Nondestructive Estimation of Hazelnut (Corylus avellana L.) Terminal Velocity and Drag Coefficient Based on Some...

Area of Science:

• Agricultural Engineering
• Aerodynamics
• Machine Learning Applications

Background:

• Hazelnut production is globally significant, originating in Turkey.
• Understanding hazelnut aerodynamic properties is vital for designing agricultural machinery, especially for harvesting and post-harvest operations.

Purpose of the Study:

• To estimate the terminal velocity and drag coefficient of hazelnuts.
• To evaluate the effectiveness of machine learning models for predicting these aerodynamic properties.

Main Methods:

• Utilized logistic regression (LR), support vector regression (SVR), and artificial neural networks (ANNs).
• Independent variables included moisture, mass, density, projected area, surface area, and geometric diameter.
• Model performance was assessed using R-squared (R²), mean squared error (MSE), and mean absolute error (MAE) metrics.

Main Results:

• Artificial neural networks (ANNs) demonstrated the highest accuracy.
• ANN models achieved 91.5% accuracy for terminal velocity and 85.9% for drag coefficient based on R².
• SVR and LR models also showed successful estimations, with SVR being the second most successful.

Conclusions:

• Machine learning models, particularly ANNs, can successfully estimate hazelnut terminal velocity and drag coefficient.
• Accurate aerodynamic property data supports improved design and efficiency in hazelnut processing and harvesting equipment.