AEROSPACE
A tool for real-time airfoil
aerodynamic performance prediction

CLIENT
University of Padua
CHALLENGE
Innovative airfoil parametrization, verifying performance in real time
SUCCESS
Automatic generation of high-performing airfoil shapes for wind energy, axial compressors, pumps and other applications
01
NablaFoil is a new approach to the design of airfoils. Through an interactive, real-time mode the user adapts the shape of the airfoil by hand and can check in real-time its performance. A neural network runs in the background, extrapolating results from a huge database of airfoils. The main features:
01
real-time calculation of aerodynamic performance
02
the ability to choose the aerodynamic target in terms of lift and drag coefficients
03
easy change of the parametrization techniques
04
over 500k airfoil shapes populating the airfoil map
02
The parametrization techniques, using the most advanced mathematical representations, make it possible to reproduce all the most common airfoils and to generate new, innovative, shapes. The discretization methods include:
01
Bézier curve
02
B-Spline representation
03
NURBS curve

03
The aerodynamic prediction technique of NablaFoil is based on a Neural Network:
01
first, a huge number of parametrized airfoils — over 500k — was generated
02
then each airfoil was simulated with a panel code and its performance recorded in a database
03
finally, a neural network was set up and trained on that database
The result is a complex mathematical function, able to instantaneously predict the polars of an airfoil from the coordinates of its control points.

04
The NablaFoil Graphical User Interface shows an interactive draw-mode environment linked to two dependent plots of the airfoil polars. Its commands allow the user to:
01
change the view and colour of the airfoil, and zoom the shape
02
add or subtract a point of the parametrization, and move a point along a fixed coordinate
03
identify a random airfoil shape from a target value of CL and CD
Modifying the airfoil shape by hand shows the change in its performance in real time.

05
NablaFoil cuts across a broad range of applications. It is currently used to generate high-performing airfoil shapes, and couples easily with 0D and 1D engineering codes for the estimation of performances. The main fields:
01
Wind energy — generating the sequence of airfoils of a horizontal-axis wind turbine, and identifying the slightly non-symmetric airfoil best suited to a vertical-axis one
02
Axial compressors — the rotor blade is easily optimized with NablaFoil's algorithms
03
Pumps — coupled with a 1D code to predict the performance of axial and radial pumps
04
NablaFoil can also be customized on demand: different parametrization techniques and target plots can be implemented, and airfoil performance maps of different accuracy used, for stall prediction for instance
RECOGNITION
Our team reached the final at the CAE Conference 2018 with NablaFoil.
“We are proud of our new, innovative approach to the design of airfoils.”
Andrea Dal Monte, Ph.D.

Try creating your own neural network. Leave your email to download a sample of the dataset used: