A Mixed-FEM approximation with uniform conservation of the exponential stability for a class of anisotropic port-Hamiltonian system and its application to LQ control.
Luis Mora, Kirsten Morris · 2025
Read A Mixed-FEM approximation with uniform conservation of the exponential stability for a class of anisotropic port-Hamiltonian system and its application to LQ control. by Luis Mora, Kirsten Morris (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-16388.txt.
Introduction
Luis Mora and Kirsten Morris present, in a 2025 paper, a way to discretize a class of energy-based systems for computer simulation. The systems, called port-Hamiltonian systems, lose energy at their boundaries. The discretization must preserve that energy loss at any mesh size. If the discrete model decays too fast or too slow, control designs built on it can mislead.
Their mixed finite element method uses two kinds of mesh elements at once. With a multiplier argument they prove the approximation preserves exponential stability uniformly, with a decay rate floor that does not depend on mesh size. A piezoelectric beam serves as the test case. They then show that controllers built on their model converge as the mesh refines, in an infinite-time linear quadratic design, unlike controllers built on the standard method.
Catalog record
| CREATOR | Luis Mora, Kirsten Morris |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 69 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-16388
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Mixed-FEM approximation with uniform conservation of the exponential stability for a class of anisotropic port-Hamiltonian system and its application to LQ control. Luis Mora, Kirsten Morris Abstract In this manuscript, we present a mixed finite element discretization for a class of boundary-damped anisotropic port-Hamiltonian systems. Using a multiplier method, we demonstrate that the resulting approximation model uniformly preserves the exponential stability of the uncontrolled system, establishing a lower bound for the exponential decay rate that is independent of the mesh size. This property is illustrated through the spatial discretization of a piezoelectric beam. Furthermore, we show how the uniform preservation of exponential stability by the proposed model aids in the convergence of controllers derived from an infinite-time linear quadratic control design, in comparison to models obtained from the standard finite-element method. 1 Introduction. The conservation of the exponential stability independently of the mesh size, uniform conservation, is a desired property for the numerical approximation models of distributed-parameter systems. The relevance of this property lies not only in mitigating the effects of spurious oscillations during simulations, as happen in standard finite differences approximations of the wave equation, but also in the conservation of properties as the exact controllability and observability´that are relevant in the design of controllers and estimators [ 29 ] . For example, in the linear quadratic control for partial differential equations with an infinite time horizon, the design is based on numerical approximations of the systems, and the sequence of designed controllers could not converge as the approximation order is increased and may not even stabilize the original system. The key sufficient condition for the controllers convergence and the satisfactory performance of the feedback system is the uniform exponential stability of the numerical model [ 28 , 29 ] . Unfortunately, not all discretization schemes can conserve uniformly the exponential stability of any distributed-parameter system. For example, standard finite-differences (FD) and finite-elements (FE) methods does not conserve the stability properties of the boundary-damped wave equation, as shown in [ 4 , 9 ] . In the last 30 years, several modifications of the finite-differences and finite-elements methods have been proposed to guarantee a uniform exponential stability conservation. Approaches based on numerical viscosity terms are proposed in [ 9 , 32 , 34 , 47 ] , based on order-reduction methods in [ 24 , 35 , 15 , 25 , 44 ] , and using average operators in [ 33 , 46 ] . Additionally, approaches based on mixed finite-elements with uniform exponential decay can be found in [ 4 , 1 , 11 , 10 , 12 , 8 ] , modal methods in [ 26 ] , and continuous Galerkin in [ 13 ] . On the other hand, the port-Hamiltonian systems is a framework focused on the system energy flux and the description complex interconnected processes, including those characterized by ordinary and partial differential equations, linear or nonlinear, [ 19 , 23 , 42 ] . Between the advantages of this framework is that the well-posedness of boundary-controlled distributed-parameter systems can be done by a simple rank analyzing of a set of matrices that define the boundary variable relations [ 48 , 19 ] . Similarly, these matrices are used to study the system exponential stability, see [ 3 , 43 , 40 ] for details, however, these papers do not provide the decay rate. In [ 27 ] the multiplier method was used to obtain a lower bound on the exponential decay rate in terms of the physical parameters for a class anisotropic port-Hamiltonian systems with boundary dissipation. The discretization schemes of distributed-parameter port-Hamiltonian systems center on the conservation of the