Akademik Veri Yönetim Sistemi
ULUSLARARASI 11. OTOMOTİV TEKNOLOJİLERİ KONGRESİ, Bursa, Türkiye, 9 - 10 Eylül 2024, cilt.1, ss.706-719, (Tam Metin Bildiri)
This study investigated the detectability of delamination damage modes in composite materials with artificial intelligence using machine learning and image processing methods with the YOLO algorithm. The advantages of image processing techniques compared to human inspection are discussed in detail. How these techniques contribute to preventing humaninduced errors has been examined and developed. With conventional and conventional methods, the detection of production defects is based on manual or human eye inspection. In this context, carelessness and fatigue can lead to erroneous results. On the other hand, image processing technologies can minimize such errors by working with high precision and consistency. In this case, "zero defect" can correspond to the concept of excellence in various industry sectors. The primary purpose of this study is to demonstrate the detectability of defects in composite materials using image processing technologies. The methodology used in the study starts with a detailed data collection process for structures that delaminate, which is a mode of damage in composite materials, and includes careful labeling of these data, followed by training the model. During the data collection phase, a wide variety of learning was provided with images taken in different rotations. In this way, the risk of "overfitting", which is defined as overlearning in the image processing model, has been reduced, and the ability to generalize has been increased. Each image was carefully examined during the labelling phase, and the defects that needed to be detected were identified and labelled in the appropriate format. Finally, during the model training phase, this labelled data was trained with the YOLO algorithm and the model's ability to detect defects with machine learning was optimized. In this context, it has enabled the active detection and quality control of faults in widely used industrial materials, automotive, offshore wind turbine blades, and various components.