AI Anomaly Detection for Nuclear Reactor Brick Arrays

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for inspecting nuclear reactor components are labor-intensive, risky for human operators, and inefficient, with manual data interpretation and lack of predictive capabilities for future defect development.

Innovation Solution

A computer-implemented method using artificial intelligence to automatically detect defects in nuclear reactor components from collected picture data, and predict future system states based on identified defects, integrating classification models and neural network algorithms with physical modeling.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual inspection methods are used by human operators, then detailed visual assessment can be performed, but operator safety is compromised and inspection time increases

Engineering Contradiction:
Improvevisual assessment accuracyVSAvoidoperator safety
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent uses Remote Visual Inspection (RVI) cameras to capture images of brick arrays, creating visual copies that can be analyzed without human operators physically present. The system processes these images through AI algorithms to detect anomalies, thereby maintaining inspection accuracy while eliminating operator exposure to hazardous radiation environments

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent replaces manual mechanical inspection processes with an automated system combining RVI cameras, image processing algorithms, and AI anomaly detection. This substitution eliminates the need for human operators to physically access hazardous areas while maintaining or improving detection capabilities through computational analysis

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If frequent inspections are conducted to prevent defects, then system reliability improves, but productivity decreases due to regular shutdowns

Engineering Contradiction:
Improvesystem safetyVSAvoidpower plant output
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent implements AI-based anomaly detection that identifies potential defects in brick arrays before they develop into critical failures. By detecting early signs of degradation patterns, the system enables predictive maintenance scheduling that prevents catastrophic failures without requiring frequent preventive shutdowns, thus maintaining both safety and productivity

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system continuously monitors brick array conditions through RVI and AI analysis, providing feedback on degradation trends. This ongoing assessment allows operators to schedule maintenance based on actual condition rather than fixed intervals, optimizing the balance between reliability and productivity

Inventive Principle:
Principle #23Feedback

3Measurement precision

If Finite Element Methods are used for defect prediction, then physical accuracy improves, but computational time increases significantly

Engineering Contradiction:
Improvedefect prediction accuracyVSAvoidcomputational time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent employs lightweight machine learning models trained on historical RVI data that can quickly assess current brick array conditions without requiring intensive computational resources. These models provide sufficient accuracy for operational decision-making while executing in real-time or near-real-time, unlike heavy FEM simulations

Inventive Principle:
Principle #27Cheap short-living objects (Disposable)

Solution Approach 2:

The system uses AI algorithms that analyze only the most relevant features from RVI images (such as crack patterns, displacement, and anomaly signatures) rather than performing complete physical simulations. This selective analysis provides adequate prediction accuracy for maintenance scheduling while significantly reducing computational burden

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4567722A1Artificial intelligence method for extracting anomalies and forecasting a future state of a nuclear system
Publication Date: 2025.06.11 ELECTRICITE DE FRANCE
  • EP4567722A1 patent drawingFigure 1
  • EP4567722A1 patent drawingFigure 2
  • EP4567722A1 patent drawingFigure 3A~3D

AI summary

The invention relates to a computer implemented method for monitoring a state of a system in a nuclear site, at least parts of the system being pictured by at least one camera, the method comprising: - collecting (S1) pictures of at least one part of the system, taken by at least one camera over time, - implementing (S2) a first artificial intelligence to detect at least one defect in said pictures, - implementing (S3) a second artificial intelligence to provide a prediction of a future state of said part of the system, from the at least one defect and according to a current operation mode of the system, - operating (S4) the system according to an adapted operational procedure, depending on said prediction.