DEVELOPMENT OF DRAG COEFFICIENT CORRELATIONS FOR CIRCULAR CYLINDER USING SYMBOLIC REGRESSION

Abstract

The current paper provides a symbolic regression-based correlation for the drag coefficient for circular cylinder. The correlation is intended to be applicable over a wide range of flow regimes namely that range from the creeping flow regime up to the turbulent flow regime. Demo version of TuringBot symbolic regression software was used to develop different correlations using different sets of data. Experimental set of data was used in one run whereas steady numerical results for Reynolds number up to ~ 25 were used in generating a second set of formulas. In a different run data generated using Sucker and Brauwer (Wärme-und Stoffübertragung, 1975. 8: p. 149-158) correlation with uniform distribution in each order of magnitude was used to obtain a different set of correlations. The data was generated across five orders of magnitude of change of Re. Among all suggested formulas in the three cases, four correlations are considered for their simplicity and accuracy. The predictions of the correlations ranged from reasonably good to very good as compared to existing data and correlations. The relative error of the four developed correlations ranged between 9% and 16% when compared to experimental data for Reynolds numbers ranging from 1-10⁵. In particular, one correlations was able to capture all the qualitative and quantitative changes in the drag coefficient over the different flow regimes for 0.15 < Re < 10⁵. The relative error of this correlation was comparable to Sucker and Brauwer correlation.