Determining the relaxation kernel in nonlinear one-dimensional viscoelasticity (Q1301247): Difference between revisions

The author is concerned with the problem of recovering the unknown memory kernel \(k:[0,T]\to {\mathbb R}\) in a viscoelastic string governed by the nonlinear Volterra integrodifferential equation \[ \begin{multlined} \rho(x)D_t^2u(x,t) - D_x\Biggl\{ \varphi_0(x)D_xu(x,t) + \int_0^t k(s)\psi(x,D_xu(x,t-s)) ds\Biggr\}=f(x,t),\\ (x,t)\in Q_T=(0,L)\times (0,T), \end{multlined}\tag{1} \] subject to the initial conditions \[ u(x,0)=u_0(x),\qquad D_tu(x,0)=u_1(x),\qquad x\in [0,L], \tag{2} \] to the boundary conditions \[ u(0,t)=\alpha(t),\qquad D_tu(0,t)=\beta(t),\qquad t\in [0,T], \tag{3} \] and to the (additional) flux condition \[ \varphi_0(0)D_xu(0,t) + \int_0^t k(s)\psi(0,D_xu(0,t-s)) ds= g(t),\qquad t\in [0,T]. \tag{4} \] Here \(\rho\), \(\varphi_0\), \(\psi\), \(f\), \(u_0\), \(u_1\), \(\alpha\), \(\beta\), \(g\) are (smooth) given functions satisfying compatibility conditions up to the third order. Moreover, functions \(\rho\), \(\varphi_0\) and \(D_xD_z\psi\), \(D_z^2\psi\) are assumed to be, respectively, strictly positive, and uniformly locally Lipschitz continuous in their second argument \(z\). Under the basic inequality \(\psi(0,u_0'(0))\neq 0\) the author proves a local in time existence and uniqueness result for the identification problem (1)--(4) in the context of classical spaces \(C^2({\overline Q_T})\) for the displacement \(u\) and the velocity \(D_tu\) and of the Sobolev space \(W^{1,1}((0,T))\) for the memory kernel \(k\). Furthermore, a global in time continuous dependence result with respect to data \(f\), \(u_0\), \(u_1\), \(\alpha\), \(\beta\), \(g\) is proved under the assumption that \(k\) may be a-priori estimated in \(W^{1,1}((0,T))\). The present paper is related to a previous one by the reviewer [Nonlinear Anal., Theory Methods Appl. 22, No. 1, 21-44 (1994; Zbl 0818.93014)], where the flux condition (4) was replaced by the measurement of the displacement \(u\) at some point \(x_0\) \((u(x_0,t)=g(t))\). However, the techniques used by M. Grasselli, which replace energy estimates by \(C^2({\overline Q_T})\)-ones [cf. \textit{R. Grimmer} and \textit{E. Sinestrari}, Differ. Integral Equ. 5, No. 2, 421-432 (1992; Zbl 0782.47037)], in the flux case considerably improve and simplify those employed by the reviewer (loc. cit.).