Reduced-order modelling based on Koopman operator theory

Diana A. Bistrian,Gabriel Dimitriu,Ionel M. Navon

2024-09-05

Abstract:The present study focuses on a subject of significant interest in fluid dynamics: the identification of a model with decreased computational complexity from numerical code output using Koopman operator theory. A reduced-order modelling method that incorporates a novel strategy for identifying the most impactful Koopman modes was used to numerically approximate the Koopman composition operator.

Numerical Analysis,Dynamical Systems

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to identify a hydrodynamic model with reduced computational complexity through numerical code output, based on Koopman operator theory. Specifically, the authors focus on extracting a low - dimensional and computationally less - expensive model from complex nonlinear dynamic systems to improve computational efficiency and model accuracy.

### Summary of Main Problems:

1. **Reducing Computational Complexity**: The research aims to extract a model with lower computational complexity from numerical simulation results through Koopman operator theory. This can significantly reduce the consumption of computational resources while ensuring the accuracy of the model.

2. **Modal Decomposition and Reduced - Order Modeling**: Using modal decomposition methods (such as Dynamic Mode Decomposition, DMD), the dominant modes are extracted from high - dimensional data to construct a reduced - order model. This method can effectively capture the dynamic characteristics of the system while reducing unnecessary computational burden.

3. **Application of Koopman Operator**: The Koopman operator provides a tool for transforming a nonlinear dynamic system into a linear system. In this way, it is easier to analyze the frequency - domain characteristics of the system and perform numerical simulations more efficiently.

4. **Improved Modal Selection Strategy**: The authors propose a new strategy to select the most influential Koopman modes. This strategy can automatically select the modes that best represent the system behavior according to the weights of the modes, even if these modes may exhibit rapid changes or low amplitudes.

### Formula Presentation:

- **Definition of Koopman Operator**:
\[
K_t\phi=\phi(F_t)
\]
where \(F_t\) is the flow semigroup of the vector field \(f\).

- **Infinitesimal Generator**:
\[
G_K\phi = \lim_{t\rightarrow0}\frac{K_t\phi-\phi}{t}=\frac{d\phi}{dt}
\]

- **Koopman Eigenfunction**:
\[
G_K\varphi(y) = s\varphi(y)
\]
where \(s\in\mathbb{C}\) is a complex eigenvalue.

- **Koopman Spectral Decomposition**:
\[
\phi(y)=\sum_{j = 1}^{\infty}a_j(\phi)e^{s_jt}\varphi_j
\]
where \(s_j=\sigma_j + i\omega_j\), \(\sigma_j\) is the characteristic decay/growth rate, and \(\omega_j\) is the characteristic frequency.

### Application Examples:

The shallow - water equations (Saint - Venant equations) are used as a test problem in the paper to show how to apply the above methods to construct a reduced - order model and evaluate its performance. By comparing the results of the full model and the reduced - order model, the effectiveness of this method is verified.

In conclusion, this research provides an effective tool for reduced - order modeling of complex hydrodynamic systems, which can significantly reduce computational complexity while maintaining model accuracy.