Study of fatigue behaviour of composite patch repaired aluminium panels using artificial neural networks
This study investigates the enhancement of the fatigue resistance of aluminium panels using composite patch repairs. Aluminium panels are widely used owing to their light weight, alloyability and higher resistance to fatigue. CFRP, with its high strength-to-weight ratio, offers an ideal solution for reinforcing cracked aluminium panels. The research concentrates on assessing the fatigue life of the Aluminium panel with composite patches of different geometries through the LEFM approach. The simulations are performed in ANSYS to determine the Stress Intensity Factor values, and the obtained values are used in calculating the fatigue life of the given specimen. Calculations based on different crack lengths and stress ratios have been conducted. Further, in order to prepare a dataset, DOE has been used along with polynomial regression. For each patch configuration, a dedicated multi-layer perceptron ANN model with two inputs (crack length and stress ratio) was developed in MATLAB for prediction of fatigue life. The validation of the ANN model has been done with the data obtained by testing an aluminium specimen with different shapes of patches. Among the geometries, the trapezoidal patch exhibited the best performance, with an average increase of 184% in fatigue life, and the experimental validation confirmed the results predicted by the ANN with a maximum deviation of 3% for fatigue life, showing the ANN’s capability in fatigue applications. The model greatly reduces prediction time compared to conventional FEA, providing a framework for optimising repairs.
Authors
- Srinivasa Prasad K S
- Abhishek R
- Prathita Kumar M
- Sharath Vamsi Reddy A
- Sadashiva R
Institutions
- PES University (IN)
Publication Details
- Journal
- Australian Journal of Mechanical Engineering
- Published
- 2026-11-13
- DOI
- https://doi.org/10.1080/14484846.2026.2762261
- Primary Topic
- Fatigue and fracture mechanics
- Type
- article
- Field-Weighted Citation Impact
- 0.00