Hydrodynamic Flow Characteristics Prediction for Bluff Body Wake via Novel Adaptive Neuro-Fuzzy Controller Avoiding Fuzzy Rule Explosion

Abstract

<div class="section abstract"><div class="htmlview paragraph">This study analyses the effect of Reynolds number (<i>Re</i>) and bluff body shape (quantified by shape factor <i>SF</i>) variation on various hydrodynamic characteristics of unsteady bluff body flow, such as Strouhal number, maximum lift coefficient, and mean drag coefficient. The study initially examines a relationship among these characteristics and further utilizes artificial neural networks (ANNs) and adaptive neuro-fuzzy inference system (ANFIS) controllers for their precise prediction. The results from real-time computational fluid dynamics (CFD) experimentations were gathered and considered to train ANN controllers. A novel ANFIS controller has been designed using only three membership functions thus solving the problem of fuzzy rule explosion. The results indicate that both the ANN and ANFIS controllers can precisely predict these hydrodynamic flow characteristics as validated through minimal values of root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). It is observed that ANFIS controller provides better results compared to the proposed feed-forward ANN controller. The RMSE, MAE, and MAPE obtained for ANFIS model for different shape factors for maximum lift coefficient were 0.0024, 0.002, and 0.85%, respectively.</div></div>