Low-cost system for visual inspection of corrosion: an industrial case study

The use of digital technologies around the world has increased considerably, modifying the way in which daily activities are conducted. The manufacturing sector is no exception. Over the last decade, digital technologies have become a key element for manufacturing companies to deliver high quality products, which is a critical factor in their success and competitiveness. However, for most companies, cost and lack of understanding of the benefits of digital solutions are still a main barrier to digitalisation of their operations. In this paper, a low-cost visual system is proposed and developed for in-process detection of defective parts in a manufacturing company that produces hoses with fittings and connectors. The company often faces problems with the angle, length, and condition of parts, particularly with brazing residues and corroded surfaces, causing rework and rejection of products by clients. The proposed low-cost system is based on a deep learning approach and uses an off-the-shelf camera to identify corrosion in parts before assembly. Using a low-cost solution contributes to optimizing costs and operations in the production line, reducing waste and rework due to the assembly of defective parts, and minimizing human errors due to differing expertise or diverse criteria. Also, the solution can potentially be implemented in other production lines, reducing the learning curve as regards implementation and extending the lessons learned to the solution of other problems.