scientific article; zbMATH DE number 1172057

From MaRDI portal
Publication:4397315





The authors deal with the problem of identifying an unknown memory kernel \(m:[0,T]\to {\mathbb{R}}\) in the integrodifferential equation \[ \rho D_t^2u(x,t) - \lambda \text{ div} (\beta \nabla u(x,t)) - (m*\text{ div} (\beta \nabla))(x,t) = \gamma(x,t)\qquad (x,t)\in Q=D\times (0,T) \tag{1} \] where \(\lambda \in \{0,1\}\), \(f*g(t)=\int_0^t f(t-s)g(s) ds\), \(D\) is a bounded domain in \({\mathbb{R}}^n\) with piecewise smooth boundary \(S\), while \(\rho\), \(\beta\) and \(\gamma\) are continuous functions in \({\overline D}\) and \({\overline Q}\), respectively. We recall that equation (1) governs the wave propagation in viscoelastic media of Boltzmann type, \(\rho\) denoting the density of the body. The displacement \(u\) is assumed to satisfy the initial conditions \[ u(x,0)=\varphi(x),\quad D_tu(x,0)=\psi (x)\quad \text{ in\;} D \tag{2} \] as well homogeneous boundary value conditions of the form \[ u(x,t)=0\quad \text{ or}\quad \beta(x)D_\nu u(x,t)+\alpha(x)u(x,t)=0,\quad (x,t)\in \partial D\times (0,T) \tag{3} \] where \(\nu\) denotes the outer normal to \(S\) and \(\varphi\), \(\psi\), \(\alpha \geq 0\) are continuous functions. \noindent The additional condition, which is needed to determine the kernel \(m\), and usually corresponds to physical measurements, is given by \[ \Psi[u(\cdot,t)]=h(t)\quad \text{ or}\quad \lambda \Psi[u(\cdot,t)] - m*\Psi[u(\cdot,t)]=h(t),\qquad t\in [0,T] \tag{4,} \] where \(\Psi\) and \(h\) are a linear continuous functional and a \(C^1([0,T])\)-function, respectively. The authors make the basic assumption that the linear differential operator \((1/\rho)\text{ div} (\beta \nabla)\) endowed with boundary conditions (3) admits a system of eigenfunctions which is complete in \(L^2(D)\). Using Fourier expansions, the problem is first transformed into a nonlinear operator Volterra integral equation of the second kind for \(m\) \[ m = K(m)*m + l,\tag{5} \] where the datum \(l\) is assumed to belong to \(L^p((0,T))\) \((p\in (1,+\infty))\) or to \(C_p([0,T])=\{ m\in C((0,T]): t^{1-p}m\in C((0,T])\}\) \((p\in (0,1])\). Then, using an abstract existence and stability theorem for operator equations in weighted Banach spaces [cf. \textit{J. Janno} and \textit{L. v. Wolfersdorf}, Z. Angew. Math. Mech. 77, 243-257 (1997; Zbl 0883.35054)] the authors can prove the existence and uniqueness of the solution to problem (5) (and consequently to problem (1)--(4) in both cases \(\lambda=0\) and \(\lambda=1\)) as well as its continuous dependence on the data.




Cited in
(22)








This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4397315)