A comprehensive approach to analyze the discrepancies in heat transfer characteristics pertaining to radiant ceiling heating system

Abstract

Radiant heating/cooling systems are being popular thanks to their ability of regulating the living-environment with the use of low temperature heating and high temperature cooling. In this work, an artificial neural network investigation is carried out to predict heat transfer characteristics over a heated radiant ceiling. Experimental tests consisting of 28 case studies, obtained through varying supply water temperature, are conducted. A computational method, including the Boussinesq approach using k-epsilon RNG model, is also employed to increase the number of case studies in order to use them in artificial neural networks investigation that applies Levenberg-Marquardt training function. Thus, total data number have been increased from 28 to 74 by a simulation software. Estimations of artificial neural networks method are compared with experimental data, and seen that the outputs are compatible with each other, where most of deviations are within the range of +/- 15%. According to this result, experimental data can be increased by a numerical simulation software and evaluated by one of the artificial intelligence techniques, successfully. In conclusion, the heat transfer coefficients to use in the radiant ceiling heating applications are proposed as 0.9 W/m(2)K, 5.3 W/m(2)K, and 7.0 W/m(2)K for convective, radiative, and total heat transfer coefficients, respectively.