Parametric study and application of a data-mining model in 2D and 3D micro-fin tubes

Abstract

• The application of least squares-support vector regression method in micro-fin study. • The best 2D and 3D micro-fins had similar design parameters. • The best 2D micro-fin tube enhanced the efficiency index up to %18. • Importance of micro-fin’s axial pitch over height ratio on the efficiency index. Micro-fins (<0.5 mm tall) are an engineered roughness with the advantage of reducing thermal resistance and the disadvantage of increased pressure drop when applied inside a tube in heat-exchange applications. The competing effects highlight the need for careful optimization that identifies micro-fin surfaces with the potential to match heat exchanger design needs. Hence, the objectives of this study are chosen to enable efficient optimization in future studies. The main goals are: (1) study the effects of micro-fin design variables on heat transfer and friction factors; and (2) evaluate the potential of a data-mining model as a surrogate of computational fluid dynamic ( CFD ) models in 2 dimensional ( D ) and 3 D micro-fin tubes. This study applied conductive and convective heat transfer and turbulent fluid flow simulation to evaluate the performance of different 2 D and 3 D micro-fin tubes. Different configurations were generated by varying micro-fin height ( e ), helix angle ( α ), number of starts ( N f ), and discontinuity features. Coupled solid and periodic fluid domains were applied in ANSYS Fluent 19.1 . Performance was mapped for 210 different simulations (including a smooth tube) using a realizable k - ε turbulence model at Reynolds number ( Re ) of 48,928. Two different least squares-support vector regression ( LS - SVR ) models were employed to estimate the Colburn j factor as a function of geometric variables and the Fanning friction factor ( f ) as a function of geometric variable and j factor. Results of the parametric study showed that the best 2 D micro-fin tube can enhance efficiency index ( η ) up to 1.18. Results of the LS - SVR model showed that the percentage of average absolute error ( AAE ) between simulated and estimated j and f factors are 2.05% and 2.93% for 3 D micro-fin tubes, respectively.