O potencial das redes neurais artificiais como suporte para o desenvolvimento de índices de sustentabilidade para edifícios habitacionais (ISE-H)

Abstract

The construction industry, by moving a vast production chain, contributes with several and high levels of impact. If, on the one hand, it generates employment, income, taxes, wealth and well-being, on the other hand, it uses natural resources intensely and harmfully, generates a lot of waste and still represents a high risk to the lives of workers. The concept of sustainability has changed the way of planning, designing, building, operating, maintaining buildings and even deconstructing and disposing of waste. From this context, the challenge to measure the effects of this new thinking and acting arises. But, until then, the most recognized methods of analyzing sustainability in civil construction, among them: BREEAM, LEED, DGNB, AQUA-HQE and CASA AZUL+CAIXA, focus on the attribution of points and credits for the implementation of guidelines, strategies and systems, solutions, finally, mainly focused on environmental aspects, not always objectively evaluating the results obtained or the impacts generated, including in the social and economic dimensions. It was found that a more complete, reliable and objective analysis involves a significant number of variables to be aggregated, requiring a tool capable of dealing with this complexity. Thus, the general objective of this work was to analyze the potential of Artificial Neural Networks (ANNs) as a form of support for obtaining Sustainability Indexes for Housing Buildings (SIB-H), based on a balanced set of environmental, social and economic indicators.The Systematic Literature Review (SRL) was used as a research method, complemented by the Narrative Literature Review (NRL), from which 75 indicators and 25 suggestions for indicators were obtained in view of various aspects, such as: emphasis, dimension of sustainability, type of variable, impacts and types of damage, as well as the scope of these damages in geographic and temporal terms. ANN simulations were performed with the support of IBM’s SPSS v23 software, adopting the R8-N Standard Project as a Reference Building, according to ABNT NBR 12721:2006, used to obtain the Basic Unit Cost (BUC/m2) of the buildings. Artificial Neural Networks proved to be a very viable alternative, with strong potential to aggregate a large number of variables (indicators) and generate indices that reveal the distance for better, in relation to the conventional way of construction, expressing different levels of sustainability of the buildings constructed in Brazil.