MACHINE LEARNING MODEL FOR PREDICTING DIMENSIONAL STABILITY IN TEXTILE DYEING AND FINISHING PROCESSES
Mounika S, Dharshini C, Muhammed Zaid M
Keywords:Dimensional stability, Machine learning, Predictive modeling, Process optimization, Sustainability, Textile finishing.
Abstract
Among the textile dyeing and finishing processes, dimensional stability is considered one of the main quality variables. Existing methods for the control of dimensional stability are highly experience dependent and empirical. The paper explains how the use of machine learning will be able to predict the dimensional stability of the fabric before the finishing process. The type of fabric, yarn count, weave, and density represent the fabric properties; process variables include the dyeing and finishing processes, temperature, pH values, time, tension, and stretch. Finally, metadata is included. The output consists of a dimensional change percentage under standard conditions. A comparative performance analysis of various machine learning classifiers includes XGBoost, LightGBM, CatBoost, Random Forest, and linear regression. Cross-validation and widely used evaluation criteria will be applied: MAE, RMSE, R². In this respect, the paper will make sure machine learning is interpretable by using SHAP values and permutation feature importance. The approach will define which factors are significant in changing the dimensional stability of fabrics.The findings indicate the ability to use machine learning to accurately forecast shrink and growth, thereby minimizing guesswork, defects, and inconsistencies within the process. It formulates the best possible implementation plan, starting from the testing phase to the eventual incorporation of the process within the overall quality control system. Therefore, the method will enable predictive process control, promote sustainable manufacturing, and optimize process efficiencies in the textile processing industry.