Neural network and multi-objective optimization of confined flow characteristics on circular cylinder in standing double vortex region

• Published in: Neural computing & applications 2021-02, Vol.33 (4), p.1379-1398
• Main Authors: Senthilkumar, Rajendran, Vasudevan, Premalatha, Prabhu, Sethuramalingam
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Artificial neural networks; CAD; Circular cylinders; Computational Biology/Bioinformatics; Computational fluid dynamics; Computational Science and Engineering; Computer aided design; Computer Science; Confined flow; Data Mining and Knowledge Discovery; Diameters; Drag coefficients; Flow characteristics; Flow separation; Fluid flow; Image Processing and Computer Vision; Multiple objective analysis; Neural networks; Optimization; Original Article; Probability and Statistics in Computer Science; Reynolds number; Unsteady state
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-020-05079-z

The unsteady state and isothermal two dimensional numerical computations were carried out using Ansys Fluent-18 between the Reynolds number ranges 10 to 50. The blockage ratios (Domain height to the circular cylinder diameter) range 1.54–112. The flow characteristics such as drag coefficients and length of recirculation are optimized and correlated as a function of various Reynolds numbers at different blockage ratios. Gradual decrease in blockage ratio which means the increase in blockage effect postponed the flow separation, transition and reduces the length of recirculation and also makes the flow steady. In this study optimum flow characteristics exist at maximum blockage ratio, i.e. with minimum blockage effect and maximum Reynolds number. The artificial neural networks model proved to predict values of the total drag coefficient ( R 2  = 0.979) and length of recirculation ( R 2  = 0.992) closer to simulated data at 95% ( α  = 0.05) confident interval.