Heat transfer and pressure drop of Al2O3/water nanofluid in conically coiled tubes: Experimental and artificial neural network prediction

Abstract

A novel model of an artificial neural network (ANN) is proposed in order to predict the thermal performance of Al2O3/water nanofluid in conically coiled tubes (CCTs). Experiments were conducted at volume concentrations of 0.3 %, 0.6 %, and 0.9 % of Al2O3/water nanofluid and coil torsions ranging from 0.02 to 0.052. Feed-forward neural network (FFNN) have been modelled and trained in order to predict the experimental and non-experimental Nusselt number (Nu) and.friction factor (ƒ). Using the TRAINLM algorithm, the FFNN for predicting Nu and ƒ is well-trained, with correlation coefficients of 0.9952 and 0.9482, respectively. FFNN exhibited greater accuracy in predicting the Nu and friction factor, since the Root mean square error (RMSE) between experimental and predicted data was minimal. The average RMSE, and Mean absolute percentage error (MAPE) were 9.6166 and 3.101 for the predicted Nusselt number. The predicted results of the ANN for the average Nusselt number and friction factor at φ = 0.8 % align well with the experimental data, even though they have not yet been empirically validated. These findings demonstrate the capability of ANNs to accurately predict the Nusselt number and friction factor and yield satisfactory outcomes.

Keywords

• Artificial neural network (ANN)
• Feed forward neural network (FFNN)
• Nanofluid
• Conical coiled tubes
• Heat transfer