ANFIS grid partition framework with difference between two sigmoidal membership functions structure for validation of...

Area of Science:

• Fluid Dynamics
• Artificial Intelligence
• Computational Science

Background:

• Modeling fluid flow and temperature distribution in cavities is crucial for various engineering applications.
• Traditional Computational Fluid Dynamics (CFD) methods can be computationally intensive.
• Artificial intelligence (AI) offers potential for more efficient simulation and prediction.

Purpose of the Study:

• To model a square cavity using CFD and an AI approach.
• To assess CFD outputs using the ANFIS algorithm.
• To compare the predictive accuracy and efficiency of ANFIS against the ant colony method.

Main Methods:

• Computational Fluid Dynamics (CFD) was used to simulate fluid flow and temperature in a square cavity with copper nanoparticles as nanofluid.
• The Adaptive Neuro-Fuzzy Inference System (ANFIS) was employed to analyze CFD outputs, combining coordinate and flow parameters.
• Data classification used the grid partition method with a dsigmf membership function.

Main Results:

• The ANFIS method achieved a high prediction accuracy (R=0.99), closely matching CFD results.
• ANFIS demonstrated superior performance and pattern recognition compared to the ant colony method (R=0.91).
• The combined CFD-ANFIS approach enables rapid prediction of flow and temperature distribution.

Conclusions:

• The ANFIS method is a powerful and efficient tool for predicting fluid flow and thermal distribution in cavities.
• This hybrid CFD-ANFIS approach significantly reduces computational time.
• ANFIS offers a meshless learning step without numerical instability or boundary condition limitations.