Distinct approaches to reproduce hygrothermal behavior of building materials based black-box models
-
Roberto Zanetti Freire
Abstract
The presence of moisture in building envelopes caused by infiltration or condensation, especially in insulation layers, can have serious consequences in the whole-building energy performance and thermal comfort. Accurate prediction of moisture transport in buildings depends on properly understanding how water migrates across an interface, and it is usually performed by associating experimental analysis of different types of porous media or by numerical simulation. With the objective of reducing the energy consumption of buildings, computational tools are being used to simulate new and retrofitting buildings. In this type of application, it is common to find nonlinear behavior affecting temperature and relative humidity profiles in building structures, mainly due to modeling difficulty and highly moisture-dependent properties, increasing the difference between the results found by computational simulations and what happens inside building materials. Based on these concepts, this chapter presents two black-box approaches, with nonlinear identification focus, adopting Multivariate Adaptive Regression Splines (MARS) and Least Squares Support Vector Machines (LS-SVM). The first technique was considered to reproduce the behavior of highly hygroscopic building materials. Considering an experimental data set acquired using an experimental plant developed to study moisture effects on building surfaces. MARS models were built in order to predict heat flux, mass flow rates, and both temperature and relative humidity profiles considering just indoor and outdoor surface temperature and relative humidity as inputs. In the second approach, by adopting multiples MISO (multiple-input, single-output) Nonlinear Auto-Regressive with eXogenous inputs (NARX) models, LS-SVM, a maximum margin model based on structural risk minimization, was used to predict vapor flux, sensible heat flux, latent heat flux and mould growth risk in roofs surfaces. In this second case study, outdoor weather conditions were considered as input for the models. To evaluate the proposed black-box regression and identification techniques, five performance coefficients were analyzed for both training and validation phases. Results of applying artificial intelligence based approaches in predicting the hygrothermal behavior of building materials showed consistent precision when compared to the results of both experimental and numerical model results.
Abstract
The presence of moisture in building envelopes caused by infiltration or condensation, especially in insulation layers, can have serious consequences in the whole-building energy performance and thermal comfort. Accurate prediction of moisture transport in buildings depends on properly understanding how water migrates across an interface, and it is usually performed by associating experimental analysis of different types of porous media or by numerical simulation. With the objective of reducing the energy consumption of buildings, computational tools are being used to simulate new and retrofitting buildings. In this type of application, it is common to find nonlinear behavior affecting temperature and relative humidity profiles in building structures, mainly due to modeling difficulty and highly moisture-dependent properties, increasing the difference between the results found by computational simulations and what happens inside building materials. Based on these concepts, this chapter presents two black-box approaches, with nonlinear identification focus, adopting Multivariate Adaptive Regression Splines (MARS) and Least Squares Support Vector Machines (LS-SVM). The first technique was considered to reproduce the behavior of highly hygroscopic building materials. Considering an experimental data set acquired using an experimental plant developed to study moisture effects on building surfaces. MARS models were built in order to predict heat flux, mass flow rates, and both temperature and relative humidity profiles considering just indoor and outdoor surface temperature and relative humidity as inputs. In the second approach, by adopting multiples MISO (multiple-input, single-output) Nonlinear Auto-Regressive with eXogenous inputs (NARX) models, LS-SVM, a maximum margin model based on structural risk minimization, was used to predict vapor flux, sensible heat flux, latent heat flux and mould growth risk in roofs surfaces. In this second case study, outdoor weather conditions were considered as input for the models. To evaluate the proposed black-box regression and identification techniques, five performance coefficients were analyzed for both training and validation phases. Results of applying artificial intelligence based approaches in predicting the hygrothermal behavior of building materials showed consistent precision when compared to the results of both experimental and numerical model results.
Chapters in this book
- Frontmatter I
- Preface V
- Contents VII
- About the editors XI
-
Part I: Future cities
- Biopotent social technology: occupations park and university extensions 1
- Performance potentials: the optimization of buildings in operation 21
- Climate culture building: comparison of different computer generated building envelope designs for different Brazilian climate zones 35
- Electrical energy efficiency in urban infrastructure systems: nonintrusive smart meter for electrical energy consumption monitoring 47
- Distinct approaches to reproduce hygrothermal behavior of building materials based black-box models 61
-
Part II: Modern urban agriculture
- Investigating the challenges and opportunities of urban agriculture in global north and global south countries 95
- Social technology and urban agriculture in Brazil: the social technology network and the social technology DataBank project 111
- Orchards from the forest: Urban agriculture as a lab for multiple learning 121
-
Part III: Renewable energy
- The challenges of the new energy revolution 137
- Synthesis of inorganic energy materials 159
-
Part IV: Sustainable smart materials
- Nature-inspired smart materials for multifunctional applications 177
- Smart fiber-reinforced polymer composites and their resource-efficient production by means of sensor integration 191
- The role of biologically inspired design to 4D printing development 205
- Influence of different carbon nanotubes types in dynamic-mechanical properties of lightweight carbon felt/CNTs composites 215
- Light-assisted synthesis of colloids and solid films of metallic nanoparticles 225
- The influence of polymeric interlayers on damping behavior of a fiber metal laminate 239
- Piezoresistivity of low carbon nanotubes content in elastomeric polymer matrix 259
- Improvement of fatigue strength of carbon fiber reinforced polymers by matrix modifications for ultrafast rotating flywheels 279
- Experimental study of thermal conductivity, viscosity and breakdown voltage of mineral oil-based TiO2 nanofluids 290
Chapters in this book
- Frontmatter I
- Preface V
- Contents VII
- About the editors XI
-
Part I: Future cities
- Biopotent social technology: occupations park and university extensions 1
- Performance potentials: the optimization of buildings in operation 21
- Climate culture building: comparison of different computer generated building envelope designs for different Brazilian climate zones 35
- Electrical energy efficiency in urban infrastructure systems: nonintrusive smart meter for electrical energy consumption monitoring 47
- Distinct approaches to reproduce hygrothermal behavior of building materials based black-box models 61
-
Part II: Modern urban agriculture
- Investigating the challenges and opportunities of urban agriculture in global north and global south countries 95
- Social technology and urban agriculture in Brazil: the social technology network and the social technology DataBank project 111
- Orchards from the forest: Urban agriculture as a lab for multiple learning 121
-
Part III: Renewable energy
- The challenges of the new energy revolution 137
- Synthesis of inorganic energy materials 159
-
Part IV: Sustainable smart materials
- Nature-inspired smart materials for multifunctional applications 177
- Smart fiber-reinforced polymer composites and their resource-efficient production by means of sensor integration 191
- The role of biologically inspired design to 4D printing development 205
- Influence of different carbon nanotubes types in dynamic-mechanical properties of lightweight carbon felt/CNTs composites 215
- Light-assisted synthesis of colloids and solid films of metallic nanoparticles 225
- The influence of polymeric interlayers on damping behavior of a fiber metal laminate 239
- Piezoresistivity of low carbon nanotubes content in elastomeric polymer matrix 259
- Improvement of fatigue strength of carbon fiber reinforced polymers by matrix modifications for ultrafast rotating flywheels 279
- Experimental study of thermal conductivity, viscosity and breakdown voltage of mineral oil-based TiO2 nanofluids 290
