Deep Learning Based Anomaly Detection in the 3D Modeling Production Process

3D printing is a technology that is predicted to be one of the most widely used production technologies in the coming years. This technology allows three-dimensional output of products designed using 3D modelling software. Although this technology is an effective method in production, it also has disadvantages. The disadvantages are abnormal developments that may occur during printing. To solve this problem, the study aims to develop a CNN deep learning model to create a image processing system that monitors the printing work and cancels or stops the work when there is a problem. In the study, a dataset consisting of various possible anomaly images from the open access Kaggle website was used. The size of the training, validation and test sets were determined as 1040, 150 and 310 images, respectively, and the partitioning process was randomised. According to the performance measurements made in the testing of the model, it was observed that it successfully detected 98% of the anomalies that may occur during printing. It is predicted that the developed model is one of the successful algorithms that can be used in anomaly detection and will prevent material and time loss due to abnormal printing.