A comprehensive review, CFD and ML analysis of flow around tandem circular cylinders at sub-critical Reynolds numbers

Abstract

The hybrid review paper meticulously examines crucial research on tandem cylinders across a broad range of Reynolds ($Re$) numbers, extending up to $170,000$ for Strouhal ($St$) and $300,000$ for pressure coefficients ($C_P$). By consolidating findings on various flow parameters, including Strouhal number, drag ($C_D$), lift ($C_L$), and pressure coefficients ($C_P$), the paper advocates the use of experimental and three-dimensional numerical data, exclusively omitting two-dimensional numerical data, especially at higher $Re$ numbers. To this end, the predictive performance of different machine learning techniques-such as XGBoost, genetic optimization, ensemble modeling, and Random Forest-was evaluated using numerical simulations and data sourced from literature. The results demonstrate that, given a sufficiently large dataset, these techniques can accurately predict flow variables like Strouhal number and pressure coefficients with minimal computational cost. However, it is crucial to use only three-dimensional datasets for such analyses. The study identifies Random Forest and XGBoost models as the most accurate in forecasting flow-induced oscillations and pressure distributions around the cylinders, exhibiting the lowest mean squared errors for Strouhal number and pressure coefficient predictions.