AI-based automation for structural design of buildings

Resumo

With the recent advancement of technology and computational empowerment, processes that were executed manually have the potential to be automated. In Civil Engineering, this is being seen through the implementation of technologies such as Building Information Modeling (BIM), Generative Design, Machine Learning and Artificial Intelligence. The objective of this work is to develop a computational system aiming to optimize the development of architectural and structural projects, implementing technologies such as artificial intelligence and generative design to automate and systematize processes inherent to the project pre-dimensioning phase when designing a building.

Initially, the user interacts with an AI-based chatbot interface which allows the definition of a buildings main characteristics, such as number of stories, apartments per story and minimum areas. The system has access to a database that contains different architectural BIM models and will filter them in order to provide the best option according to the user specifications. The selected model is processed by a Dynamo script in Autodesk Revit that automatically creates and allocates structural elements, e.g. beams, columns, slabs, and foundation. The script considers the intersection between walls and building corners, also positioning columns in wider spans avoiding doors and windows.

This structure is converted into an analytical model, so it can be exported to Autodesk Robot Structural Analysis where another Dynamo script is executed, which contains loads determined by technical standards as the Brazilian ABNT NBR 6118:2023 for concrete design. This script analyzes bending moments, shear effects and axial forces of the structural elements and, through limiting parameters, sends messages to Autodesk Revit, stating if a structural element needs to be modified to attend applied loads. This process runs continuously until all elements are adequate.

Using this technology, enterprises specialized in architectural and structural projects will save time, since allocating and pre-dimensioning structural elements are processes that are executed manually. Therefore, engineers are then able to focus on greater challenges involved in structural analysis and architects can simulate the dimensions of elements more accurately. There are still many processes in Autodesk Revit and Robot Structural Analysis that can be automated, and these will be further developed as a sequence of this study.