Approximation results for parameter estimation in a class of abstract nonlinear hyperbolic systems (Q1808445)

The object of this paper is the following class of abstract nonlinear hyperbolic systems envolving in a compact separable Hilbert space \(H\): \[ w_{tt}+A_1(q) w+ A_2(q) w_t+ N^* g(q)N(w)= f(t,q);\quad w(0)= \varphi_0;\quad w_t(0)= \varphi_1.\tag{1} \] Here, \(A_1(q)\), \(A_2(q)\) are unbounded operators depending on some parameter \(q\), \(g(q)\) is a nonlinear operator in \(H\), \(N\) is an unbounded operator, and \(f\) is a parameter dependent forcing term. A general parameter estimation problem is formulated, and its convergence properties are analyzed. To solve the least-square minimization problem \[ I(q,z)= \|\widetilde C_2\{\widetilde C_1\{w(t_i, q)\}- \{z_i\}\}\|^2,\tag{2} \] a Galerkin type approximation to equation (1) is considered, and a family of approximating parameter estimation problems is defined. In (2), \(\{z_i\}^K_{i=1}\) is a set of observations taken at time \(t_i\), \(w(t_i,q)\) is the parameter dependent solution of equation (1) evaluated at \(t_i\) \((i= 1,2,\dots,K)\), \(\|\cdot\|\) is an appropriate chosen Euclidean norm, and the observation operators \(\widetilde C_1\), \(\widetilde C_2\) depend on the type of collected data.