Automating excavator productivity measurement using deep learning

Heavy equipment represents a major cost element and a critical resource in large infrastructure projects. Automating the measurement of its productivity is important to remove the inaccuracies and inefficiencies of current manual measurement processes and to improve the performance of projects. Existing studies have prevalently focused on equipment activity recognition using mainly vision-based systems that require intrusive field installation and the application of more computationally demanding methods. This study aims to automate the measurement of equipment productivity using a combination of smartphone sensors to collect kinematic and noise data and deep learning algorithms. Different combination inputs and deep learning methods were implemented and tested in a real-world case study of a demolition activity. The results demonstrated a very high accuracy (99.78%) in measuring the productivity of the excavator. Construction projects can benefit from the proposed method to automate productivity measurement, identify equipment inefficiencies in near real time and inform corrective actions.