TY - GEN
T1 - Inverse airfoil design algorithm based on multi-output least-squares support vector regression machines
AU - Zhu, Xinqi
AU - Gao, Zhenghong
N1 - Publisher Copyright:
© Springer Nature Singapore Pte Ltd 2019.
PY - 2019
Y1 - 2019
N2 - Inverse airfoil design algorithm can obtain the airfoil geometry according to the target pressure coefficient (Cp) distribution. Recently, the rapid development of machine learning method provides new idea to solve engineering problem. Multi-output least-squares support vector regression machines (MLS-SVR) is a multi-output regression machine learning method which can make prediction for several outputs simultaneously through learning a mapping from a multivariate input feature space to a multivariate output space. In this paper, MLS-SVR is used to learn the mapping from the Cp distribution to the geometry, which can be seen as a multi-output regression problem. Through iteratively adding the predicted airfoil geometry and its pressure coefficient distribution into the sample database, the precision of MLS-SVR to predict the right airfoil geometry corresponding to the target Cp distribution is improved. A low speed, transonic and supersonic airfoil inverse design problem are used to validate the efficiency of the proposed algorithm, and the experimental results show that the proposed algorithm can save 34.1% and 58.6% CFD evaluations for low speed and transonic cases respectively to obtain satisfactory airfoil.
AB - Inverse airfoil design algorithm can obtain the airfoil geometry according to the target pressure coefficient (Cp) distribution. Recently, the rapid development of machine learning method provides new idea to solve engineering problem. Multi-output least-squares support vector regression machines (MLS-SVR) is a multi-output regression machine learning method which can make prediction for several outputs simultaneously through learning a mapping from a multivariate input feature space to a multivariate output space. In this paper, MLS-SVR is used to learn the mapping from the Cp distribution to the geometry, which can be seen as a multi-output regression problem. Through iteratively adding the predicted airfoil geometry and its pressure coefficient distribution into the sample database, the precision of MLS-SVR to predict the right airfoil geometry corresponding to the target Cp distribution is improved. A low speed, transonic and supersonic airfoil inverse design problem are used to validate the efficiency of the proposed algorithm, and the experimental results show that the proposed algorithm can save 34.1% and 58.6% CFD evaluations for low speed and transonic cases respectively to obtain satisfactory airfoil.
KW - Airfoil inverse design
KW - Machine learning
KW - Multi-output regression
KW - Support vector machines
UR - http://www.scopus.com/inward/record.url?scp=85070754087&partnerID=8YFLogxK
U2 - 10.1007/978-981-13-3305-7_112
DO - 10.1007/978-981-13-3305-7_112
M3 - 会议稿件
AN - SCOPUS:85070754087
SN - 9789811333040
T3 - Lecture Notes in Electrical Engineering
SP - 1412
EP - 1426
BT - The Proceedings of the Asia-Pacific International Symposium on Aerospace Technology, APISAT 2018
A2 - Zhang, Xinguo
PB - Springer Verlag
T2 - Asia-Pacific International Symposium on Aerospace Technology, APISAT 2018
Y2 - 16 October 2018 through 18 October 2018
ER -