AI Applications in the Workplace
1. Building BIM Models in Infrastructure Design
1.1. Problem Statement and Context
In previous civil engineering projects, 3D models were primarily used for design review before 2D drawings were developed. Conversely, bridge and road projects prioritize 2D drawings first, with 3D models serving only illustrative purposes.
These 3D models are not yet true BIM models because BIM requires:
- The 3D model must contain information.
- The 3D model and 2D drawings must be exported from the same synchronized data source.
Many current projects (such as the Dong Da Interchange and National Highway 26B) have 3D models that are recreated from finalized 2D drawings, causing a lack of consistency and failing to leverage the core value of BIM.
1.2. Optimizing BIM Models: From Information Assignment to Automation
In road design using Civil 3D, objects are typically 3D solids assigned property sets. However, default information is often missing (what component, what material, geometric parameters, etc.). Manually assigning these properties is time-consuming and prone to errors.
To make the most of BIM, it’s necessary to classify and automatically assign three main types of information:
- Type 1 – Shared information: Project name, design unit, contract code. Easy to update in bulk.
- Type 2 – Specific information about the object: Land plot number, owner, type of construction. Requires customization tools.
- Type 3 – Geometric calculation information: Area, volume, height, coordinates. Automated calculation tools are needed to reduce errors.
Overview of BIM in infrastructure design
Categorizing Property Sets
2. BIM Automation with AI Agents
2.1. The process of writing new programming code
The emergence of AI agents has completely changed the process of developing automation tools:
Traditional (6 steps)
With AI (4 steps)
With AI Agent (3 steps)
2.2. Analysis of the effectiveness of the AI Agent (Github Copilot)
Outstanding advantages
- Automatically generate code upon request.
- Perform effective self-checking and error correction (only stop when all errors are resolved).
- Multiple requests can be run in parallel.
- High success rate.
Limitations & Challenges
- High cost for premium models.
- Monthly resource limit.
- It’s difficult to control the code without programming knowledge.
- Effectiveness depends on the ability to issue commands (Prompt Engineering).
2.3. Developing the “Vibe Coding” platform
The research team at T27 is developing sample programming projects to make them easily accessible to employees:
> Install Visual Studio + Github Copilot
> Activate AI Agent
> Start “Vibe Coding” – Create tools
Combining Visual Studio and AI Agent not only supports programming for Civil 3D, Revit, and Excel, but can also be used for website design or game creation.
3. Practical Applications That Have Been Implemented
🛠️ Automatic Information Assignment Tool
Quickly assigning information to the Property Sets of a 3D solid object, such as component name, material type, area, etc., helps standardize input data for the BIM model.
📍 Tool for Exporting Coordinates of Prestressed Cables
Automatically export prestressed cable coordinates from plan and longitudinal section cable layout drawings, minimizing errors caused by manual data entry.
🔌 API for Civil 3D & Revit
- Civil 3D: Create infrastructure design automation tools similar to Visual Infrastructure.
- Revit: Developing utilities for bridge design, optimizing modeling operations.
- Dynamo: Create custom nodes for visual programming.
4. Conclusion
Building BIM models with the help of AI Agents not only increases efficiency and improves the accuracy of design documents but also opens up potential opportunities for developing “Make in T27” utilities .
This is a significant step forward for T27 to apply AI more extensively in the future, affirming its pioneering position in the digital transformation of the construction industry.
The future of AI applications in infrastructure design at T27