On-site process monitoring and autonomous control of a CNC milling machine using CNN and Bayesian optimization

Abstract

In this thesis, the feasibility of implementing a cheap and reliable monitoring system to check the quality of the machined surfaces was researched. The proposed system only needs a small camera mounted next to the tool. Different monitoring CNN architectures were trained and compared in order to choose the most suitable one. Additionally, an optimisation system using Bayesian optimisation was implemented to choose the best machining parameters at priori and use that knowledge to take actions when the machining results were not as expected, testing the model’s performance as an online optimiser directly connected to the monitoring system. The created system, if further improved, will be capable of performing in an industrial machining environment and can easily be implemented, opening the possibility of introducing autonomous and automated systems even in the small to medium companies’ reality.