Predictive Risk Signatures for Weak Signal Detection in Industrial Systems

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods struggle to effectively detect and characterize weak risk exposure signals in industrial systems, particularly in identifying precursors to significant events, due to challenges in distinguishing these signals from random noise and the limitations of existing symbolic and numerical approaches.

Innovation Solution

A method that calculates a predictive risk signature using data from sensors, incorporating parameters determined by artificial intelligence, combining symbolic and digital approaches to identify weak signals by comparing them to reference signatures and displaying associated threat scenarios.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If symbolic approach with models and rule-based reasoning is used, then detection can be performed for specific cases, but the approach is limited to specific cases and lacks generalizability

Engineering Contradiction:
Improvedetection accuracyVSAvoidapplicability scope
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent transforms the symbolic approach by changing parameters from fixed rule-based thresholds to dynamic parameters learned through machine learning. The system learns optimal detection parameters from historical data, enabling the same model to adapt to different industrial scenarios while maintaining high detection accuracy for weak signals.

Inventive Principle:
Principle #35Parameter changes

2Loss of information

If numerical approach with artificial neural networks is used, then data mining capability is improved, but the system becomes complex and primarily enables a posteriori detection after events occur

Engineering Contradiction:
Improvedata mining capabilityVSAvoidsystem complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent implements preliminary action by training the neural network model beforehand on historical data containing patterns preceding significant events. This pre-training enables the system to perform a priori detection, identifying weak signals before events occur, rather than only analyzing data after events have happened.

Inventive Principle:
Principle #10Preliminary action

3Ease of manufacture

If traditional detection methods are used, then implementation is simpler, but the ability to distinguish weak signals from random noise is insufficient

Engineering Contradiction:
Improveimplementation simplicityVSAvoidsignal-to-noise discrimination
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent incorporates feedback mechanisms where the system continuously learns from detected events and adjusts its detection parameters. The model is refined using feedback from actual operational data, improving its ability to distinguish weak signals from noise while maintaining practical implementability through iterative optimization.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4198668B1Method and system for detection and characterization of low risk exposure signals in an industrial system
Publication Date: 2024.09.04 COMMISSARIAT A LENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVES
  • EP4198668B1 patent drawingFigure 1
  • EP4198668B1 patent drawingFigure 2
  • EP4198668B1 patent drawingFigure 3

AI summary

The invention relates to a method and a system for detecting and characterizing weak signals of exposure to risk in an industrial system, from data relating to the industrial system collected over a given time period.This system is configured to implement: - a module (36) for calculating a predictive risk signature, from data relating to the industrial system collected, using a first term obtained by summing elementary signatures associated with elementary initiating events, dependent on parameters including a severity value, a characteristic function and a weighting function of the elementary initiating event, at least part of said parameters being determined by implementing a neural network, - a module (38) for detecting the presence of a weak signal of exposure to a risk by comparing the calculated predictive risk signature to predetermined reference risk signatures.