Çok Açılı Görüntü İşleme Kullanılarak Hassas Sıralı Fiber Optik Tel Sarımının Çevrimiçi Geometrik Kalite Kontrolü

Creative Commons License

Altın T., Orhan S., Ergül E.

Uluslararası Mühendislik Araştırma ve Geliştirme Dergisi, cilt.18, sa.5, ss.188-204, 2026 (TRDizin)

Özet

Precision sequential wire winding is widely employed in critical manufacturing applications, including fiber-optic and specialty cable production. Irregular winding structures may cause sudden interlayer tension increases during unwinding, leading to wire breakage and communication failure. In conventional processes, defects are identified only after operation completion, resulting in inevitable material and time losses. This study presents an image-processing-based online geometric quality control system for realtime winding monitoring. The coil is inspected from 12 viewing angles at 30° intervals; brightness profiles are extracted from multiple horizontal rows within a region of interest (ROI) and analyzed using adaptive moving-average thresholding. Sub-threshold regions are marked as potential defect zones, and classification is based on the maximum width (d) of consistently detected regions: Normal, Warning, or Fault. The method was validated on a 2 mm diameter fiber-optic wire with 93 rows per layer. Defects with d < 6 pixels (≈0.40 mm) were classified as Normal, 6 ≤ d ≤ 13 pixels (≈0.40–0.80 mm) as Warning, and d > 13 pixels (≈0.80 mm) as Fault. The results obtained from 20 winding layers demonstrated that the proposed method can reliably detect winding defects, achieving an 85% agreement with operator assessments. Furthermore, under a conservative worst-case binary evaluation, the method attained an accuracy of 90.0% and a recall of 100.0%, highlighting its robustness in identifying defective winding conditions.