Abstract
Systematic approaches to process optimization are essential for enhancing operational efficiency in industrial manufacturing, particularly in the textile sector. This study investigates and compares the effectiveness of Artificial Intelligence (AI) based modeling and optimization techniques with Grey Relational Analysis (GRA) in determining the optimal operational parameters of the speed frame for producing 100% polyester spun yarn. The key parameters analyzed include twist, break draft, spacer size, and overhang, with Response Surface Methodology (RSM) employed to design the experiments and ensure sufficient variability in the input space. Yarn quality was evaluated using the Grey Relational Grade (GRG), calculated through GRA based on two critical response variables: yarn evenness (CVm%) and imperfection index (IPI). A predictive model based on an Artificial Neural Network (ANN) was developed to estimate the GRG, and its performance was compared to that of the RSM model. Optimization was then conducted using both GRA and the Genetic Algorithm (GA) to assess the comparative efficiency of a multi-criteria decision-making approach and an AI-based optimization technique. The results demonstrated that the ANN model achieved significantly higher prediction accuracy (R(2) ≈ 1) compared to the RSM model (R(2) ≈ 0.65). Furthermore, the GA-based optimization outperformed GRA, resulting in improved yarn quality. The optimal parameter settings identified were twist = 23 TPM, break draft = 1.26, spacer size = 5.1 mm, and overhang = 3.5 mm. These findings highlight the superior performance and reliability of AI-based methods over conventional statistical and multi-criteria decision-making approaches, emphasizing their potential to advance intelligent optimization in textile manufacturing.