Abstract
This research addresses the computational challenges inherent in analyzing large-scale nonlinear dynamical systems. By reformulating Quadratic Dynamical Systems (QDS) using Hadamard products rather than traditional Kronecker products, the authors derive a new algebraic framework—the Quadratic Bilinear Equation with Hadamard products (QBEH). This paper establishes the mathematical foundation for this approach, proving the existence of solutions for the reachability Gramian and demonstrating the convergence of iterative solvers, thereby enabling more efficient Model Order Reduction (MOR) for complex engineering applications.
Introduction
In modern engineering disciplines such as circuit design and fluid dynamics, nonlinear behavior is frequently modeled using Quadratic Dynamical Systems (QDS). As these systems grow in dimensionality ($N$), the computational resources required to simulate, control, or simplify them increase exponentially. A critical tool in managing this complexity is Model Order Reduction (MOR), which seeks to produce a lower-dimensional approximation of the system that preserves its essential dynamics.
Central to many MOR techniques is the calculation of the "Gramian matrix," a representation of the system's energy states. However, traditional QDS formulations rely on Kronecker products to represent nonlinear interactions. The computational cost of these products scales at $O(N^4)$, creating a significant bottleneck for large-scale systems. This paper proposes a structural transformation to mitigate this burden.
The Hadamard Framework
The authors observe that many physical systems, including transmission line circuits containing diodes, possess an implicit matrix structure where nonlinearities can be expressed more efficiently. They propose a shift from standard QDS to a Quadratic Bilinear System with Hadamard products (QBSH).
Unlike the Kronecker product, which significantly expands the dimensionality of the system matrices, the Hadamard product involves element-wise multiplication. This shift reduces the theoretical computational complexity from $O(N^4)$ to $O(N^2)$ for dense matrices. By utilizing Volterra series expansions—a functional method for representing nonlinear systems—the authors map the state vector over time to define a new reachability Gramian specific to QBSH.
The Quadratic Bilinear Equation (QBEH)
The core of the mathematical derivation is the establishment of the fundamental algebraic equation for the system, termed the QBEH:
$$Q(X) = AX + XA^\top + D + MXM^\top + GXG^\top \circ FXF^\top = 0$$
In this equation, $X$ represents the reachability Gramian. The term involving the symbol $\circ$ denotes the Hadamard product. Solving this equation is essential for balancing-type MOR, where the system is simplified by identifying and retaining only the most "reachable" and "observable" states.
Existence and Convergence Proofs
The primary contribution of this work is the rigorous proof of the existence of a solution to the QBEH. The authors provide two key theorems:
- Existence of a Maximal Solution: They prove (Theorem 3.1) that under the condition of a stable state matrix $A$ and the existence of specific bounding matrices, a positive semi-definite maximal solution $X^*$ exists.
- Guaranteed Convergence: The authors propose a fixed-point iterative algorithm to solve the QBEH. They demonstrate (Theorem 3.2) that this iteration converges to the valid solution at a linear rate, provided the spectral radius of the associated linear operator remains less than one.
These proofs ensure that the transition from Kronecker-based models to Hadamard-based models is mathematically sound and that the resulting equations are solvable through standard numerical methods.
Discussion and Limitations
While the Hadamard approach offers a clear "Hadamard Advantage" in terms of computational flops, the authors identify practical challenges for future research. The convergence proof relies on the selection of an initial matrix ($X_0$) that satisfies the inequality $Q(X_0) \leq 0$. In practice, identifying such a matrix is not always straightforward.
Furthermore, while $O(N^2)$ is a significant improvement over $O(N^4)$, truly massive systems may still require specialized sparse solvers or further optimizations to the QBEH iteration process. The effectiveness of balancing-type MOR in this framework remains to be validated across a broader range of empirical physical datasets.
Conclusion
By rethinking the algebraic representation of quadratic interactions, this paper provides a robust framework for the analysis of nonlinear dynamical systems. The derivation of the QBEH and the accompanying existence proofs offer a path toward more efficient simulation and reduction of large-scale engineering models, bypassing the traditional computational barriers of Kronecker-based analysis.
Summary updated