AI Interaction with a Topological BIM

Enabling LLMs to interact with, extract from, and reason on Topological Building Information Models (TBIM) to enhance energy performance analysis processes.

Initial situation

Ecolution works in the field of analysing building energy performance, for certification and optimization purposes. Extracting relevant data and calculating energy performance metrics, like heat loss, from a building model is a tedious and time consuming process.

Project goal

This project aimed to develop tools that give LLM agents the power to autonomously do the task of extracting the relevant entities and their details from a building model, and perform energy metric calculations based upon them. These tools and agents should help ecolution streamline their current processes, and were created to be extendible for future needs and features.

Solution developed and its benefits

The final product ended up being a MCP server, that housed multiple workflows an agent can use:

• Automated entity filtering and retrieval
• Detail retrieval
• Heat loss and building performance calculations

This allows any LLM client our customer would want to use, to seamlessly integrate with the MCP server and leverage its capabilities. Allowing ecolution to expedite their workflows and improve overall efficiency, by delegating autonomous agents to do the heavy lifting.

Via our tools, and claude 4 sonnet, we were able to achieve an answer accuracy of 92% on retrieval and 87% on transmission heat loss calculations

Key terms

• TBIMTopological Building Information Model
• LLMLarge Language Model
• AgentAn autonomous LLM that can perform tasks on behalf of a user.
• Tool useCode extensions to LLMs
• MCPModel Context Protocol

Customer

Ecolution Engineering
Kevin Imhoff
Sonnenbergstrasse 53
5408 Ennetbaden
ecolution.swiss

Team