Abstract In this research the effect of yarn characteristics such as yarn count, different percent of polyester/cotton fiers in blend yarn, yarn twist and average of the weft yarn tension at different supplied air pressures and weaving machine speed on the weft yarn speed, and weft arrival time were investigated. Blend polyester-cotton yarns (45 samples( were produced by ring spinning method and the percentage of cotton in the composition changed as 0, 25, 50,7 5, and 100%. The samples were used as weft yarn in an air-jet weaving machine for measuring the weft yarn arrival time and its tension. In order to fid a correlation between the yarn properties and the experimental results, the back-propagation neural network model was adapted. The results showed that all of the parameters have a signifiant inflence on the weft arrival time. However, the most important parameter was found to be the loom speed and consequently the applied air pressure on the weft yarn. The presented neural network model can be used to predict the weft yarn velocity for cotton, polyester and cottonpolyester blend yarns on an air-jet loom with a regression coeffiient as high as R2= 0.98. Keywords: weaving, Air-jet weft insertion, Neural Network, Polyester/cotton yarn.