Forming Regression-based Mathematical Models to Predict PET POY Yarn Properties in the Case of Changing Production Parameters

Yildirim K., ULCAY Y., KOPMAZ O.

TEXTILE RESEARCH JOURNAL, vol.80, no.5, pp.411-421, 2010 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 80 Issue: 5
  • Publication Date: 2010
  • Doi Number: 10.1177/0040517509346437
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.411-421
  • Keywords: PET yarn, spinning factors, physical and performance parameters, regression-based mathematical model, NEURAL-NETWORK, AREA
  • Bursa Uludag University Affiliated: Yes


This study comprises investigations of the effect of PET (polyethylene terephthalate) POY (poly oriented yarn) production parameters on the crystallinity degree, which is involved in the structure of the yarn and performing prediction equations based on a non-linear regression mathematical model. Although there are many production parameters which affect the yarn properties, quenching air temperature, quenching air speed and winding speed were selected for the POY process. Yarn samples were produced in three different levels of each of selected parameters and tested to investigate how the crystallinity degree changes. Measured properties of PET POY were tensile strength, tensile strain, draw force, crystallinity degree based on differential scanning calorimetry technique, dye uptake (K/S), brightness and boiling water shrinkage. In order to obtain empirical formulas for predicting the change of POY properties with respect to selected production parameters, the yarns were produced in 27 different combinations. The starting point of the empirical equation is based on a completely randomized variance analyses model. The coefficients of the curves fitted were computed by means of non-linear regression analysis. R(2) values for these curves were observed to be highly reliable being about 0.8.