by Aydan Aghabayli, Regional Lead for Azerbaijan
The recent integration of Information Technologies with different disciplines has seen Machine Learning (ML) used in several areas and disciplines, including the AEC industry. The increasing use of data in a digital format, as well as the use of BIM models, creates possibilities to apply ML and learn from the available data.
Applying ML techniques to BIM models and the modelling process are among the most prominent topics found in the peer-reviewed literature. Fields of proposed ML application include clash detection (Hu and Castro-Lacouture, 2019), cost estimation (Fiske, 2019; Vitasek and Zak, 2018), conversion from 3D scan point cloud to BIM (Babacan et al., 2017; Keshavarzi et al., 2020), digital fabrication (Ramsgaard Thomsen et al., 2020), scheduling in construction (Aibinu and Jagboro, 2002; Siu et al., 2013), energy modelling or sustainability (Ma et al., 2019), among others.
The applications in the field of architectural layout design mostly involve the use of Deep Learning (DL) algorithms and Generative Adversarial Networks (GAN) algorithms in particular. The main reason for GANs popularity is that the algorithms use pixel image as an input (a common information representation form for architectural layout). GAN algorithms were applied in a number of applications, like automatic space generation, as suggested in the paper by Huang and Zheng (2018). The use of GANs in layout generation was also proposed by Stanislas Chaillou (2019). In the research work “AI + Architecture”, Chaillou studied the general relation and link between AI and architecture. The author proposed a procedure for an automatic plan generation in which the machine generates new layouts based on a predefined outer contour in addition to the vertical windows distributed along that layout boundary (Figure 1).
Moreover, the paper presented by Nauata et al. (2020) continues to explore the use of GANs in layout design. The research application called “House-GAN” creates a framework for the use of Relational GANs. Input data for the algorithm was a realistic and diverse dataset of layouts and bubble diagrams (connectivity graphs). The research aimed to generate new layouts based on input graphs; generated layouts were rated by professionals afterwards. The research also includes a comparison to previous research on the same topic (Nauata et al., 2020).
In summary, the growing popularity of the use of DL in layout identification and creation is evident. Whereas 2D information is still the most common data source widely available in the industry, the increasing momentum in representing the built environment in 3D formats justifies discussing approaches relying on this complete source of information.
Although GANs techniques started to use 3D data as an input, the utilization of BIM Models has been limited. The discussion on the ability to automatically recognize patterns from layouts, even from more unconventional work of some renowned architects, is presented in As et al. (2018) research. The authors extract the input graphs from BIM models (Figure 2) by making use of conventional BIM tools. Nodes were represented by room parameters (perimeter, area), and edges represented the connections between the spaces. Although the research was conducted using only 15 BIM models, the authors stated the need for more input data to verify obtained results and methodology. However, the authors still consider this approach verified for further development (As et al., 2018).
In conclusion, after exploring case studies of ML applications in the AEC industry, ML can be identified among future tools for extracting knowledge from the data available in the industry with particular importance of BIM models.
Acknowledgements
This paper is an output of master’s dissertation in BIM A+ program. I would like to acknowledge my supervisors Bruno Figueiredo, José Granja, Ricardo de Matos Camarinha and Manuel Esteves Luís for their contribution to current research.
References
Aibinu, A.A., Jagboro, G.O., 2002. The effects of construction delays on project delivery in Nigerian construction industry. International Journal of Project Management 20, 593–599.
As, I., Pal, S., Basu, P., 2018. Artificial intelligence in architecture: Generating conceptual design via deep learning. International Journal of Architectural Computing 16, 306–327.
Babacan, K., Chen, L., Sohn, G., 2017. Semantic Segmentation Of Indoor Point Clouds Using Convolutional Neural Network. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences 4, 101–108.
Chaillou, S., 2019. AI + Architecture.
Fiske, J., 2019. Towards Automated Cost Analysis , Benchmarking And Estimating In Construction : A Machine Learning Approach.
Hu, Y., Castro-Lacouture, D., 2019. Clash Relevance Prediction Based on Machine Learning. Journal of Computing in Civil Engineering 33, 04018060.
Huang, W., Zheng, H., 2018. Architectural drawings recognition and generation through machine learning. Recalibration on Imprecision and Infidelity – Proceedings of the 38th Annual Conference of the Association for Computer Aided Design in Architecture, ACADIA 2018 156–165.
Keshavarzi, M., Afolabi, O., Caldas, L., Yang, A.Y., Zakhor, A., 2020. GENSCAN: A generative method for populating parametric 3D scan datasets. arXiv 1, 91–100.
Krijnen, T., Tamke, M., 2015. Assessing Implicit Knowledge in BIM Models with Machine Learning. Modelling Behaviour.
Ma, G., Liu, Y., Shang, S., 2019. A building information model (BIM) and artificial neural network (ANN) based system for personal thermal comfort evaluation and energy efficient design of interior space. Sustainability (Switzerland) 11.
Nauata, N., Chang, K.H., Cheng, C.Y., Mori, G., Furukawa, Y., 2020. House-GAN: Relational Generative Adversarial Networks for Graph-Constrained House Layout Generation. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 12346 LNCS, 162–177.
Newton, D., 2019. Generative Deep Learning in Architectural Design. Technology Architecture and Design 3, 176–189.
Ramsgaard Thomsen, M., Nicholas, P., Tamke, M., Gatz, S., Sinke, Y., Rossi, G., 2020. Towards machine learning for architectural fabrication in the age of industry 4.0. International Journal of Architectural Computing 18, 335–352.
Siu, M.F., Ekyalimpa, R., Lu, M., Abourizk, S., 2013. Applying Regression Analysis to Predict and Classify Construction Cycle Time 148, 148–162.
Vitasek, S., Zak, J., 2018. BIM for Cost Estimation.
