Modeling and optimizing yarn characteristics in sustainable cotton blends through multiple linear regression and artificial neural networks

The textile industry is undergoing a rapid transition toward the use of recycled fibers, driven by environmental concerns, the adoption of circular textile strategies, fluctuating prices of virgin fibers, and increasing demand from environmentally conscious consumers for sustainable clothing. In response to these shifts, the present study employed a practical approach to develop a sustainable yarn by maximizing the use of recycled cotton derived from spinning mill hard waste. Since recycled cotton exhibits poor spinnability due to its lower fiber strength, and high content of short fibers and neps, appropriate amounts of virgin cotton and Tencel were incorporated to enhance processability and improve yarn strength and elongation, making the yarn suitable for industrial fabric production. Determining the ideal blend ratio of the three fibers—while balancing high recycled content with acceptable yarn quality—was a complex and time-intensive task. To address this, MATLAB software was employed to generate a good number of experimental blends with varying fiber proportions. The resulting yarns were tested for quality parameters and analyzed using Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) models. These models developed predictive equations and suggested optimal blend ratios with highest probability of achieving the targeted yarn properties. Between the two, the ANN-based optimized blend produced yarn properties that matched more closely with predicted values, suggesting improved fit within the investigated design space when compared with the MLR model.