Application of Machine Learning Models in Predicting Vibration Frequencies of Thin Variable Thickness Plates

Area of Science:

• Computational Mechanics
• Machine Learning Applications
• Structural Dynamics

Background:

• Predicting vibration frequencies of thin rectangular plates with variable thickness is crucial for structural analysis.
• Traditional methods like finite element (FE) analysis are computationally expensive due to repeated eigenproblem solutions.
• Genetic algorithms used in optimization exacerbate the computational burden.

Purpose of the Study:

• To investigate the application of machine learning (ML) techniques for predicting vibration frequencies of thin rectangular plates.
• To develop an efficient surrogate model using artificial neural networks (ANNs) to reduce computational costs.
• To integrate ML into structural optimization workflows for enhanced efficiency.

Main Methods:

• A dataset was created with variations in plate geometry, boundary conditions, and thickness distribution, encoded numerically.
• Artificial neural networks (ANNs) were employed as a surrogate model.
• Systematic tuning of ANN architecture and hyperparameters (hidden layers, neurons, activation functions) and data preprocessing (standardization, scaling) were performed.

Main Results:

• ANNs demonstrated high prediction accuracy for plate vibration eigenvalues.
• The ML approach significantly reduced computational effort compared to FE analysis.
• Root Mean Square Error (RMSE) and R-squared (R²) metrics confirmed model performance.

Conclusions:

• Artificial neural networks provide an accurate and computationally efficient alternative for predicting vibration frequencies of thin rectangular plates.
• This ML-based surrogate modeling approach offers a practical solution for optimizing structural analysis workflows.
• The study highlights the potential of integrating machine learning into structural engineering for reduced computational expense.