Enhancing airflow dynamics in airjet spinning: A machine learning approach to optimize nozzle design

Abstract

This research delves into using machine learning techniques to enhance the airflow dynamics in Airjet spinning. It focuses on understanding the factors that affect airflow by examining components of a spinning nozzle including the fiber inlet element, injector nozzle and spinning spindle. A prototype Airjet spinning nozzle was developed to evaluate the Intake Airflow and Airflow Rate, which serve as the basis for a simulation model. A total of 501 data points were empirically gathered, and machine learning methods were applied to uncover patterns and make predictions. The study combines conventional measurement techniques with data analysis tools, like linear regression, decision trees, random forest and support vector regression. After developing a applicable machine learning tool, the hyperparameters are optimized with the goal to improve the model’s reliability. Following this optimization process it was found that the CatBoost model outperformed ML models in terms of all performance metrics. Furthermore insights, into how nozzle features impact airflow dynamics were obtained through sHapley Additive exPlanations (SHAP) analysis. The results indicate that specific nozzle design parameters play a significant role in improving airflow dynamics. Integrating machine learning techniques into the design process marks a departure from conventional empirical methods. This scientific method allows for a more rapid and precise Airjet nozzle design process, fostering innovation in textile manufacturing technology.