A comparison between radial basis function (RBF) and adaptive neuro-fuzzy inference system (ANFIS) to model the free convection in an open round cavity

This study demonstrates the capability of radial basis function (RBF) and adaptive neuro-fuzzy inference system (ANFIS) to model and predict the free convection heat transfer in an open round cavity. In fact, the effects of the Rayleigh number (Ra) and ratio of the nonconductor barrier distance from the bottom of the cavity to the cavity diameter (H/D), on the free convection in the cavity, are modeled via the RBF and ANFIS models. To start modeling, sufficient data are gathered. Here, data are experimentally generated using a Mach-Zehnder interferometer. In the next step, the RBF and ANFIS models are trained. According to the results, there is an optimum ratio (H/D), in which the heat transfer is maximum. This maximum value increases by increasing the Rayleigh number (Ra). Moreover, based on the results obtained by the RBF and ANFIS, the predicted results for the convection heat transfer are in good agreement with similar ones obtained experimentally. The mean relative errors of the training, testing, and checking data for the RBF model were found as 0.1348%, 1.1972%, and 2.4967%, respectively. Moreover, for the ANFIS model, the error values were 0.0731%, 0.9110%, and 1.9144%, which shows that RBF and ANFIS can predict the results precisely.