AN ARTIFICIAL NEURAL NETWORK AND TAGUCHI INTEGRATED APPROACH TO INVESTIGATION OF HEAT TRANSFER AND PRESSURE DROP IN THE SOLAR AIR HEATER

ATA İ., ÇAKIROĞLU R., ACIR A.

HEAT TRANSFER RESEARCH, cilt.54, sa.3, ss.17-35, 2023 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 54 Sayı: 3
  • Basım Tarihi: 2023
  • Doi Numarası: 10.1615/heattransres.2022044331
  • Dergi Adı: HEAT TRANSFER RESEARCH
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Communication Abstracts, Compendex, INSPEC, Metadex, Civil Engineering Abstracts
  • Sayfa Sayıları: ss.17-35
  • Anahtar Kelimeler: solar air heater, heat transfer, friction factor, Taguchi method, artificial neural network
  • Gazi Üniversitesi Adresli: Evet

Özet

In this paper, an artif icial neural network (ANN) and Taguchi method integrated approach to investigation of the Nusselt number (Nu) and friction factor (f) with circular ring turbulators under solar radiation has been used. The integrated ANN and Taguchi method was applied to examine the effect of three different parameters - pitch ratio (PR), hole number (HN), and Reynolds number (Re) - on the Nusselt number (Nu) and friction factor (f) under solar radiation heat flux (I). The Taguchi experimental design was conducted using L9 orthogonal array to determine the factor levels of the optimal design. The optimum parameters affecting Nu and f in a solar air heater (SAH) were determined as A3B1C1 and A3B3C3, respectively. The contribution ratios for the test parameters affecting Nu and f were evaluated by using ANOVA analysis. It was shown that the Re number has an effect on Nu with 40.03% contribution ratio, whereas HN has a contribution ratio of 57.24% on f in SAH. Also, Nu and f empirical equations for correlations were obtained with the help of the ANN model. The results of estimation of the ANN model were compared with those of experimens and the obtained results showed a good correlation coefficient (R2) by f inding between 0.9950-0.9987 for the test data set.