A Machine-Learning-Based Surrogate Model for Internal Flow Nusselt Number and Friction Factor in Various Channel Cross Sections

Pai, Saeel S.; Visaria, Dhvaneel; Weibel, Justin A.

doi:10.1109/itherm51669.2021.9503289

Cited by 1 publication

(1 citation statement)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Pai et al conducted a study in which they used artificial neural networks to develop surrogate models for predicting Nusselt number and friction factor in ducts of different constant cross-sectional shapes. 31 They suggested to utilize machine learning techniques such as artificial neural network with three hidden layers and Huber loss function for this purpose. The results indicate that the predictions obtained through this approach were highly accurate when compared to existing data.…”

Section: Introductionmentioning

confidence: 99%

LBM curved boundary treatments for pulsatile flow on convective heat transfer and friction factor in corrugated channels

Aslan,

Ozsaban,

Kucur

et al. 2023

Proceedings of the Institution of Mechanical Engineers, Part C:

View full text Add to dashboard Cite

The present research investigates heat transfer and the flow characteristics of periodically corrugated wavy channels numerically under pulsatile flow conditions. The numerical method used here is Lattice Boltzmann Method (LBM), and the validation of the study is done by Ansys-Fluent which is finite volume based commercial Computational Fluid Dynamics (CFD) code. For modeling walls, bounce-back method, namely, staircase method and three different curved boundary treatments, which are extrapolation, Filippova-Hänel (FH) and Mei-Luo-Shy (MLS), are used. For modeling constant temperature at walls, staircase method and the same curved wall treatments are used. Corrugated channels have a sharp wavy peak, and its inclination angle is 30°. Two different minimum channel heights are considered, which are 5 and 10 mm in corrugated channels. Flow regime is assumed as laminar (50 < Re < 300) and Prandtl number is kept as 0.7. Four kinds of different sinusoidal pulsatile flows are used with a combination of two different dimensionless frequencies and dimensionless amplitudes. For varying Reynolds number range, Nusselt number and friction factor are calculated. Narrow channel creates higher Nusselt number and friction factor than wide channel. At low Reynolds number, Nusselt number does not changed with pulsatile flow conditions, however at high Reynolds number cases of lower dimensionless frequency and higher dimensionless amplitude produce higher Nusselt numbers. Lower dimensionless frequency cases produce higher Nusselt number than higher dimensionless frequency cases. Pulsatile flow conditions have no effect on friction factor and for narrow and wide channel. Nusselt number prediction of FVM is close to STR and EXT for all cases of narrow channels, and close to the FH, MLS and EXT for all cases wide channel. Correlation equations for Nusselt number and friction factor are constructed by deep neural network (DNN) algorithm.

show abstract

Section: Introductionmentioning

confidence: 99%