CNN-based flow control device modelling on aerodynamic airfoils.

Area of Science:

• Fluid dynamics
• Artificial intelligence
• Renewable energy

Background:

• Wind energy is crucial for sustainable power, but wind turbine efficiency needs enhancement.
• Flow control devices on airfoils are key to improving turbine performance.
• Computational Fluid Dynamics (CFD) is standard, but AI offers faster prediction methods.

Purpose of the Study:

• To develop and validate Convolutional Neural Networks (CNNs) for predicting flow characteristics and aerodynamic coefficients of airfoils with flow control devices.
• To assess the accuracy and computational efficiency of AI models compared to traditional CFD simulations.

Main Methods:

• Conducted 158 CFD simulations of a DU91W(2)250 airfoil with rotating microtabs and Gurney flaps.
• Trained two CNNs: one for velocity/pressure field prediction, another for aerodynamic coefficients.
• Validated CNN predictions against CFD simulation results.

Main Results:

• The CNN for field prediction accurately captured flow characteristics around the devices with minimal error.
• The CNN for aerodynamic coefficients reliably predicted trends and values.
• CNNs reduced computational time by four orders of magnitude compared to CFD.

Conclusions:

• CNNs provide a highly accurate and computationally efficient alternative to CFD for analyzing airfoil flow control.
• AI-driven predictions can accelerate the design and optimization of wind turbine components.
• This approach supports the development of more competitive and sustainable wind energy technologies.