AI and computer vision-driven structural design automation
Project Description
Structural engineering design is an integral part of the AEC industry. Traditional structural design is done manually, where engineers need to read complex engineering drawings to identify critical structural information, perform quantity surveying, and check design compliance. The process is highly time-consuming and labour-intensive. In recent years, advanced AI and computer vision technologies have been developed rapidly, which have shown great performance in image and natural language processing. Currently, there is a growing trend in the use of computer vision algorithms for structural design automation. In this project, we aim to develop advanced AI-driven computer vision algorithms to process complex engineering drawings and automate the design tasks for buildings and bridges.
Supervisor
PAN, Xiao
Quota
5
Course type
UROP1100
UROP2100
UROP3100
UROP3200
Applicant's Roles
The students will be engaged in the following tasks:
a) Data curation, augmentation, and synthesis from existing engineering drawing datasets.
b) Develop computer vision algorithms for automated processing of graphical objects, semantic labels, tables, etc., for beams, columns, walls, steel reinforcement, etc., within the engineering drawings.
c) Develop numerical models to simulate the structural response of buildings and bridges.
d) Develop an agentic AI-driven drawing-to-analysis workflow in representative downstream tasks such as quantity surveying, numerical modelling, and design checks, etc.
e) Contribute to academic writing towards the publication of conference or journal papers.
a) Data curation, augmentation, and synthesis from existing engineering drawing datasets.
b) Develop computer vision algorithms for automated processing of graphical objects, semantic labels, tables, etc., for beams, columns, walls, steel reinforcement, etc., within the engineering drawings.
c) Develop numerical models to simulate the structural response of buildings and bridges.
d) Develop an agentic AI-driven drawing-to-analysis workflow in representative downstream tasks such as quantity surveying, numerical modelling, and design checks, etc.
e) Contribute to academic writing towards the publication of conference or journal papers.
Applicant's Learning Objectives
Students are expected to learn and complete at least one of the following tasks:
a) Use of Autodesk software (e.g., AutoCAD, Revit, etc) or other related programs to process engineering drawings.
b) Use Python programming with multiple platforms such as OpenCV, PyTorch, Hugging Face, etc., to develop AI-driven computer vision algorithms for automated engineering drawing analysis.
c) Use numerical simulation software (e.g., OpenSees and Abaqus) to develop finite element models for buildings and bridges.
d) Build a basic Agentic AI pipeline to automate one of the structural design tasks.
e) Write technical reports based on the results obtained.
a) Use of Autodesk software (e.g., AutoCAD, Revit, etc) or other related programs to process engineering drawings.
b) Use Python programming with multiple platforms such as OpenCV, PyTorch, Hugging Face, etc., to develop AI-driven computer vision algorithms for automated engineering drawing analysis.
c) Use numerical simulation software (e.g., OpenSees and Abaqus) to develop finite element models for buildings and bridges.
d) Build a basic Agentic AI pipeline to automate one of the structural design tasks.
e) Write technical reports based on the results obtained.
Complexity of the project
Moderate