Numerical solution for an inverse problem about determination of volumetric heat capacity of rock mass during artificial freezing

The paper is devoted to development and implementation of algorithms for a numerical solution of a coefficient inverse Stefan problem. This problem arises in a modeling for a process of ice wall formation around a projected horizontal section of a mine shaft. The ice wall is formed by an artificial ground freezing to convert soil pore water into ice. The temperature field modeling is based on an enthalpy form of a two-dimensional Stefan problem. The aim of the study is to determine the volumetric heat capacity for the rock layer on the base of additional information about temperature evolution in thermal wells. The problem of coefficients' identification is stated as a variation form of the coefficient inverse Stefan problem. As a result, two algorithms for a numerical solution of the stated inverse Stefan problem have been developed. The first algorithm is based on the conjugate gradient iterative optimization method. The second algorithm is based on the steepest descent iterative optimization method. Under the first algorithm the calculation of the discrepancy functional gradient and determination of parameters for the optimization method are performed by solving a sensitivity problem and an adjoint problem. Forms of these problems have been obtained for the stated direct Stefan problem. For the second algorithm the discrepancy functional gradient and parameters for the descent step method are determined by calculating sensitivity coefficients. Solutions of the direct problem, the sensitivity problem and the adjoint problem are performed by the finite element method. The special feature of the used optimization methods is that these methods have regularizing properties. In order to verify effectiveness of the proposed algorithms, the computational experiments have been performed. The first and second experiments are related to determining only one unknown volumetric heat capacity for an ice domain or a cooling domain. The third experiment is devoted to determining the volumetric heat capacity for the both domains. The results of the experiments show that both algorithms allow to determine the volumetric heat capacity with a sufficiently good accuracy. However, a convergence rate of the second algorithm is higher than the rate of the first algorithm. The presented approach to modeling the process of ice wall formation as the coefficient inverse Stefan problem and the developed algorithms can be used for designing and improving the initial data for building mine shafts with a technology of artificial ground freezing.