Case Study

GenAI for data extraction from technical drawings

We realized a solution using generative AI for the automatic acquisition of information from technical drawings or modification documents of a plant.

Where: Italy
Challenge
The customer, a company in the energy sector that operates a gas transmission network, asked for our support in automating the acquisition of relevant data from the various technical drawings provided by the designers of regulation and measurement plants. With this solution, the customer aimed to streamline the validation procedure for plant drawings and modifications documents.
Approach
We used a step-by-step and iterative approach, starting with a limited set of plants and then extending the scope of the solution to the entire fleet of meters. To enable the extraction of relevant information from different technical plant drawings, we combined various algorithms of Generative AI, Machine Learning and Image Analysis.
Solution
We have realized a system that, by integrating customized models and enhancing GPT4Vision models, is able to analyze a technical drawing of a plant and extrapolate its characteristics by entering them into a relational DB. Given a .pdf file containing plant documents as input, the solution extracts the drawing data and the list of components using Machine Learning algorithms, validating them through GenAI models. Using an advanced customized Image Analysis model, the solution automatically recognizes the symbols in technical drawings and indicates whether they are connected. From the detected symbols and relationships, it then uses a probabilistic system to reconstruct the features and lines that make up the individual control and measurement plant.
Results
Technologies

To know more

STARLIGHT: Artificial Intelligence to fight against high priority threats

Better data analysis and management: a project to equip law enforcement agencies with secure and resilient AI technologies and tools.

Use Case

Enable Traceability for AgriFood

How to use digital technologies to improve food chain traceability

Easy and Invisible Payment Experience

How to use “invisible payments” to make the shopping experience easier