Improvement of Thermal Insulation Properties of Polyester Nonwoven and Estimation of Thermal Conductivity Coefficients Using Artificial Neural Network


Eyupoglu C., Eyupoglu S. , Merdan N.

JOURNAL OF TESTING AND EVALUATION, cilt.47, ss.1075-1086, 2019 (SCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 47 Konu: 2
  • Basım Tarihi: 2019
  • Doi Numarası: 10.1520/jte20180129
  • Dergi Adı: JOURNAL OF TESTING AND EVALUATION
  • Sayfa Sayıları: ss.1075-1086

Özet

In this study, polyester, i.e., Poly(ethylene terephthalate) (PET) nonwoven fabric, was coated with white tuff, perlite, and volcanic stone powder at rates of 10, 20, 30 and 40 % so as to increase the thermal insulation properties of PET nonwoven fabric. In order to apply white tuff, perlite, and volcanic stone powder to PET nonwoven fabric, polyurethane-based coating material was used as a cross-linking agent. The porosity and thermal conductivity coefficients of samples were then investigated as regards the type and concentration of stone powder. Furthermore, three-layer feed-forward artificial neural network (3FFNN) was used to estimate and verify the accuracy of the thermal conductivity coefficient of PET nonwovens coated with white tuff, perlite, and volcanic stone powder. The results showed that perlite stone powder provided higher thermal insulation compared to white tuff and volcanic stone powder. Moreover, thermal insulation coefficient of samples increased with the rise in concentration of white tuff, perlite, and volcanic stone powder. Besides, the accuracy of 3FFNN was 99 %. Artificial neural network (ANN)-based results showed that the thermal conductivity coefficients of samples with four different concentrations obtained from white tuff, perlite, and volcanic stone powder were almost the same for experimental and ANN-trained models. According to the results, it was seen that 3FFNN was correctly modeled, and the prediction of the thermal conductivity coefficients was successfully realized.