Multi-Objective Optimization of Low Reynolds Number Airfoil Using Convolutional Neural Network and Non-Dominated Sorting Genetic Algorithm

Abstract

The airfoil is the prime component of flying vehicles. For low-speed flights, low Reynolds number airfoils are used. The characteristic of low Reynolds number airfoils is a laminar separation bubble and an associated drag rise. This paper presents a framework for the design of a low Reynolds number airfoil. The contributions of the proposed research are twofold. First, a convolutional neural network (CNN) is designed for the aerodynamic coefficient prediction of low Reynolds number airfoils. Data generation is discussed in detail and XFOIL is selected to obtain aerodynamic coefficients. The performance of the CNN is evaluated using different learning rate schedulers and adaptive learning rate optimizers. The trained model can predict the aerodynamic coefficients with high accuracy. Second, the trained model is used with a non-dominated sorting genetic algorithm (NSGA-II) for multi-objective optimization of the low Reynolds number airfoil at a specific angle of attack. A similar optimization is performed using NSGA-II directly calling XFOIL, to obtain the aerodynamic coefficients. The Pareto fronts of both optimizations are compared, and it is concluded that the proposed CNN can replicate the actual Pareto in considerably less time.

On the investigation of the aerodynamics performance and associated flow physics of the optimized tubercle airfoil

“…6. Pareto-optimal front 54 showing the possible solutions for the present multiobjective optimization problem (Note that the x axis and y axis are not set to equal scale). The maximum drift in the force reading along three axes is found to be 60.01N, which corresponds to a non-dimensional force coefficient of 60.0125.…”

A Review of Intelligent Airfoil Aerodynamic Optimization Methods Based on Data-Driven Advanced Models

“…Artificial neural networks essentially analyze the algorithmic laws between input and output parameters through the connected neurons, and cannot locally sense and learn from the nodes in the flow field. With the development of machine learning techniques and aerodynamic evaluation, convolutional neural networks (CNN) with feature learning capability are applied to finely identify flow field images and perform the hierarchical learning of aerodynamic characteristics [88][89][90][107][108][109]. Zhang et al [88] developed an appropriate CNN structure for varying flow conditions and geometric configurations, which could predict airfoil's lift coefficients under different Mach numbers, Reynolds numbers, and various attack angles.…”

“…Yu et al [89] proposed an enhanced deep CNN method by introducing the "feature enhance-image" strategy for airfoil images. Bakar et al [90] presented a framework for designing low Reynolds number airfoils based on a CNN-based aerodynamic coefficient prediction model. The trained model is validated that can reproduce the authentic Pareto front in a very short time compared with other models.…”

Investigation and Optimisation of High-Lift Airfoils for Airborne Wind Energy Systems at High Reynolds Numbers

“…The NSGA-II genetic algorithm [20] as implemented in CAESES was chosen, as it is well suited for constrained multi-objective problems and generally achieves good convergence for nonlinear problems. A study using NSGA-II in conjunction with XFOIL can be found in [21]. Furthermore, it enables parallel evaluation of up to the population size cases (depending on the computational resources available) and combines design-space exploration and optimisation.…”