Open AccessSupervised Learning for Generating Fair Curves with Curvature Boundary Conditions
Author(s) -
Samuel Sudhof,
Norimichi Uemura,
Masaharu Uchiumi
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1909/1/012030
Subject(s) - curvature , computer science , aerodynamics , artificial neural network , measure (data warehouse) , field (mathematics) , flow (mathematics) , rotor (electric) , mathematical optimization , boundary (topology) , quality (philosophy) , supersonic speed , mathematics , artificial intelligence , geometry , data mining , mathematical analysis , aerospace engineering , mechanical engineering , engineering , philosophy , epistemology , pure mathematics
Fairness measures are means to guess if a curve is suited for an aerodynamic application if no knowledge of the flow field is available. Fair curves are proposed as a means to reduce the number of parameters for generating a supersonic rotor blade. A specific new measure of fairness is proposed with the goal of being numerically less demanding to evaluate than previously known measures of fairness. Gradient search for an optimum of this fairness value is nevertheless computationally expensive. Therefore, a neural network was trained using Adamax optimization that can produce curves of acceptable quality instantly for applications where real time response is needed.