Publications / 2018 Proceedings of the 35th ISARC, Berlin, Germany

Machine Learning for Assessing Real-Time Safety Conditions of Scaffolds

Chunhee Cho, Jeewoong Park, Kyungki Kim and Sayan Sakhakarmi
Pages 56-63 (2018 Proceedings of the 35th ISARC, Berlin, Germany, ISBN 978-3-00-060855-1, ISSN 2413-5844)

Researchers have taken advantage of technological advancements to automate construction processes; as a result, significant progress has been made in designing and planning temporary structures. Despite this effort, relatively little attention has been placed on automating the monitoring of safety issues of scaffolding structures, which are one of the major elements used in the construction industry. A need has emerged for a reliable means to assess the safety conditions of scaffoldings. This paper proposes a method of integrating strain - gage sensing with a machine - learning algorithm (support vector machine) to assess the real - time safety conditions of scaffolds. Based on actual strain data of scaffolding members, which were collected using wireless sensors for various loading cases on the scaffolding structure, a support vector machine was applied to differentiate the scaffolding conditions into 'safe', 'overturning', 'uneven settlement', or 'overloading' conditions. Such an automated differentiation of the condition of a scaffold could help to determine whether or not the scaffolding is safe to use without deploying safety inspectors throughout the site. The proposed method was experimentally validated to be successful in estimating the safety condition of a scaffold with an average accuracy of 97.66% for the cases that were tested. The proposed methodology could serve as a real - time monitoring system to determine the status of scaffolding structures. Its application is expected to significantly improve reliability in assessing the safety conditions of scaffolding structures, compared to conventional safety inspections, and to resolve the related safety issues.

Keywords: Scaffolding, real-time monitoring, strain sensor, machine learning, safety, condition assessment, support vector machine