The time needed to insert 50/50 polyester/cotton yarns with a nominal count of 20 tex on air-jet looms is investigated. Because weft insertion on air-jet looms is extremely complicated, a statistical approach is followed in order to relate weft insertion behavior to yarn properties. A model is developed using a backpropagation neural network. Weft insertion behavior is related to yarn properties such as actual yarn count, diameter, hairiness, and irregularity. There is a high coefficient of correlation between experimental and calculated values. The results of the neural network are compared with those of multiple linear regression modeling. The effect of the parameters is illustrated using a factorial design on the neural network model.
展开▼
