Productivity estimation of bulldozers using generalized linear mixed models

The productivity estimation of construction machinery is a significant challenge faced by many earthmoving contractors. Traditionally, contractors have used manufacturers' catalogues or have simply relied on the site personnel's experiences to estimate the equipment production rates. However, various studies have demonstrated that typically, there are large differences between the estimated and real values. In the construction research domain, linear regression and neural network methods have been considered as popular tools for estimating the productivity of equipment. However, linear regression cannot provide very accurate results, while neural network methods require an immense volume of historical data for training and testing. Hence, a model that works with a small dataset and provides results that are accurate enough is required. This paper proposes a generalized linear mixed model as a powerful tool to estimate the productivity of Komatsu D-155A1 bulldozers that are commonly used in many earthmoving job sites in different countries. The data for the numerical analysis are collected from actual productivity measurements of 65 bulldozers. The outputs of the proposed model are compared with the results obtained by using a standard linear regression model. In this manner, the capabilities of the proposed method for accurate estimations of productivity rates are demonstrated.