Capability to predict the steady and unsteady reduced aerodynamic forces on a square cylinder by ANN and GEP

摘要

The reduction in the drag and lift forces on a square cylinder is obtained by attaching an extended solid (thorn) with it positioned at the front stagnation point and at the rear stagnation point separately at low Reynolds number, Re = 40, 100 and 180. The effect of variation of thorn length (l' = 0.2, 0.4 and 0.6) and inclination angle (theta = 5 degrees, 10 degrees, 15 degrees and 20 degrees) on the drag, lift and this effect is forecasted by the ANN (artificial neural network) and GEP (gene expression programming). The required input and output data to train the ANN and GEP models have been taken from the available published results of the present authors (Dey and Das in Eng Sci Technol 18: 758-768 [2015]; in Ain Shams Eng J 6: 929-938 [2015]). It is found that the drag and lift are minimized by 16 and 46 % for Re = 100, respectively, and 22 and 60 % for Re = 180 compared to a square model. It has been perceived that both the prediction tools (ANN and GEP) can forecast the aerodynamic behavior correctly and quickly within the given range of the training data than simulation, which takes much longer time to complete the computation. It is further observed that the GEP model is much efficient than the ANN model.