A Method of Automated Work Observation for Ship Production Using Deep Neural Networks

Abstract

It is important to increase the productivity of every shipyard. Visualizing the actual work status during any industrial activity is essential. Work observation as one of the methods of industrial engineering has been applied in various fields in shipyards in Japan to increase productivity. However, current work observation requires both time and labor, and in some cases, shipyards hesitate to implement work observation. The aim of this study is to develop a methodology that uses deep neural networks to reduce the disadvantages of current work observation approaches while identifying work tasks and the accuracy of this observation.