Abstract:This paper is concerned with the design of a non-intrusive model order reduction (MOR) for the system of parametric time-domain Maxwell equations. A time- and parameter-independent reduced basis (RB) is constructed by using a two-step proper orthogonal decomposition (POD) technique from a collection of full-order electromagnetic field solutions, which are generated via a discontinuous Galerkin time-domain (DGTD) solver. The mapping between the time/parameter values and the projection coefficients onto the RB space is approximated by a Gaussian process regression (GPR). Based on the data characteristics of electromagnetic field solutions, the singular value decomposition (SVD) is applied to extract the principal components of the training data of each projection coefficient, and the GPR models are trained for time- and parameter-modes respectively, by which the final global regression function can be represented as a linear combination of these time- and parameter-Gaussian processes. The extraction of the RB and the training of GPR surrogate models are both completed in the offline stage. Then the field solution at any new input time/parameter point can be directly recovered in the online stage as a linear combination of the RB with the regression outputs as the coefficients. In virtue of its non-intrusive nature, the proposed POD-GPR framework, which is equation-free, decouples the offline and online stages completely, and hence can predict the electromagnetic solution fields at unseen parameter locations quickly and effectively. The performance of our method is illustrated by a scattering problem of a multi-layer dielectric cylinder.
What problem does this paper attempt to address?
### What problems does this paper attempt to solve?
This paper aims to design a non - invasive model order reduction method (Model Order Reduction, MOR) to effectively solve the parameterized time - domain Maxwell's equations. Specifically, this research attempts to solve the following problems:
1. **Computational cost issues in high - fidelity electromagnetic field simulation**:
- In electromagnetic radiation/scattering problems, the traditional full - order model (Full - order Model, FOM) requires a large number of degrees of freedom (Degrees of Freedom, DOFs) to ensure the accuracy of the solution. This leads to a huge computational burden, especially in terms of CPU time and memory.
- When the parameters change, the cost of repeatedly solving the high - fidelity model is too high, making it difficult to perform multiple simulations.
2. **Effective solution of parameterized electromagnetic scattering problems**:
- The research objective is to develop a method that can quickly and accurately predict the electromagnetic field distribution under different parameter conditions. These parameters may include different frequencies, incident directions, or material properties.
- By introducing non - invasive model order reduction techniques, the limitations of traditional Galerkin - projection - based methods in dealing with complex nonlinear problems are avoided.
3. **Efficient modeling combining POD and GPR**:
- Use Proper Orthogonal Decomposition (POD) to extract a low - dimensional basis (Reduced Basis, RB), and perform regression modeling on the projection coefficients through Gaussian Process Regression (GPR).
- This combination enables the rapid recovery of the electromagnetic field solution at new time/parameter points in the online stage without having to re - solve the complex full - order model.
4. **Challenges in handling large - scale snapshot matrices**:
- To address the problem of overly large snapshot matrices in time - domain electromagnetic problems, a two - step POD algorithm is proposed, which effectively reduces the computational cost.
- By decomposing the time mode and the parameter mode, the difficulties of global GPR in the time - parameter domain are resolved.
### Overview of the solution
This paper proposes a non - invasive model order reduction framework based on POD and GPR. The specific steps are as follows:
1. **Offline stage**:
- Use the discontinuous Galerkin time - domain method (Discontinuous Galerkin Time - Domain, DGTD) to generate the full - order electromagnetic field solution and construct the snapshot matrix.
- Apply the two - step POD algorithm to extract the time - independent and parameter - independent reduced - order bases (RB).
- Use SVD to decompose the time mode and the parameter mode of the training data and train GPR models respectively.
2. **Online stage**:
- For any new time/parameter point, quickly obtain the projection coefficients through the trained GPR models.
- Use these coefficients and the reduced - order bases to linearly combine to obtain an approximate electromagnetic field solution.
Through this method, this research achieves effective solution of the parameterized time - domain Maxwell's equations, significantly reduces the computational cost, and maintains high accuracy at the same time.
### Summary of mathematical formulas
- **Maxwell's equations**:
\[
\begin{cases}
\mu_r\frac{\partial H}{\partial t}+\nabla\times E = 0, & \text{in }\Omega\times T, \\
\epsilon_r\frac{\partial E}{\partial t}-\nabla\times H = 0, & \text{in }\Omega\times T,
\end{cases}
\]
where \(\Omega\) is the spatial domain, \(E\) and \(H\) are the electric field and the magnetic field respectively, \(T=[0, T_f)\) is the time interval, and \(\epsilon_r\) and \(\mu_r\) are the relative permittivity and permeability respectively.
- **Boundary conditions**: