Experimental and numerical investigations on the heat transfer characteristics of a real-sized radiant cooled wall system supported by machine learning

Abstract

Despite the extensive utilization of radiant air conditioning units in rooms, challenging points of design associated with the calculation of cooling load are still present. Except for the radiant wall cooling studies aimed at conducting heat transfer-focused analyses carried out by the authors of this investigation, neither experimental nor computational studies exist in the related literature. The current experimental and computational study aims to address the deficiencies in the radiant-cooled wall problem. Differing from other conditioned rooms, the heat is exposed through the back surface of the analyzed wall, whose heat flux range lies between 1.60 and 10.84 W/m2. The total, radiative, and convective heat transfer coefficients of 7.78, 5.13, and 2.52 W/m2.K are acquired as values for use in building energy simulation programs. Seven different artificial neural network models are designed to estimate the total, radiative, and convective heat transfer coefficients and heat transfer rates. Dependency analyses are also performed using various inputs in the investigated numerical models. The margin of deviation values computed for six different output factors are found below −1.80%, the mean square error values are less than 1.51E-04, the R values are greater than 0.98, and the data points do not surpass the 10% deviation line. Artificial neural networks have been found to outperform well-known correlations in estimating experimental results. Extensive measured experimental data are presented for the sake of other researchers numerical modelling and validation issues. Building energy simulation software designers and engineers in the field of thermal comfort are thought to benefit from these findings.