<![CDATA[Quality control in cotton yarn manufacturing is affected by the daily variation in the occurrence of defects and the late detection of abnormal events at the packing stage. This study proposes a decision-support workflow that integrates Random Forest regression for daily defect forecasting with Critical-to-Quality (CTQ) analysis to enable proactive quality control in an industrial case. The model is trained and validated with time ordered evaluation using historical production volume and defect counts, and the performance is reported using MAE, RMSE and sMAPE. Results indicate that the model learns to capture routine up-down defect patterns during the training period, but prediction accuracy is degraded on test days with extreme defect spikes. However, despite this limitation, the forecast outputs provide an early warning signal to prioritize inspection and corrective actions on high risk days. The predicted risk levels are then further linked to CTQ mapping to identify dominant defect types and convert model outputs to actionable defect-specific control recommendations. The proposed integration of machine-learning forecasting with shop-floor quality planning enables more timely and targeted quality interventions in cotton yarn packaging operations.]]>
Copyrights © 2026
