Differential Riccati equation for the active control of a problem in structural acoustics (Q2564170)

In fluid acoustics, one encounters a fluid-structure interaction, while in many cases the structure (or container) is considered rigid, this is not the case in several problems of aerospace science, submerged crafts etc.\dots, when structural elasticity must be taken into account. Here, the authors consider a classical wave equation coupled with a beam-like equation (in the words of the authors), which could be a plate or shell equation in which one dimension is heavily preferred, as was the original approximate model of Sophie Germaine-Lagrange plate theory. The system of equations is: \[ z_{tt}= \Delta z\text{ on }\Omega\times(s,T),\quad v_{tt}=-\Delta^2v- \Delta^2v_t- z_t+ Bu, \] with appropriate initial and boundary conditions. \(\partial z/\partial\nu=v_t\) on \(\Gamma_0\) and \(\partial z/\partial\nu=0\) on the complement of \(\Gamma_0\). \(B\in{\mathcal L}(U,H^{-\alpha}(\Gamma_0))\), where \(\alpha\) is specified for specific shape of the domain. Since the operators considered by the authors are positive definite, their fractional (and also negative) powers are well defined. The combined hyperbolic-parabolic system is rewritten as an abstract dynamical system: \[ dY/dt= Ay+ Bu\text{ in }D(A^*)',\quad Y(s)= Y_0\in H_1\times H_0. \] The authors set up the functional analytic background for minimization of a quadratic functional which may be regarded a square of a Hilbert space norm. The control is an unbounded trace-type operator. Unbounded property of admissible controls generally precludes continuity of the optimal control. In fact it is known that for the purely hyperbolic systems unbounded controls imply unbounded gain. In this case, the gain turns out to be bounded because of the smoothing property of the analytic part of the dynamic system. The rather difficult study of the resulting Riccati equation combined with local microanalysis of the hyperbolic part produces the following main result. For prescribed data \(Y_0\in H_1\times H_0\) there exists a unique optimal control. Certain regularity conditions are true for the optimal control and for the optimal solution \(Y^*(s)\), which continuously depends on \(Y_0\).