Reduced-cost microwave modeling using constrained domains and dimensionality reduction.

Area of Science:

• Electromagnetics
• Computational Engineering
• Microwave Engineering

Background:

• Modern microwave device development relies on full-wave electromagnetic (EM) simulations, which are computationally expensive.
• Existing surrogate modeling methods, particularly data-driven techniques, face challenges with nonlinearity, high dimensionality, and broad operating parameter ranges in real-world microwave components.
• Performance-driven modeling offers partial solutions but requires costly initial database acquisition.

Purpose of the Study:

• To develop a novel, cost-efficient framework for modeling microwave components that overcomes the limitations of existing methods.
• To reduce the computational expense and setup cost associated with surrogate modeling for microwave devices.
• To enhance the applicability of surrogate modeling for complex, real-world microwave components.

Main Methods:

• Replaced traditional database designs with random observables for surrogate model domain definition.
• Implemented explicit dimensionality reduction of the surrogate model domain using spectral analysis.
• Developed a complete cost-efficient framework integrating these novel techniques for microwave component modeling.

Main Results:

• The introduced framework demonstrates excellent predictive power of the generated surrogate models.
• The framework exhibits favorable scalability properties for complex microwave components.
• Achieved low computational overhead during the model setup phase, significantly reducing costs.

Conclusions:

• The proposed framework offers a cost-efficient and effective solution for surrogate modeling of microwave components.
• Explicit dimensionality reduction and the use of random observables are key innovations for improving surrogate modeling.
• This approach enhances the practical utility of data-driven modeling in microwave engineering by addressing computational and data acquisition challenges.