Finite Element and Machine Learning-Based Prediction of Buckling Strength in Additively Manufactured Lattice Stiffened Panels

The current research aimed to investigate the critical buckling load of a simply supported aerospace-grade stiffened panel made of additively manufactured cubic lattice unit cell arrays, namely simple cubic, face-centered cubic (FCC) and body-centered cubic (BCC) structures. Ansys Design Modeler was chosen to design and analyze the critical buckling load of the panel, while a popular material, Ti-6Al-4V, was used as the build material. Numerical validation on both the stiffened panel and a lattice beam structure was established from multiple resources from the literature. Finally, the panels were tested against increments of a strut diameter ranging from 0.5 mm to 2 mm, which corresponds to a relative density of 6% to 78%. It was found that considering the relative density and fixed relative density, the simple cubic lattice cell outperformed the buckling results of the FCC and BCC panels. Moreover, the relationship of the parameters was found to be non-linear. Finally, the data samples collected from numerical outcomes were utilized to train four different machine learning models, namely multi-variable linear regression, polynomial regression, the random forest regressor and the K-nearest neighbor regressor. The evaluation metrics suggest that polynomial regression provides the highest accuracy among all the tested models, with the lowest mean squared error (MSE) value of 0.0001 and a perfect R2 score. The current research opens up the discussion of using cubic lattice cells as potential structures for future stiffened panels.