Prediction of environmental performance indicators for construction sites based on artificial neural networks

Abstract

The construction sector still contributes on a major scale to negative impacts on the environment, but it is aware of its responsibility for sustainability in the life cycle of a building. The sector has been using performance measurement, specifically environmental indicators, to monitor and manage its impacts. However, managers have not been using the monitored indicators for any managerial decision because they do not have benchmarks to establish performance targets and evaluate their results. This research is an experimental study that aims to develop an equation for the prediction of environmental performance indicators based on the construction progress. For the development of the study, a database of 15 construction sites in Brazil concerning three environmental indicators (water consumption, energy consumption and construction waste generation) established by the PBQP-H (The Brazilian Program for Quality and Productivity in Habitat) was used. The developed software was written in Python language and the model was built with the use of Artificial Neural Networks. From the application of the resulting equations, managers can obtain a benchmark based on the construction progress in which the building is and promote improvements in their environmental performance as well as use such information in the planning stage.