Computational fluid dynamics and machine learning integration for evaluating solar thermal collector efficiency...

Area of Science:

• Renewable Energy Engineering
• Computational Science
• Thermal Engineering

Background:

• Optimizing solar thermal collectors for food dehydration is crucial for efficient renewable energy utilization.
• Balancing computational efficiency and prediction accuracy is a key challenge in thermal system design.
• Micro-heat pipe arrays (MHPA) offer potential for enhanced thermal performance in solar collectors.

Purpose of the Study:

• To develop and validate a novel hybrid computational framework integrating CFD and machine learning.
• To optimize solar thermal collectors with MHPA for food dehydration applications.
• To enhance computational efficiency and prediction accuracy in thermal system design.

Main Methods:

• A validated CFD model generated 935 numerical cases for training and evaluating machine learning algorithms (LR, SVR, ANN).
• Entropy analysis was employed to quantify information transfer and identify influential design parameters.
• Interpretability analysis determined optimal operating ranges for key collector parameters.

Main Results:

• Optimized SVR and ANN models achieved high prediction accuracy (R² = 0.96 and 0.94) compared to LR (R² = 0.61).
• MHPA thermal conductivity, air inlet temperature, and air velocity were identified as the most influential parameters.
• Optimal performance was achieved with over 600 data samples, with a critical transition observed between 100-300 samples.

Conclusions:

• The hybrid CFD-ML framework offers an efficient approach for optimizing solar thermal collector designs.
• The methodology provides valuable insights for improving solar food drying systems with reduced computational costs.
• Identified key parameters and optimal operating ranges can guide future solar collector design and development.