en
Digital solutions
BIM, digital twins, data, AI: practical solutions to support projects
At setec tpi, innovation is first and foremost at the service of our projects. Initiatives led by our operational teams are nurtured within a structured and managed framework. They all aim to push the boundaries of our expertise or improve our operational efficiency.
Stwin: data structured around the model
Bridges, as key infrastructure assets with a lifespan that can exceed a century, require ongoing maintenance to ensure their safety and performance. To support managers in this task, setec tpi has developed Stwin, a web platform that centralises all data relating to a structure: original drawings, calculation notes and maintenance histories.
Navigation is based on a 3D model, allowing information to be aggregated and visualised intuitively. Each element of the structure provides access to its key data: inspection reports, technical documents and historical records. This approach facilitates understanding, decision-making and communication between stakeholders.
A structured document repository complements the platform, with a configurable classification system (notably in accordance with ITSEOA) ensuring the long-term preservation and integrity of the data.
Finally, Stwin incorporates detailed access rights management, enabling information to be shared securely and tailored to the needs of the various stakeholders.
Chat-Decoda: AI for computation and modelling
Chat-Decoda was born from a simple idea: recent language models excel at generating code (Python, C++). Why not apply them to DECODA, the data input language used by Pythagore and Armatec?
Chat-Decoda thus offers Pythagore users a modelling assistant that allows them to consult documentation, comment on, modify or generate finite element models.
The approach relies on providing context to the AI model, equipping it with the resources needed to produce relevant DECODA code. This knowledge is drawn both from the model’s initial training and from internal databases: up-to-date user documentation, a library of representative models, and a dynamic database of errors and best practices.
Integrated into VS Code Copilot, Chat-Decoda enables direct interaction with files and tools. The document database, which is vectorised, is accessible via dedicated tools that allow the model to utilise this information effectively.
Pythagiec: designing optimised, low-carbon structures
As part of an effort to reduce the environmental footprint of structures, Setec TPI is developing optimisation methods aimed at reducing structural carbon emissions. Several approaches have been explored.
Parametric optimisation was first applied to high-rise buildings, generating numerous design variants to identify optimal solutions based on a cost-carbon trade-off. However, this method is only used at an advanced stage of the design process.
Lagrangian optimisation was then used to improve the design of reinforced concrete elements by incorporating construction constraints, with a limited number of iterations.
Generative methods have also been developed to optimise the shape of regular structures, by rapidly exploring numerous variants without resorting to computationally intensive calculations.
Finally, recent work has drawn on artificial intelligence to predict the behaviour of structures and simultaneously optimise their geometry and dimensioning, notably through neural network models.
Business web applications: simplifying the deployment and use of tools
The project aims to facilitate the deployment, sharing and maintenance of business tools within operational teams. It is structured around several complementary developments:
- a web portal for distributing applications, both internally and externally, incorporating authentication, user management and data backup;
- an in-house developed wrapper, based on a Python library, to standardise and accelerate application development;
- the development of the business web applications themselves, tailored to the teams’ needs.