The concept of inverse design is widely used in the aviation industry to design and optimize efficient components. Aerodynamic shape optimization [1] is a complex process where the optimal shape of the airfoil is achieved from the specified pressure distribution. The algorithms used to optimize airfoil shapes can be broadly divided into gradient-based and gradient-free algorithms [2]. The gradient-free algorithm finds the global solution of the optimization problem [3] but the algorithm is complex, and the rate of convergence is very slow. On the other hand, gradient-based algorithm requires the design shape to be continuous, any discontinuity in the airfoil shape hampers the robustness and convergence of the algorithm [4].Deep learning (DL) has received considerable attention in recent years due to its robust predictive capability to model complex problems. The is an increasing interest in the community to use DL for various applications, including RANS modelling, aerodynamic predictions and turbomachinery design using MLP (Multi-Layered Perceptron), etc.This paper demonstrates the predictive capability of two neural network-based models to inverse design the airfoil shapes. The data for modelling the neural networks is obtained from Xfoil software, monitored by a programmed Python code. Convolutional Neural Network (CNN) and Deep Neural Network (DNN) models are developed to predict the airfoil shapes using the pressure coefficient distribution (Cp). The pressure coefficient distribution in pictorial form is provided as an input for training the deep CNN model. On the other hand, the numerical coordinates of pressure coefficient (Cp) serves as input for Deep Neural Networks (DNN). For both the models, the critical hyperparameters, namely – learning rate, number of epochs and batch size, are tuned to reduce the mean squared error (MSE) and increase the accuracy of the airfoil shapes. Further information on the deep learning models, results from tuning hyperparameters, and the airfoil shapes will be provided in the detailed version of the manuscript.References:[1] A. Jameson, “Aerodynamic design via control theory,” Journal of scientific computing, vol. 3, no. 3, pp. 233–260, 1988.[2] S. N. Skinner and H. Zare-Behtash, “State-of-the-art in aerodynamic shape optimisation methods,” Applied Soft Computing, vol. 62, pp. 933–962, 2018.[3] G. Venter and J. Sobieszczanski-Sobieski, “Particle swarm optimization,” AIAA journal, vol. 41, no. 8, pp. 1583–1589, 2003[4] L. A. Catalano, D. Quagliarella, and P. L. Vitagliano, “Aerodynamic shape design using hybrid evolutionary computing 
and multigrid-aided finite-difference evaluation of flow sensitivities,” Engineering Computations, 2015

Inverse Design of Airfoils Using Convolutional Neural Network and Deep Neural Network

Paper Proceedings: https://asmedigitalcollection.asme.org/GTINDIA/GTINDIA2021/volume/85536

ASME 2021 Gas Turbine India Conference (GTINDIA2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.