port-Hamiltonian structure, i.e., the finite-dimensional approximation model is also port-Hamiltonian and conserves the energy flux structure between the system components. In the literature we can find schemes based on finite-differences [ 39 ] , finite volumes [ 22 ] , partitioned finite-elements [ 37 , 36 , 16 ] , mixed finite-elements [ 8 , 14 , 45 ] , discontinuous Galerkin methods [ 38 ] , and pseudo-spectral methods [ 31 ] . The uniform stability conservation by these discretization schemes is commonly studied evaluating numerically the eigenvalues of the obtained finite-dimensional model [ 31 , 38 , 36 ] . Other challenge for the analysis of the approximation schemes is the choose of the method used to study the uniform conservation of the exponential stability. For example, the eigenvalue method can be used to obtain an analytic expression of the boundary-damped wave equation [ 7 ] , however, for the discretized model, obtaining an analytic formula for the eigenvalues of generic matrices with arbitrary size is a complex task. Alternative approaches based on a frequency analysis were used in [ 26 , 44 ] , among others. The multiplier methods are a powerful tool to obtain an explicit bound for the exponential decay rate of ordinary and partial differential equations [ 41 , 21 ] and their numerical approximations [ 12 , 46 , 33 , 15 ] . In this work, we present a mixed finite-element method (MFEM) the spatial discretization of a class of anisotropic distributed-parameters port-Hamiltonian system with uniform conservation of the exponential stability properties and its application on the linear quadratic control design. To preserves the port-Hamiltonian structure, we express the dynamical system in terms of the co-energy variables [ 42 , 37 ] . Then, the discretization scheme using linear piecewise bases and constant piecewise test functions is applied. For the exponential stability analysis, we use the multiplier method described in [ 41 ] . This method is useful to obtain an explicit expression of the exponential decay for distributed-parameter port-Hamiltonian systems [ 27 ] and numerical approximations [ 8 ] . We use a piezoelectric beam model as example of the uniform conservation of the exponential stability analysis. Finally, using the wave equation, we show how the control law obtained from a linear quadratic controller design based on the MFEM model proposed, converges to a control law expressed in terms of the continuous system variables. 2 Distributed-parameters systems anisotropic port-Hamiltonian systems. This work considers a class of anisotropic distributed-parameters port-Hamiltonian systems on interval [x_{l},x_{r}] . We separate the state variables into two vectors of size n\in\mathbb{N} , \boldsymbol{q}(x,t)=\begin{bmatrix}q_{1}(x,t),\dots,q_{n}(x,t)\end{bmatrix}^{\top}\in H^{1}(x_{l},x_{r};\mathbb{R}^{n}) and \boldsymbol{p}(x,t)=\begin{bmatrix}p_{1}(x,t),\dots,p_{n}(x,t)\end{bmatrix}^{\top}\in H^{1}(x_{l},x_{r};\mathbb{R}^{n}) . Similarly, the space-varying physical parameters are grouped in two sets \{\theta_{i}^{q}(x)\}_{i=1}^{n} and \{\theta_{i}^{p}(x)\}_{i=1}^{n} , strictly positive on C^{1}(x_{l},x_{r}) . Then, the system dynamics are expressed as: \displaystyle\frac{\partial}{\partial t}\begin{bmatrix}\boldsymbol{q}(x,t)\\ \boldsymbol{p}(x,t)\end{bmatrix}-\underset{P_{1}}{\underbrace{\begin{bmatrix}0&A\\ A^{\top}&0\end{bmatrix}}}\frac{\partial}{\partial x}\left(\underset{\Theta(x)}{\underbrace{\begin{bmatrix}\Theta_{q}(x)&0\\ 0&\Theta_{p}(x)\end{bmatrix}}}\begin{bmatrix}\boldsymbol{q}(x,t)\\ \boldsymbol{p}(x,t)\end{bmatrix}\right)=\begin{bmatrix}B_{q}(x)\\ B_{p}(x)\end{bmatrix}\boldsymbol{u}(t),\quad x\in[x_{l},x_{r}], (1) \displaystyle A^{\top}\Theta_{q}(x_{r})\boldsymbol{q}(x_{r},t)=-K\Theta_{p}(x_{r})\boldsymbol{p}(x_{r},t), (2) \displaystyle\Theta_{p}(x_{l})\boldsymbol{p}(x_{l},t)=0, (3) where \mathbf{u}(t)\in\mathbb{R}^{l} denotes control signal, A\in\mathbb{R}^{n\times n} is invertible, \Theta_{q}(x)=diag\left(\theta^{q}_{1}(x),\dots,\theta^{q}_{n}(x)\right) , \Theta_{p}(x)=diag\left(\theta^{p}_{1}(x),\dots,\theta^{p}_{n}(x)\right) , B_{q}(x),B_{p}(x)\in\mathcal{L}\left(\mathbb{R}^{l},L^{2}(x_{l},x_{r};\mathbb{R}^{n})\right) , K=K^{\top}>0\in\mathbb{R}^{n\times n} is the boundary dissipation matrix. Letting \left\langle\cdot,\cdot\right\rangle_{L^{2}} indicate the usual inner product on L^{2}(x_{l},x_{r}), the system energy is \displaystyle\mathcal{H}(t)=\frac{1}{2}\left\langle\begin{bmatrix}\boldsymbol{q}(x,t)\\ \boldsymbol{p}(x,t)\end{bmatrix},\Theta(x)\begin{bmatrix}\boldsymbol{q}(x,t)\\ \boldsymbol{p}(x,t)\end{bmatrix}\right\rangle_{L^{2}}. (4) It is straightforward to verify that \displaystyle\dot{\mathcal{H}}(t)= \displaystyle-\boldsymbol{p}^{\top}(x_{r},t)\Theta_{p}(x_{r})K\Theta_{p}(x_{r})\boldsymbol{p}(x_{r},t)+\left\langle\Theta(x)\begin{bmatrix}\boldsymbol{q}(x,t)\\ \boldsymbol{p}(x,t)\end{bmatrix},\begin{bmatrix}B_{q}(x)\\ B_{p}(x)\end{bmatrix}\boldsymbol{u}(t)\right\rangle_{L^{2}}. (5) and so the system is dissipative; that is, without external stimulus the energy is non-increasing along trajectories. The exponential stability of this system without control, \boldsymbol{u}(t)=0 can be analyzed using the approach in [ 27 ] , if the physical parameters satisfy the following assumption. Assumption 1 . Defining m(x)=x-x_{l}, there is a constant \delta^{c}>0 such that for every i=1\ldots