Automation and optimization of in-situ assessment of wall thermal transmittance using a Random Forest algorithm

Abstract

Reducing energy consumption and greenhouse gases emissions is among the main challenges of building sector. It is therefore crucial to know the characteristics of envelopes. There are experimental methods to determine thermal transmittance, but limitations are presented. By using techniques of artificial intelligence, this article solves the limitations of current methods by predicting correctly the thermal transmittance value of ISO 6946 and the building period of a wall with monitored data. The methodology used is extrapolated to any country: 163 real monitorings and 140 different typologies of walls have been combined to generate the dataset (22,820 items). The results show the optimal operation of the Random Forest algorithm because both the thermal transmittance of ISO 6946 and the building period are determined by using the most common methods: the heat flow meter method and the thermometric method. This study makes progress towards more automatized processes to characterize thermal transmittance. © 2019 Elsevier Ltd