SurfMon: a learning-based optical in-process surface monitoring system

Metrology is the key to ensuring the quality of manufactured parts. With the fast development in the manufacturing sector, particularly with the sustainable manufacturing requirement, there is a significant demand for manufacturing metrology to ensure not only the accuracy of the final parts but also the quality of the whole manufacturing process. As such, defects in the manufacturing process can be identified as early as possible and can potentially lead to saving of materials, energy, and cost and contribute to sustainability. In this paper, we present SurfMon, an in-process optical surface monitoring system that is portable and powered by laser speckle pattern measurements and a machine learning algorithm, making it highly flexible and able to learn from normal surfaces and detect anomalies in manufacturing processes. SurfMon essentially consists of a laser, a camera, a microcontroller with wireless connection functionality, and some optics. It is battery-powered and rechargeable, making it hassle-free to integrate into manufacturing facilities. Reflective laser speckle patterns from manufactured surfaces are captured, and a principal component analysis model is trained and used to detect defects. The system can be mounted on a machine tool and in-process surface measurements can be performed following the cutting tool. Anomalies are detected whenever they affect the surface topographies. The effectiveness of SurfMon was demonstrated on a CNC machine, and the results showed that SurfMon was promising for in-process surface monitoring in manufacturing processes.