Conference Agenda

Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

Session Overview
Session 05: Artificial neural network (ANN)
Wednesday, 25/Aug/2021:
10:30am - 11:15am

Session Chair: Prof. Ryozo Ooka, The University Of Tokyo
Location: Room 5 - Room 019, Building: 116

Show help for 'Increase or decrease the abstract text size'
10:30am - 10:45am

Thermal transmittance prediction based on the application of artificial neural networks on heat flux method results

Sanjin Gumbarević, Bojan Milovanović, Mergim Gaši, Marina Bagarić

University of Zagreb, Faculty of Civil Engineering, Croatia

Deep energy renovation of building stock came more into focus in the European Union due to energy efficiency related directives. Many buildings that must undergo deep energy renovation are old and may lack design/renovation documentation, or possible degradation of materials might have occurred in building elements over time. Thermal transmittance (i.e. U-value) is one of the most important parameters for determining the transmission heat losses through building envelope elements. It depends on the thickness and thermal properties of all the materials that form a building element. In-situ U-value can be determined by ISO 9869-1 standard (Heat Flux Method – HFM). Still, measurement duration is one of the reasons why HFM is not widely used in field testing before the renovation design process commences. This paper analyzes the possibility of reducing the measurement time by conducting parallel measurements with one heat-flux sensor. This parallelization could be achieved by applying a specific class of the Artificial Neural Network (ANN) on HFM results to predict unknown heat flux based on collected interior and exterior air temperatures. After the satisfying prediction is achieved, HFM sensor can be relocated to another measuring location. Paper shows a comparison of four ANN cases applied to HFM results for a measurement held on one multi-layer wall – multilayer perceptron with three neurons in one hidden layer, long short-term memory with 100 units in the hidden layer, gated recurrent unit with 100 units in the hidden layer and combination of 50 long short-term memory units and 50 gated recurrent units in two hidden layers. The analysis gave promising results in term of predicting the heat flux rate based on the two input temperatures. Additional analysis on another wall showed possible limitations of the method that serves as a direction for further research on this topic.

10:45am - 11:00am

Utilizing High Performance Computing to Improve the Application of Machine Learning for Time-Efficient Prediction of Buildings’ Energy and Daylighting Performance

Rania Labib

Prairie View A&M University, Texas, United States of America

Architects and engineers often investigate the energy and daylighting performance of hundreds of design solutions and configurations to ensure an energy-efficient solution for their designs. To shorten the time required for daylighting simulations, architects usually reduce the number of variables or parameters of the building and facade design. This practice usually results in the elimination of design variables that could contribute to an energy-optimized design configuration. Therefore, recent research focused on incorporating machine learning algorithms that require the execution of only a relatively small subset of the simulations to predict the daylighting and energy performance of buildings. Although machine learning has been proven to be accurate, it still becomes a time-consuming process due to the time required to execute a set of simulations to be used as training and validation data. Furthermore, in an effort to save time, designers often decide to use a small simulation subset leading to a poorly designed machine learning algorithm that produces inaccurate results. Therefore, this research study aims to introduce an automated and user-friendly framework that utilizes high performance computing to execute the simulations that are needed for the machine learning algorithm while saving time and effort. High performance computing facilitates the execution of thousands of tasks simultaneously in parallel for time-efficient simulation process therefore allowing designers to increase the size of the simulation’s subset. Pairing high performance computing with machine learning allows for accurate and almost-instant building performance predictions.

11:00am - 11:15am

Neural network for indoor airflow prediction with CFD database

Qi Zhou1, Ryozo Ooka2

1Department of Architecture, The University of Tokyo, Japan; 2Institute of Industrial Science, The University of Tokyo, Japan

Energy efficiency and indoor thermal comfort have become important issues in built environment, making it necessary to simultaneously take into consideration of the two issues, building energy performance and indoor environmental quality, at the designing stage. Coupled simulation between building energy simulation (BES) and Computational fluid dynamics (CFD) enables providing each other complementary information with regard to building energy performance and detailed indoor environment conditions; however, longer computation time is generally required by the CFD due to the difference of heat transfer time scale between BES and CFD, and thus annual prediction is almost impracticable.

With the great revolution of deep learning in the last decade, neural networks (NNs) have become one of the most popular artificial intelligence algorithms due to their advanced modelling abilities, high-accuracy prediction capabilities and high-speed computational powers. In this regard, the NN is suitable to work as a surrogate of CFD in the coupled simulation. This research aims to confirm the feasibility of NN for indoor airflow prediction, which extends our previous study from two-dimensional to three-dimensional indoor space for more realistic conditions. The cubic room has one airflow inlet and outlet, respectively, with a heated blockage at the centre as internal heat source. Boundary conditions including inlet flow conditions and envelope surface temperatures are appropriately set to cover multiple status. Validated CFD simulations are carried out to establish database for the NN. The NN receives boundary conditions as input values and outputs velocity and temperature distributions. The results show that the NN accurately reproduces the indoor airflow distribution and temperature stratification of different status. Meanwhile, the computation time required by the NN is significantly less than the CFD. The feasibility of the NN for fast and accurate indoor airflow simulation has been confirmed.

Contact and Legal Notice · Contact Address:
Privacy Statement · Conference: IBPC 2021
Conference Software - ConfTool Pro 2.6.142+TC
© 2001 - 2021 by Dr. H. Weinreich, Hamburg, Germany