Abstract:We present a novel certified model order reduction (MOR) algorithm for switched descriptor systems applicable to large-scale systems. Our algorithm combines the idea of [Hossain \& Trenn, Technical report, 2023] to reformulate the switched descriptor system as a switched ordinary differential equation with jumps and an extension of the balanced truncation for switched ODE from [Pontes Duff et al., IEEE Trans.~Automat.~Control, 2020]. Besides being the first MOR method for switched descriptor systems applicable to the large-scale setting, we give a detailed numerical analysis by incorporating the error in the computation of the system Gramians in the a-priori error bound for the output of the reduced system. In more detail, we demonstrate, theoretically and numerically, that the standard error bound is not applicable, and a certificate must account for the numerical approximation errors.
What problem does this paper attempt to address?
### What problem does this paper attempt to solve?
This paper aims to solve the model order reduction (MOR) problem of large - scale switched descriptor systems. Specifically, the paper proposes a new and verified MOR algorithm, which is suitable for large - scale switched descriptor systems. The following are the main objectives and contributions of this research:
1. **Propose a new MOR algorithm**:
- This algorithm combines the method of [Hossain & Trenn], reformulating the switched descriptor system as a switched ordinary differential equation (ODE) with jumps, and the extension of the balanced truncation method proposed by [Pontes Duff et al.] on the switched ODE.
- This is the first MOR method that can be applied to large - scale switched descriptor systems.
2. **Error analysis and verification**:
- The paper analyzes in detail the influence of the error of the system Gramian in numerical calculations on the a priori error bound.
- Theoretical and numerical results show that the standard error bound is not applicable, and the error bound must be corrected by considering the numerical approximation error (see Theorem 4.14).
3. **Solve the limitations of existing methods**:
- Existing methods are usually limited to small - scale or medium - scale problems and lack error verification. The method in this paper can be effectively applied in large - scale settings by introducing the numerical solution method of the generalized Lyapunov equation and the subspace projection framework.
4. **Theoretical and practical verification**:
- The effectiveness and feasibility of the new method are theoretically proven.
- The application effect of this method in the constrained mass - spring - damping system and the discrete Stokes equation is verified through numerical experiments (see Section 5).
### Key formulas involved
- Representation form of the switched descriptor system:
\[
\Sigma_q \begin{cases}
E_q(t) \dot{x}(t) = A_q(t)x(t) + B_q(t)u(t), \\
y(t) = C_q(t)x(t),
\end{cases}
\]
where \( q: \mathbb{R} \to J := \{1, \dots, M\} \) is the external switching signal.
- Generalized Lyapunov equation:
\[
AP + PA^T + \sum_{j = 1}^{M} (F_j P F_j^T + B_{\text{diff}, j} B_{\text{diff}, j}^T) = 0,
\]
\[
A^T Q + QA + \sum_{j = 1}^{M} (F_j^T Q F_j + C_{\text{diff}, j}^T C_{\text{diff}, j}) = 0.
\]
- Construction of the projection matrix:
\[
V = S U_1 \Sigma^{-1/2}_1, \quad W = R V_1 \Sigma^{-1/2}_1,
\]
where \( H = S T R = [U_1, U_2] [\Sigma_1, 0; 0, \Sigma_2] [V_1, V_2]^T \) is the Hankel matrix.
### Summary
This paper solves the model order reduction problem of large - scale switched descriptor systems by proposing a new MOR algorithm, and provides detailed numerical analysis and error verification, ensuring the reliability and effectiveness of the algorithm in practical applications.