Measuring and benchmarking the productivity of excavators in infrastructure projects: A deep neural network approach

Abstract Inefficiencies in the management of earthmoving equipment greatly contribute to productivity gap infrastructure projects. This paper develops and tests a Deep Neural Network (DNN) model for estimating excavators establishing measure their benchmark. After investigating current practices measuring earthwork during 13 interviews with selected industry experts, DNN was developed tested one ‘High Speed rail second phase’ (HS2) sites. The accuracy prediction achieved by evaluated using coefficient determination (R2) Weighted Absolute Percentage Error (WAPE) resulting 0.87 69.64%, respectively. is an adequate level when compared other similar studies. However, according WAPE method, still 10.36% below threshold (i.e. 80%) expected experts. An inspection results over testing period (21 days) revealed better precision days high excavation volumes low volumes. attributed likely involvement manual work archaeologists case site) alongside some excavators, which caused gaps telematics data. indicates that attained adequate, but proposed approach more accurate highly mechanised environment predominantly limited interventions) mixed mechanised-manual working environment. A bottom-up benchmark rate) can be used performance individual or group equipment, through area, whole site also discussed.