Remanufacturing process improvement by image recognition methods. Application of the mechanical part

Abstract

The paper describes the possibility of using, building, and implementing an image recognition system in a company performing remanufacturing processes. It is based on a thesis prepared with the help of Wabco Reman Solutions. The tests were conducted using one of the parts remanufactured by the company – a manifold. The research focuses on different variants of the obtained image recognition models in order to identify differences that may affect their effectiveness and possible application in real work conditions. The environment used to build the models is Jupyter Notebook, and convolutional neural networks were implemented.