Sensor Causal Learning for Deterministic Anomaly Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for determining causal relationships in industrial settings, such as factories, face challenges in accuracy due to reliance on human experience and struggle with large-scale systems, especially when deterministic factors and nonlinear relationships are involved, and are computationally intensive.

Innovation Solution

A system that includes a causal relationship learning device with a determination unit to assess correlations between sensor measurements and an estimation unit to identify causal relationships, using Convergent Cross Mapping (CCM) for low correlation pairs and bidirectional polynomial models for high correlation pairs, to accurately determine cause-effect relationships between sensors.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If causal relationship is set based on human experience, then automation is improved, but accuracy deteriorates due to incorrect causal relationships and difficulty in large-scale factories

Engineering Contradiction:

Improveautomation of causal relationship identificationVSAvoidaccuracy of causal relationship

Core Design Contradiction:

Extent of automationVSMeasurement precision

Solution Approach 1:

The patent replaces the mechanical system of human experience-based causal relationship setting with a computational system using Granger causality and motion entropy calculations. This substitution enables automated identification of causal relationships through mathematical analysis of sensor data, eliminating reliance on human expertise while maintaining or improving accuracy through objective computational methods.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the parameters used for causal relationship determination from subjective human judgment to objective quantitative parameters including Granger causality metrics and motion entropy values. By transforming the problem into parameter-based computational analysis, the system achieves both automation and improved accuracy in identifying causal relationships among sensors.

Inventive Principle:

Principle #35Parameter changes

2Extent of automation

If Granger causality method is used, then automation is improved, but accuracy deteriorates for deterministic systems and nonlinear relationships

Engineering Contradiction:

Improveautomation of causal relationship estimationVSAvoidaccuracy in deterministic systems

Core Design Contradiction:

Extent of automationVSMeasurement precision

Solution Approach 1:

The patent segments the causal relationship estimation process into two distinct approaches: Granger causality for probabilistic/linear relationships and motion entropy for deterministic/nonlinear relationships. By dividing the problem space and applying appropriate methods to each segment, the system achieves both automation and high accuracy across different system types without the limitations of a single method.

Inventive Principle:

Principle #1Segmentation

3Measurement precision

If motion entropy method is used, then accuracy is improved for deterministic systems, but device complexity increases due to large calculation requirements

Engineering Contradiction:

Improveaccuracy in deterministic systemsVSAvoidcalculation complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies local quality by selecting different causal estimation methods based on the local characteristics of the system being analyzed. Motion entropy with higher computational complexity is applied locally to deterministic systems where it provides superior accuracy, while Granger causality with lower computational complexity is applied to probabilistic systems. This localized application optimizes both accuracy and computational efficiency.

Inventive Principle:

Principle #3Local quality

4Adaptability or versatility

If operator visually references sensor values, then flexibility is maintained, but productivity deteriorates due to time consumption and human error

Engineering Contradiction:

Improveoperator flexibility in anomaly analysisVSAvoidspeed of anomaly cause identification

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent implements self-service by enabling the system to automatically identify anomaly causes through computational analysis of sensor data and causal relationships. Instead of relying on operators to manually analyze sensor values, the system performs the analysis itself using automated causal relationship identification, thereby eliminating human error and dramatically improving productivity while maintaining the flexibility to handle various anomaly scenarios.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS11073825B2Causal relationship learning method, program, device, and anomaly analysis system

Publication Date: 2021.07.27 NEC CORP

AI summary

The present invention provides a method, a program, and a device that can accurately learn a causal relationship in a deterministic system and a system that performs anomaly analysis by using the causal relationship. A causal relationship learning device according to one example embodiment of the present invention includes: a correlation determination unit that determines a correlation between measurement values measured by two sensors; and a low correlation causal relationship estimation unit that, when the correlation is lower than a predetermined reference, determines a causal relationship between the two sensors by estimating one of the measurement values which is a cause from the other of the measurement values which is a result.