Semantic search engine (GDI) for the decommissioning of a nuclear reactor

Industrial data science Nuclear France

Context and challenges

This nuclear reactor is located on a major nuclear site in France. This reactor was commissioned in 1964, and then operated for about ten years. By decree issued on 17 April 1980, the operator of this site was authorised to re-use this installation to store radioactive substances.

This installation, which does not comply with the current safety requirements for storage facilities, has not received any radioactive substances for storage purposes since 2008. The permanent shutdown of the facility is planned by the end of 2023. The decommissioning file must be sent to the ASN (French Nuclear Safety Authority) before 15 December 2019 in accordance with the Environment Code and the Nuclear Safety and Radiation Protection Mission letter (6 July 2018). The facility involves many documents necessary to organise the decommissioning, and the teams are facing difficulties using all of them.

In order to manage the decommissioning operations on this site and boost the efficiency of information management, this client tasked Assystem with implementing a “document to data” approach using the GDI solution (Global Data Inquirer: an Assystem Artificial Intelligence solution to read, understand, and structure the technical archives of complex installations).

Project scope

  • Data conversion and OCRisation (Optical Character Recognition)
  • Automatic generation of a database of the site inventory: buildings, equipment, waste, etc.
  • Setting up a smart query semantic search engine on all input data
  • Generating an ontology and searchable databases of the site inventory: extraction of tables and lists in the documents – Natural Language Processing, reading and understanding text by AI, (Deep Learning)
  • Setting up an “Installation Explorer” web application designed with several features: links between documents, extraction of the field of a document, etc.
  • Detection of duplicate documents and similar documents, and automatic extraction of tables from pdf files
  • Setting up a question & answer chatbot on the input data (AI DL QANet model)

Client benefits

  • Time savings thanks to the indexing and structuring of documents for fast and exhaustive access to information, enabling engineers and technicians to dedicate their time to design and production
  • Extraction of data to perform analyses and feed into simulation or decision-making tools
  • Major time savings for the analysis of decommissioning scenarios
  • The client has high confidence in the data thanks to its reliability and exhaustiveness
  • Decision-making support to decide if any additional studies are needed
  • Possibility of re-using this tool to other facilities in decommissioning phase (same topics, reusable structure)

In figures


GB of input data


input documents reduced to 14,280 documents


tables extracted


Questions/Answers created for the chatbot

Nuclear: related projects

optimisation de projets complexes nucléaire
Industrial data science Nuclear

Optimisation of complex schedules for nuclear unit shutdowns

Diesels d'ultime secours nucléaire.
Project management Nuclear

Assistance in Ultimate Diesel Generators test monitoring

Mise en conformité réglementaire
Compliance Nuclear

Regulatory compliance of anchorages on nuclear power plants

Digitalisation des essais et mise en service EPR - Digital test
Digital field services Nuclear

Digital test & commissioning for an EPR

Etude de sites et permis Assystem
Siting & Permitting Nuclear

Siting and permitting for a Nuclear Power Plant

contrôle-commande site nucléaire assystem
Industrial control systems Nuclear

Renovation of the ventilation instrumentation and control system

Jumeau numérique dans le secteur nucléaire
Digital field services Nuclear

Digital Twin of a nuclear waste treatment plant

Protection de site nucléaire SIS
Security Nuclear

Site protection and functional safety for a nuclear area


Contact us