Designing and optimizing a novel heat sink for the enhancement of hydrothermal performances: Modelling and analysis using artificial neural network

Benouis F. Z., Ould Amer Y., ARICI M., Meziane S.

Engineering Analysis with Boundary Elements, cilt.155, ss.766-778, 2023 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 155
  • Basım Tarihi: 2023
  • Doi Numarası: 10.1016/j.enganabound.2023.07.002
  • Dergi Adı: Engineering Analysis with Boundary Elements
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Aquatic Science & Fisheries Abstracts (ASFA), Communication Abstracts, INSPEC, Metadex, zbMATH, DIALNET, Civil Engineering Abstracts
  • Sayfa Sayıları: ss.766-778
  • Anahtar Kelimeler: Artificial neural network, Heat transfer enhancement, Pin-fin heat sink, Pressure drop, PV cooling
  • Kocaeli Üniversitesi Adresli: Evet

Özet

This paper investigates the hydrothermal performance of a new hybrid pin fin heat sink design (HPFHS) in order to enhance its overall performance, the aim was to develop a design that enables efficient cooling while minimizing energy consumption. The outcomes were compared with those of conventional solid pin fin heat sinks (SPFHS), perforated pin fin heat sinks (PPFHS), and pin fin heat sinks with semi-spheres attached on (PFHSSS) for a Reynolds number range of 8731˂Re<26671. The results confirmed that the new heat sink design augments the fluid-solid interaction surface by approximately 62%, resulting in a significant improvement in heat transfer. Additionally, a 7% reduction in pressure drop was achieved compared to classic pin fins, and the overall efficiency of the HPFHS improved by up to 50% compared to SPFHS. To estimate the local Nusselt number, a back-propagation artificial neural network with feed-forward training was employed, considering Reynolds number, pin spacing in both x and y directions, and pin diameter. The linear regression analysis demonstrated the excellent training of the network. This network can be utilized to determine the local Nusselt number for any desired combination of inputs.