In recent years, stochastic modeling has been increasingly applied to investigate the uncertainties of input parameters in hygrothermal simulation and the moisture damage risks of building envelopes. Generally, stochastic modeling requires hundreds or even thousands of simulations to take into account the uncertainties of input parameters, which is computationally intensive and timeconsuming. This paper aims to apply polynomial and neural network metamodel as a substitute for the traditional hygrothermal model, to predict the hygrothermal performance of building envelopes. In the previous study carried out by the authors, stochastic simulations have been performed based on the traditional hygrothermal model, to investigate the hygrothermal performance of wood-frame walls under different rain leakage levels. The material properties and rain deposition factors were considered as stochastic variables, and stochastic simulations were performed under three rain leakage scenarios: 1%, 0.5% and 0.1% of wind-driven rain. In this paper, the stochastic inputs (the hygric material properties and rain deposition factor) and outputs (the maximum moisture content and mold growth index over a 5-year period of the simulation) of a conventional 2×6 wood-frame wall are used to develop the metamodels through polynomial regression and neural network methods. The metamodels are developed for each rain leakage scenario, and the stochastic data of the three rain leakage scenarios are aggregated together to train another metamodel. It is found that the metamodels generally perform well to predict the maximum moisture content and mold growth index. The metamodels for low rain leakage scenarios are better than those for high rain leakage scenarios and the neural network metamodel is more accurate than polynomial metamodel for high rain leakage scenarios, i.e. 1% of rain leakage.

Full document

The PDF file did not load properly or your web browser does not support viewing PDF files. Download directly to your device: Download PDF document
Back to Top

Document information

Published on 25/09/20
Submitted on 21/09/20

DOI: 10.23967/dbmc.2020.095
Licence: CC BY-NC-SA license

Document Score


Views 16
Recommendations 0

Share this document

claim authorship

Are you one of the authors of this document?