Implementation of Machine and Deep Learning Algorithms to Detect Gaps in a Sheet Metal Industry: A Case Study

Abstract

AbstractThis research project is supported by a case company engaged in sheet metal works. The problem of the case company was to detect the minimum gaps on a sheet, as necessary to handle the sheet automatically by a gripper after punching operation. In order to solve such critical problem, the case company is provided images, which were analyzed by using machine learning and deep learning techniques. These images were consisted of the pictures of components produced from the sheet and waste materials after punching it. With human eyes, it can be observed that the gaps between some components are too small. These narrow gaps on the metal sheet after punching the predesigned components become so stuffed for the automated grippers to move the rest of the sheet safely to the warehouse or to the trash bin. Due to such small and weak gaps, automated sheer movement cannot grip properly the rest of the sheet skeleton after punching it. This results in some of the sheet parts being left behind on the worktable and needing human operators to intervene, which resulted an extra trigger and production break. In this study, an approach to machine learning and deep learning algorithms is used for the given images in order to identify which parts of the pictures are actual components and which parts are the waste material. The outcomes from this study contributes successfully to solve the gap problem of the sheet metal and find out a way to optimize the gaps between components that facilitate the easy and safe movement of the grippers and minimize the wastes materials too. This study concludes with several useful recommendations and further study perspectives in future.KeywordsMachine learningDeep learningDetection of gapsSheet metal industryCase study