From waste to wear: ANN-driven optimization of 100% sustainable tri-blend yarns from recycled cotton, recycled PET, and ecovero

The textile sector is rapidly transitioning toward recycled fibers, driven by environmental concerns, circular economy initiatives, fluctuating virgin fiber prices, and growing demand for sustainable clothing. This study aimed to develop a sustainable, high-performance tri-blend yarn composed of recycled cotton (r-cotton), recycled polyester (r-PET), and Ecovero (a certified sustainable viscose). The blend was designed to achieve a synergistic balance, with r-cotton providing softness, eco-friendliness, and cost efficiency; r-PET contributing durability, quick-drying, and wrinkle resistance; and Ecovero enhancing comfort, absorbency, luster, and drape. Optimizing fiber proportions to maximize recycled content while maintaining low unevenness, imperfections, and hairiness, along with adequate strength and elongation of yarns for industrial fabric production, posed a complex challenge. To address this, MATLAB software was employed, which generated thirteen experimental blends. The yarns were spun, and the yarn properties were evaluated using Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) models. Both models developed predictive equations, and recommended optimized blend ratios with corresponding yarn characteristics. Final yarn samples were produced and compared against predicted values. The results showed that while both models demonstrated reliable predictive capability, the ANN model provided superior accuracy, with its optimized blends aligning more closely with the targeted yarn properties.