Estimating Construction Productivity: Neural‐Network‐Based Approach

Abstract

A neural‐network (NN) and observation‐data‐based approach to estimating construction operation productivity is presented. The main reason for using neural networks for construction productivity estimation is the requirement of performing complex mapping of environment and management factors to productivity. A generic description of the proposed approach is provided, followed by an example of an excavation and hauling operation. The example consisted of two neural‐network modules: (1) Estimating excavator capacity based on job conditions; and (2) estimating excavator efficiency based on the attributes of operation elements. An experiment with a desktop excavator model was developed generating sample cycle‐time data for training the first neural network. To provide the training set for the second neural network, a simulation program was developed generating sample production‐rate data. Test results show that the NN approach can produce a sufficiently accurate estimate with a limited data‐collection effort, and thus has the potential to provide an efficient tool for construction productivity estimation.