Optimization of number and locations of discrete heaters in a two-dimensional radiant heating furnace using artificial neural networks

Abstract

In this paper, the inverse solution of determining the optimum location and number of discrete heaters in a two-dimensional radiant furnace with participating media is presented as an optimization problem and solved using an exhaustive search. The regular radiative transfer equation is replaced by a neural network-based surrogate to make the solution process faster. This surrogate model is coupled with an exhaustive search algorithm to determine the best distribution of heaters in a radiant furnace containing H2O and CO2 as participating gases. The RTE solution method, when replaced with ANN results in 1000 times saving in the computational time. The accuracy of the heat flux estimations for the predictions with RTE and ANN were found to be under 5%. The optimum configuration of the heaters is found to be sensitive to the medium properties and design surface temperature. This new method of using ANN in place of the RTE and coupling it with search algorithms proves to be an accurate, fast and robust tool in solving the inverse problem of optimization of heater configurations.