraft mission classification, crack growth prediction and fracture surface classification through Multi-Layer Perceptron and Convolutional Neural Network, resulting in accuracy, F1 score and R2 values of approximately 90%. Results from the first part of this study can be used by the Portuguese Air Force engineers to adjust the aircraft maintenance program according to the actual operation regime and consider a possible life extension of this fleet. The second part of this study provided alternative and complementary methods to classify missions according to the severity related with fatigue and predict the propagation of fatigue cracks, considering the actual operation regime and the presence of eventual damages. Additionally, it comprises a tool to support Portuguese Air Force engineers in conducting failure investigations through the evaluation of fracture surfaces of broken components.