Prediction of heat transfer coefficient, pressure drop, and overall cost of double-pipe heat exchangers using the artificial neural network

Typically, success in the estimation of machine learning is expected to rise with increasing input parameters, whereas the noise issue may rarely arise owing to redundant input factors undesirably influencing the learning algorithm. The parameters such as overall heat transfer coefficient, pressure drop, and overall cost have been determined by two different artificial neural networks evaluated by a multi-layer perceptron model. Using the Levenberg-Marquardt training algorithm, in the first model input layer, a total of 10 input parameters ρ, np, k1, Re1, fi, Re2, fo, ns, P1 and P2 have been defined, while the second involves 8 input parameters by subtracting pumping powers from the first one, thus the noise issue has been investigated using unnecessary input parameters. Overall heat transfer coefficient, tube/annulus sides pressure drops, and overall cost have been estimated with deviations of 0.16%, 0.23%, 0.02%, and 0.003% via Model 1, 0.02%, 0.18%, 0.16%, and 0.15% via Model 2, respectively. Moreover, Model 1 results in the best mean squared errors for annulus side pressure drop and overall cost with the values of 2.54E-04 and 1.93E-04, correspondingly, whereas Model 2 yields the best values of 1.11E-04 and 1.90E-04 for overall heat transfer coefficient and tube side pressure drop, sequentially.

Abstract licence: CC BY-NC

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.