Image: Research Overview (Image visual rendering was optimised using the GPT Image-2 model) 

starting point:

How are practice data (BIM, GIS, Lidar models, etc.) structured to enable knowledge accessibility, interoperability, and security?  How are errors, risks, and file sharing handled?

This topic may include consideration of: vocabulary and semantics/ontology across architectural sectors and how it is covered by standards; the impacts of interfaces and communication methods needed to feed analysed practice data back to practice.

project summary: 

This study addresses the difficulty architects face when selecting building materials under multiple criteria across iterative design stages, in a context where material data is fragmented, and searched mainly by keywords. Existing multi criteria decision making frameworks and online material libraries formalise factors such as performance, cost, and sustainability, yet they rarely capture how designers actually interpret ambiguous scenario, balance trade-offs over time, or reuse prior decisions, because they overlook the implicit and tacit knowledge in material selection. Current LLM based BIM model query prototypes remain tightly coupled within prompt engineering to translate user’s natural language query into SQL for material databases, with little integration of external knowledge or transparent memory mechanisms. Which raised two questions for this study:   

• What knowledge taxonomy best characterises knowledge of architects used in building material selection across design stages? 

• How can a knowledge-organisation and retrieval framework be designed to support iterative building material selection and cross-project knowledge reuse? 

In response, the research will develop an LLM powered material selection system based on the integration of relational material database and unstructured material design code documents, combines a modular retrieval and query translation layer that maps multi criteria natural language requirements. Building on an empirically grounded knowledge taxonomy, the system separates explicit material facts, graph based contextual relations, and human tacit intuition, the first two are formalised while designers retain control over final decisions.  

The overall aim is to support multi criteria, iterative and transparent material selection in early design, enabling non-technical designers to search, compare and justify suitable materials more efficiently across projects, while improving the reuse of past project experience and contributing to more digitally integrated practice in the AEC sector. 

Reference 

1. Bajwa, A. U. R., Siriwardana, C., Shahzad, W., & Naeem, M. A. (2025). Material selection in the construction industry: A systematic literature review on multi-criteria decision making. Environment Systems and Decisions, 45(1), 8. https://doi.org/10.1007/s10669-025-10001-w 

2. Brunsmann, J., & Wolfgang, W. (2009). Enabling Product Design Reuse by Long-term Preservation of Engineering Knowledge | International Journal of Digital Curation. https://ijdc.net/index.php/ijdc/article/view/114 

3. Gao, Y., Xiong, Y., Gao, X., Jia, K., Pan, J., Bi, Y., Dai, Y., Sun, J., Wang, M., & Wang, H. (2024). Retrieval-augmented generation for large language models: A survey (arXiv:2312.10997). arXiv. https://doi.org/10.48550/arXiv.2312.10997 

4. Sanchez, R. (2004). ‘Tacit Knowledge’ versus ‘Explicit Knowledge’: Approaches to Knowledge Management Practice. https://research.cbs.dk/en/publications/tacit-knowledge-versus-explicit-knowledge-approaches-to-knowledge/ 

5. Yang, J., & Ogunkah, I. C. B. (2013). A multi-criteria decision support system for the selection of low-cost green building materials and components. Journal of Building Construction and Planning Research, 01(04), 89–130. https://doi.org/10.4236/jbcpr.2013.14013 

6. Zheng, C., Wong, S., Su, X., Tang, Y., Nawaz, A., & Kassem, M. (2025). Automating construction contract review using knowledge graph-enhanced large language models. Automation in Construction, 175, 106179. https://doi.org/10.1016/j.autcon.2025.106179 

7. Zheng, J., & Fischer, M. (2023). Dynamic prompt-based virtual assistant framework for BIM information search. Automation in Construction, 155, 105067. https://doi.org/10.1016/j.autcon.2023.105067 

8. Zhou, Y., Shi, Q., Luo, Z., Cai, X., Huang, Y., Kim, D. H., Weng, D., & Wu, Y. (2026). Cerebra: Aligning implicit knowledge in interactive SQL authoring. Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems, CHI ’26, 1–22. https://doi.org/10.1145/3772318.3790974