CFD-based optimisation of base pressure behaviour on suddenly expanded flows at supersonic Mach numbers. (27th May 2022)

• Record Type: Journal Article
• Authors: Quadros, Jaimon Dennis; Khan, S.A.; Prashanth, T.
• Is Part Of: Progress in computational fluid dynamics. Volume 22:Number 3(2022)
• Page Start: 159
• Page End: 173
• Publication Date: 2022-05-27
• Subjects: base pressure -- CFD -- response surface methodology -- RSM -- neural networks -- optimisation; Fluid dynamics -- Data processing -- Periodicals; Fluid dynamics -- Mathematics -- Periodicals; 620.10640285
• Languages: English
• ISSNs: 1468-4349

Abstract

The base pressure developed in a suddenly expanded flow process majorly depends on Mach number (M), nozzle pressure ratio (NPR), area ratio (AR), and length to diameter ratio (L/D). Numerical analysis of the flow process was carried out using the computational fluid dynamics (CFD) technique, and was validated by experiments. The input-output test cases for CFD analysis were developed by two statistical methods, namely central composite design (CCD) and Box-Behnken design (BBD). The BBD model yielded better prediction accuracy and was used for generating data that trained the recurrent and backpropagation neural networks. The recurrent neural network outperformed both the backpropagation neural network and Box-Behnken design. Furthermore, to assess the right range of conditions for maximising base pressure, the genetic algorithm (GA), desirability function approach (DFA), and particle swarm optimisation (PSO) techniques were implemented. The PSO and GA techniques were found to be better, as they carried out search operations in many directions at multi-dimensional space simultaneously.