Thermal optimization of PCM‐based heat sink using fins: A combination of CFD, genetic algorithms, and neural networks

• Published in: Journal of engineering (Stevenage, England) Vol. 2024; no. 10
• Main Authors: Bahrami, Hamid‐Reza, Saberi, Mehdi
• Format: Journal Article
• Language: English
• Published: Wiley 01.10.2024
• Subjects: artificial neural network; genetic algorithm; heat transfer; phase change materials; sustainable development
• ISSN: 2051-3305; 2051-3305
• DOI: 10.1049/tje2.70015

Sustainable development requires focusing on renewable energy, such as solar power, due to the limited availability of fossil fuels. Solar energy, available only during daylight, can be stored in thermal batteries using phase change materials (PCM). However, PCM has low thermal conductivity, slowing its melting process. Fins have been added to enhance heat diffusion, but their optimal design requires further study. This research examines an annular chamber with fixed inner and outer wall temperatures and explores fin configurations. Two series of evenly spaced fins (four and five) were analyzed, with melting times calculated using CFD simulations. Larger fins reduce melting time but limit PCM storage. Effectiveness, defined as the PCM space over melting time, was used to evaluate performance. Higher boundary temperatures decreased melting time but did not always increase effectiveness. The study also used artificial neural networks (ANN) combined with a genetic algorithm to predict optimal conditions. The configuration with five fins, a 90°C boundary temperature, a fin length of 45 mm, and a thickness of 5.4 mm achieved the highest effectiveness of 4.7, showing that smaller fins at lower temperatures are more efficient. This study investigates using fins to enhance heat transfer in a cylindrical thermal energy storage system with phase change materials (PCMs). The researchers optimized fin configuration for maximum effectiveness, defined as available PCM storage space over melting time, using computational fluid dynamics (CFD) simulations and artificial neural networks. The optimal configuration was five fins with a length of 45 mm, thickness of 5.4 mm, and boundary temperature of 90°C, achieving an effectiveness of 4.7.