Data-driven natural frequency tuning in cantilever structures through strategic perforation patterns

The objective of this study is to investigate the dynamic behavior of a cantilever rectangular beam by tuning its natural frequency through a novel perforation technique. The research aims to optimize the number and position of holes in the beam by integrating machine learning models with an optimization algorithm. Three distinct datasets were created to capture both geometric and statistical characteristics of the perforation pattern. Four predictive models were evaluated including linear regression, decision tree, support vector regression, and gradient boosting. The vulture optimization algorithm was then employed to determine the optimal configuration that achieves a target natural frequency. Results show that while models based on the binary and ten-feature datasets performed well on the training data, they encountered significant overfitting and reduced accuracy on unseen data. The derived dataset combined with the gradient boosting model achieved the best performance, with an R² of 0.99 in the training phase and 0.97 in the test phase. This high accuracy is due to the effective extraction of meaningful features that enhance pattern recognition. Example target normalized frequencies of 0.2, 0.6, and 0.9 were used to demonstrate the success of the proposed approach in fine-tuning the perforation configuration to achieve the desired natural frequency.