Inverse Identification of Circular Cavity in a 2D Object via Boundary Temperature Measurements Using Artificial Neural Network

Abstract

In geometric inverse problems, it is assumed that some parts of domain boundaries are not accessible and geometric shape and dimensions of these parts cannot be measured directly. The aim of inverse geometry problems is to approximate the unknown boundary shape by conducting some experimental measurements on accessible surfaces. In the present paper, the artificial neural network is used to solve these kinds of problems in conduction heat transfer in 2D objects. In order to train the neural network, some direct problems are solved by using the finite element method. In order to evaluate the applicability of the proposed method, different cases with different number of measuring points and different error levels are examined. The results show that the ANN can effectively be used in solving inverse geometry problems in heat conduction.