Journal of Theoretical and Applied Vibration and Acoustics

Journal of Theoretical and Applied Vibration and Acoustics

A hybrid machine learning and particle swarm system for configuring holes on cantilever beams to achieve desired natural frequencies

Document Type : Research Article

Authors
Amir Hossein Rabiee
Amir Mohammad Jalali
School of Mechanical Engineering, Arak University of Technology, 38181-41167, Arak, IRAN
10.22064/tava.2025.2040402.1244
Abstract
In this paper, a hybrid machine learning/optimization system is developed to identify the optimal configuration of holes on a cantilever beam to achieve a desired natural frequency. Based on a design of experiments, 100 configurations are selected from the vast possible combinations of placing five holes on a 5x21 matrix grid over the beam. The natural frequencies for these configurations are obtained using frequency analysis in COMSOL. A dataset containing the hole configurations and their corresponding normalized first natural frequency is constructed to build a machine-learning model using the LightGBM method. The particle swarm optimization algorithm is employed to find the optimal hole configuration that yields the desired natural frequency. The results demonstrate the success of the developed hybrid system, as the machine learning model accurately predicts both the training and testing data. Additionally, the optimization algorithm successfully identifies hole configurations that closely match the desired natural frequency in various test cases, validating the system's effectiveness.

Highlights

  • Hybrid machine learning/optimization is used to identify optimal hole configuration.
  • 100 hole configurations are selected using design of experiments.
  • Natural frequencies are obtained using COMSOL frequency analysis.
  • Light GBM model is used to predict training and testing data accurately.
  • Optimization successfully finds configurations matching desired natural frequency.

Keywords
Frequency tuning
Machine learning
Hole configuration
Frequency analysis
Frequency adjustment
Subjects