Numerical Simulation of Inverse Retrospective Problems for a Two-Dimensional Heat Equation

Abstract

The paper proposes a new, unique method for numerically solving inverse problems for nonlinear conditions using the example of the problem of restoring the initial condition for a two-dimensional heat equation with boundary conditions of the third kind. In this problem, the initial condition is an unknown function of two variables, and is determined from experimental temperature values. The proposed method is based on using the parametric identification method, the implicit gradient descent method and Tikhonov’s regularization method. An algorithm and a software package for numerical solution have been developed. The use of the implicit gradient descent method allowed for faster convergence (number of iterations) compared to zero-order methods. Numerous results of numerical experiments have been obtained and discussed. An analysis of the behavior of solution functions with and without the use of Tikhonov’s regularizing functional along with the impact of the regularizing parameter has been carried out. The results of computational experiments using the proposed numerical method showed that the error in the results obtained does not exceed the error in the experimental data due to the correct choice of the regularizing parameter.