Integration of digital quality control for intelligent manufacturing of industrial ceramic tiles

Conventional quality control (QC) is carried out at the end of the manufacturing for surface defect (SD) detection and separating the products as to quality in ceramic tile manufacturing (CTM) manually. Manual inspection may cause misclassify the products as to quality and so SD may not be analyzed in detail. Unfavorable results may have occurred concerning manufacturing costs in such cases. In this study, Gabor, Steerable Digital Filter and Wiener filter methods-based operations were proposed experimentally for plain tiles to determine surface quality (SQ) as to ISO 10545-2 standard. Besides, deep learning-based methods were used for classifying SD. The defect dataset was created with 150 tile images from crack, fleck, pore, scratch, spot defects. The relative error of defects' calculated areas was found as 8.78508E-4. The ability of the detected areas to represent the actual defect area was proved with the defect location and the angle of the defect centroid with the X-axis unlike the studies which evaluate this ability visually in literature. The classification of defects (crack, scratch) was made with an accuracy of 96%. The effect of the data augmentation method on classification success was evaluated. The studies in this article were conducted with Us,ak Seramik. Consequently, more cognitive, accurate results than the conventional QC were obtained. A computational and, informative system was presented for integration of the digital QC and SD classification in CTM.