Real-time visual detection and correction of automatic screw operations in dimpled light-gauge steel framing with pre-drilled pilot holes

Martinez Rodriguez, Pablo, Ahmad, Rafiq and Al-Hussein, Mohamed (2019) Real-time visual detection and correction of automatic screw operations in dimpled light-gauge steel framing with pre-drilled pilot holes. Procedia Manufacturing, 34. pp. 798-803. ISSN 2351-9789

Preview

Text 1-s2.0-S2351978919309345-main.pdf - Published Version Available under License Creative Commons Attribution Non-commercial No Derivatives 4.0. Download (1MB) | Preview

Abstract

Modular and panelized construction have been promoted and recognized globally as advanced construction techniques for residential and commercial industries alike. Light-Gauge Steel (LGS) panels have become more popular for commercial buildings and high-rise residential buildings in the last decades. When constructing such panels, for ease of manufacturing and assembling, a common practice in the construction industry is the use of dimples and pre-drilled pilot holes. Current automatic LGS machinery, however, is not designed to operate with such constraints. In this study, a real-time vision-based approach is proposed to enable current machinery to use dimpled studs with pre-drilled pilot holes. An algorithm designed for hole detection inside the dimples on LGS steel studs, based on edge detection and ellipse fitting is proposed. Finally, an adaptive approach is proposed to adjust the screw driving manipulators to ensure that the drilling operation occurs accurately, avoiding any possible damage to the LGS studs or failure of the screwing operation. The proposed algorithm is validated on a real steel assembly and a comparison is provided with other well-known algorithms for ellipse detection to demonstrate the effectiveness of the proposed method. This real-time algorithm gives real-time results for pilot hole detection and screwing location estimation within 3 mm tolerance. When compared with other well-known approaches in the literature, the proposed approach is 59% more accurate than the fastest available algorithm.

Actions (login required)

View Item

Downloads