Development of GEP and ANN model to predict the unsteady forced convection over a cylinder. Issue 8 (November 2016)

• Record Type: Journal Article
• Authors: Dey, Prasenjit; Sarkar, Abhijit; Das, Ajoy
• Is Part Of: Neural computing & applications. Volume 27:Issue 8(2016)
• Page Start: 2537
• Page End: 2549
• Publication Date: 2016-11
• Subjects: Square cylinder -- Sharp and rounded corner -- Forced convection -- GEP -- ANN; Neural networks (Computer science) -- Periodicals; Neural circuitry -- Periodicals; Artificial intelligence -- Periodicals; Neural Networks (Computer) -- Periodicals; Réseaux neuronaux (Informatique) -- Périodiques; Réseaux nerveux -- Périodiques; Intelligence artificielle -- Périodiques; 006.32
• DOI: 10.1007/s00521-015-2023-8
• Languages: English
• ISSNs: 0941-0643

Abstract

Abstract In the present study, forced convection heat transfer flow past a square cylinder with a rounded corner edge in an unsteady two-dimensional low Reynolds number laminar regime is numerically performed and predicted by the artificial neural network (ANN) and gene expression programming (GEP). In this study, Reynolds number (Re ) is varied from 80 to 180 with Prandtl number (Pr ) variation from 0.01 to 1000 for various corner radii (r = 0.50, 0.51, 0.54, 0.59, 0.64 and 0.71). Finite volume-based commercial software FLUENT is used in the present numerical solution for the output results that are used to train the present ANN and GEP model. The local Nusselt number (Nu local ) and average Nusselt number (Nu avg ) at various Reynolds numbers, Prandtl numbers and various corner radii are utilized to forecast the heat transfer characteristics. It is found that the heat transfer rate of a circular cylinder can be enhanced by 12 % whenRe is varying and 14 % when the Prandtl number is varying by introducing new cylinder geometry of corner radiusr = 0.51. The forced convection characteristics can be envisioned properly and very promptly by using back-propagation ANN and GEP in contrast to a regular CFD approach. It is also found that GEP is more efficient in predicting than the ANN. The outcomes obtained by the present numerical study confirm a fine accord with the available results in the literature.