Author: Ochilov, Murodjon Ashurqulovich; Tojiboyev, Suhrobxon Ja’far o‘g‘li

Annotation: The packaging stage in modern industrial production is one of the most critical links in quality control. This paper presents a combined intelligent control system for a drywall board packaging line integrating three modules: a ResNet-50 visual classifier via transfer learning; a discrete Kalman filter sensor fusion mechanism; and a PLC PI controller with D-decomposition parameters. Industrial validation was conducted at four plants (Tashkent, Samarkand, Fergana, Bukhara) over 504 working days. Results: accuracy 93.5 ± 0.8% (F1 = 0.939; ROC-AUC = 0.978); missed-defect rate reduced from 11.7% to 2.9%; throughput increased 4.4×; waste reduced 17.0 pp; operator cognitive load reduced 57.6% (NASA-TLX, p < 0.001). Payback: 14–15 months.

Keywords: computer vision; neural network; ResNet-50; Kalman filter; sensor fusion; quality control; drywall packaging; PLC control; Industry 4.0; multi-site validation; NASA-TLX; D-decomposition.

Pages in journal: 502 - 511