Approximate inverse for a one-dimensional inverse heat conduction problem (Q4950645)

This paper is devoted to the application of the method of the approximate inverse to a one-dimensional inverse heat conduction problem. The method was published by \textit{A. K. Louis} [see ibid. 12, No. 2, 175-190 (1996; Zbl 0851.65036) and ibid. 15, No. 2, 489-498 (1999; Zbl 0933.65060)]. It is a mollification method, where for a linear ill-posed problem \(Af= g\), \(A: X\to Y\) the topology of \(X\) is changed such that the approximate problem becomes well-posed. For a suitable function \(m_\gamma\) a reconstruction kernel \(\psi_\gamma\) with \(A^*\psi_\gamma= m_\gamma\) will be used to formulate an approximation \(S_\gamma g:=\langle g,\psi_\gamma\rangle\) of solution moments \(\langle f,m\rangle\). In the inverse problem under consideration \(f\) is the unknown heat flux at \(x=1\) for the heat equation \(u_{xx}= u_t\) on the unit interval and \(g\) is the observed boundary temperature at \(x=0\). There are given results about the regularization effect of the approximate inverse and error estimates when smoothness assumptions are satisfies. The mollification parameter \(\gamma\) plays the role of the regularization parameter. The authors also give an algorithm in pseudo-code for that very fast method. Some numerical examples complete the paper and indicate that online reconstruction seems to be possible for the inverse heat equation problem.