Source identification for the wave equation on graphs (Q470128): Difference between revisions

The authors consider source identification problems for the wave equation on an interval and on trees. The main advantage of their approach is its locality: to recover unknown functions on a part of the graph, a piece of information is used which is only relevant to this subgraph. This feature of the method allows the authors to propose a very efficient identification algorithm, which is new, even in the one-dimensional case, and much simpler than similar algorithms proposed in the past. In the case of an interval, consider the initial boundary value problem \[ \begin{aligned} u_{tt}- u_{xx} & = f(t,g(x),\quad 0< x< l,\quad 0<t<T,\tag{1}\\ u(0,t) &= u(l,t)= 0,\quad 0< t< T,\tag{2}\\ u(x,0) &= u_t(x,0)= 0,\quad 0<x<1,\tag{3}\end{aligned} \] Here, \(f\) is a known function, while \(g\) is an unknown function. The following cases are considered: (a) \(g\in L^2(0,T)\), (b) \(g(x)= \sum^N_{j=1} \alpha_j\delta(x- \xi_j)\), \(0< \xi_1<\cdots< \xi_N<l\). Then, the authors prove the following theorem. (a) If \(f\in L^2(0,T)\) and \(g\in L^2(0,l)\) in the IBVP (1)--(3), then \(u_x(0,\cdot)\in C(0,T)\) and it can be represented by \[ u(0,t)= \int^t_0 g(x)w(x,t)\,dx. \] (b) If \(f\in C[0,T]\) and \(g\) is a linear combination of \(\delta\)-functions, then \(u_x(0,\cdot)\in C[0,T]\) and it can be represented by \[ u_x(0,T)= \sum_{j;\xi_j\leq t}\alpha_j w(\xi_, t),\quad 0\leq l, \] where \(w\) is a solution of the problem \[ \begin{aligned} w_{tt}-w_{xx} & =0,\quad 0<x<l,\, 0<t<T,\\ u(0, t) &= f(t),\;0<t< T,\\ w(x, 0) A &= w_t(x,0)= 0,\;0<x< l.\end{aligned} \] It should be remarked that the above approach can be extended to equations on graphs.