A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Neural networks for metamodelling the hygrothermal behaviour of building components
https://orcid.org/0000-0002-8315-3955
Abstract When simulating the hygrothermal behaviour of a building component, there are many inherently uncertain parameters. A probabilistic evaluation takes these uncertainties into account, allowing a more dependable assessment of the hygrothermal behaviour. However, this often necessitates many Monte Carlo simulations, which easily become computationally inhibitive. To overcome this time-expense problem, the hygrothermal model can be replaced by a metamodel, a much simpler mathematical model which aims at mimicking the original model with a strongly reduced calculation time. In this paper, a metamodel is developed to directly predict hygrothermal time series (e.g. temperature, relative humidity, moisture content), rather than single-valued derived performance indicators (e.g. maximum mould index), as these hygrothermal time series yield more information, and also allow the user to post-process the output as desired. So far, no metamodelling strategies able to tackle time series are available in the field of building physics. Because the hygrothermal response of a building component is highly non-linear and transient, this paper focuses on neural networks for time series, as they have proven successful in many other fields. The performance and training time of three popular types of networks (multilayer perceptron, recurrent neural network, convolutional neural network) is evaluated based on an application example of a massive masonry wall. The results indicate that only the recurrent and convolutional networks are able to capture the complex patterns of the hygrothermal response. Additionally, the convolutional network performed significantly better and was 10 times faster to train for the current application example, compared to the recurrent network.
Highlights A metamodel with a memory mechanism is required to accurately predict hygrothermal time series. Recurrent neural networks and dilated causal convolutional networks are able to capture the complex patterns of the hygrothermal response. To predict the relative humidity, dilated causal convolutional neural networks perform significantly better than recurrent neural networks. Dilated causal convolutional networks are 10 times faster to train on the current example, compared to recurrent neural networks (LSTM, GRU).
Neural networks for metamodelling the hygrothermal behaviour of building components
https://orcid.org/0000-0002-8315-3955
Abstract When simulating the hygrothermal behaviour of a building component, there are many inherently uncertain parameters. A probabilistic evaluation takes these uncertainties into account, allowing a more dependable assessment of the hygrothermal behaviour. However, this often necessitates many Monte Carlo simulations, which easily become computationally inhibitive. To overcome this time-expense problem, the hygrothermal model can be replaced by a metamodel, a much simpler mathematical model which aims at mimicking the original model with a strongly reduced calculation time. In this paper, a metamodel is developed to directly predict hygrothermal time series (e.g. temperature, relative humidity, moisture content), rather than single-valued derived performance indicators (e.g. maximum mould index), as these hygrothermal time series yield more information, and also allow the user to post-process the output as desired. So far, no metamodelling strategies able to tackle time series are available in the field of building physics. Because the hygrothermal response of a building component is highly non-linear and transient, this paper focuses on neural networks for time series, as they have proven successful in many other fields. The performance and training time of three popular types of networks (multilayer perceptron, recurrent neural network, convolutional neural network) is evaluated based on an application example of a massive masonry wall. The results indicate that only the recurrent and convolutional networks are able to capture the complex patterns of the hygrothermal response. Additionally, the convolutional network performed significantly better and was 10 times faster to train for the current application example, compared to the recurrent network.
Highlights A metamodel with a memory mechanism is required to accurately predict hygrothermal time series. Recurrent neural networks and dilated causal convolutional networks are able to capture the complex patterns of the hygrothermal response. To predict the relative humidity, dilated causal convolutional neural networks perform significantly better than recurrent neural networks. Dilated causal convolutional networks are 10 times faster to train on the current example, compared to recurrent neural networks (LSTM, GRU).
Neural networks for metamodelling the hygrothermal behaviour of building components
https://orcid.org/0000-0002-8315-3955
Abstract When simulating the hygrothermal behaviour of a building component, there are many inherently uncertain parameters. A probabilistic evaluation takes these uncertainties into account, allowing a more dependable assessment of the hygrothermal behaviour. However, this often necessitates many Monte Carlo simulations, which easily become computationally inhibitive. To overcome this time-expense problem, the hygrothermal model can be replaced by a metamodel, a much simpler mathematical model which aims at mimicking the original model with a strongly reduced calculation time. In this paper, a metamodel is developed to directly predict hygrothermal time series (e.g. temperature, relative humidity, moisture content), rather than single-valued derived performance indicators (e.g. maximum mould index), as these hygrothermal time series yield more information, and also allow the user to post-process the output as desired. So far, no metamodelling strategies able to tackle time series are available in the field of building physics. Because the hygrothermal response of a building component is highly non-linear and transient, this paper focuses on neural networks for time series, as they have proven successful in many other fields. The performance and training time of three popular types of networks (multilayer perceptron, recurrent neural network, convolutional neural network) is evaluated based on an application example of a massive masonry wall. The results indicate that only the recurrent and convolutional networks are able to capture the complex patterns of the hygrothermal response. Additionally, the convolutional network performed significantly better and was 10 times faster to train for the current application example, compared to the recurrent network.
Highlights A metamodel with a memory mechanism is required to accurately predict hygrothermal time series. Recurrent neural networks and dilated causal convolutional networks are able to capture the complex patterns of the hygrothermal response. To predict the relative humidity, dilated causal convolutional neural networks perform significantly better than recurrent neural networks. Dilated causal convolutional networks are 10 times faster to train on the current example, compared to recurrent neural networks (LSTM, GRU).
Neural networks for metamodelling the hygrothermal behaviour of building components
Tijskens, Astrid (author) / Roels, Staf (author) / Janssen, Hans (author)
Building and Environment ; 162
2019-07-15
Article (Journal)
Electronic Resource
English
Hygrothermal Behaviour and Building Pathologies
| UB Braunschweig | 2021
Hygrothermal behaviour and building pathologies
| TIBKAT | 2021
Hygrothermal Loads of Building Components in Bathroom of Dwellings
| Trans Tech Publications | 2014
Hygrothermal Loads of Building Components in Bathroom of Dwellings
| British Library Conference Proceedings | 2014