Behavior modeling for a new flexure-based mechanism by Hunger Game Search and physics-guided artificial neural...

Area of Science:

• Mechanical Engineering
• Computational Science
• Artificial Intelligence

Background:

• Compliant mechanisms are vital for precise positioning systems but face modeling challenges like unstable results and limited training data.
• Existing research has not explored meta-heuristics for optimizing neural network parameters in compliant mechanism modeling.
• Physics-Guided Artificial Neural Networks (Phy-GANNs) show promise but haven't been applied to compliant mechanisms.

Purpose of the Study:

• To develop a novel approach for modeling compliant mechanism behavior.
• To address limitations in existing modeling techniques, including unstable results and data dependency.
• To integrate meta-heuristics optimization with advanced neural network architectures for improved performance.

Main Methods:

• The study pioneers the use of the Hunger Game Search (HGS) algorithm for optimizing neural network weights and biases.
• Physics-Guided Artificial Neural Networks (Phy-GANNs) are employed, incorporating physical principles into the neural network model.
• A target function is minimized, leveraging both physical insights and data-driven information.

Main Results:

• Hunger Game Search demonstrates superior performance over the backpropagation method, yielding smaller modeling errors.
• Investigations across various training set ratios and ANOVA tests confirm the robustness of HGS.
• Integrating HGS with Phy-GANNs significantly reduces error and accelerates convergence compared to conventional neural networks.

Conclusions:

• The proposed method using Hunger Game Search and Physics-Guided Artificial Neural Networks offers a potent solution for compliant mechanism modeling.
• This approach overcomes previous limitations, providing more stable and accurate results.
• The findings highlight the potential of HGS-Phy-GANNs for advancing compliant mechanism design and application.