AVALIAÇÃO DE COMPORTAMENTOS ADAPTATIVOS MOTIVADOS POR DESCONFORTO TÉRMICO EM ESCRITÓRIOS:

UMA ABORDAGEM PROBABILÍSTICA

Abstract

This article aims to assess the occupants' adaptive behaviours in offices when they feel hot or cold discomfort. A questionnaire was applied to users of office spaces at the Federal University of Santa Catarina and used as a basis for the analyses. Probabilistic approaches were applied to evaluate occupants' first and second actions when they feel thermal discomfort at work. Those approaches include both uni and bivariate probability density function, as well as a machine learning algorithm (Bayesian Network) to dynamically assess how personal and contextual variables in the work environment impact occupants' actions. Results show that the most frequent adaptations to cold discomfort involve personal adaptations and adjustments in building systems that do not directly consume energy. Although these options are also significant when considering adaptations to hot discomfort, air conditioning systems are also frequent in this scenario. Results from the Bayesian Networks showed significant variations in the most likely adaptive behaviours from different occupant profiles: 93.7% chance of personal adaptation as the first response to hot discomfort for a female user profile and 88.5% chance of using the air-conditioning system as the first option for a male user profile. It was concluded that the use of Bayesian Networks in studies of this field allows estimating information that was not initially evident, relying on conditional interaction between variables.