Defect Detection Using Convolutional Neural Network (CNN) for Automotive Parts Remanufacturing

Remanufacturing is recognised as one of the keys to sustainability. However, there is a certain task during remanufacturing that requires labour and skill intensive, which leads to increasing remanufacturing time, cost of production, inconsistency of remanufactured parts and others. There is a great opportunity for a technology shift in the overall remanufacturing process which may reduce costs and improve remanufacturing efficiency. Integration of AI and AM is one of the key areas due to the ability of AI in detecting and determining defects and the ability of AM to manufacture complex parts automatically from computer-aided design (CAD) data. This is considered as one of the potential areas to automate and improve overall remanufacturing activities which are currently too dependent on skilled workers. The study develops Convolutional Neural Network (CNN) for automotive component with the aim of using AM for component remanufacturing process. The CNN model has been developed using YOLOv5 in Google Colab for surface defect detection. The images of defect surfaces have been compiled and trained using the CNN model. The effectiveness of the model is then evaluated and areas for improvement are highlighted. The study demonstrates the application of the CNN model and its significance in supporting automated remanufacturing of automotive components.