Metamaterial absorber optimization method based on an artificial neural network surrogate.

Area of Science:

• Electromagnetics and Materials Science
• Computational Physics and Engineering

Background:

• Optimizing metamaterial absorber design for specific electromagnetic responses remains a significant challenge.
• Existing methods often require extensive computational resources and time.

Purpose of the Study:

• To develop an efficient surrogate-based optimization method for metamaterial absorbers.
• To combine artificial neural networks (ANN) with a trust region algorithm for improved design parameter optimization.

Main Methods:

• A surrogate-based optimization approach using ANN to model the metamaterial absorber's response.
• Incorporation of a trust region algorithm to enhance convergence of the ANN surrogate model.
• Iterative training of the ANN model using solutions from previous iterations and surrounding sample points.

Main Results:

• The proposed method demonstrates quick convergence to target solutions with minimal training data.
• Achieved more efficient optimization of geometric and material parameters compared to three alternative methods.
• Successfully validated through two examples of electromagnetic optimization for metamaterial absorbers.

Conclusions:

• The ANN and trust region algorithm combination offers an efficient and effective solution for metamaterial absorber design.
• This method reduces the computational burden and accelerates the optimization process.
• The approach is validated and applicable to complex electromagnetic optimization problems.