Correlation Model Selection for Time-Varying System Failure Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing operations management systems face challenges in generating accurate correlation models for system analysis due to the lack of information on modeling periods, leading to incorrect detection of system abnormalities, especially when system characteristics change over time.

Innovation Solution

An operations management apparatus and method that generate multiple correlation models for different time periods and select a basic model with the highest fitting degree, along with specific models, to apply them to specific time slots for failure detection, ensuring accurate analysis even without knowledge of system characteristics.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a single correlation model is generated using time series information from a predetermined period, then the model generation process is simple, but the model becomes inaccurate when system characteristics change over time, leading to incorrect failure detection

Engineering Contradiction:
Improveaccuracy of failure detectionVSAvoidcomplexity of model generation process
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the time series information into multiple segments based on different time periods (e.g., daytime, nighttime, weekdays, weekends). Instead of generating a single correlation model from the entire period, separate correlation models are generated for each time period segment. This allows the system to capture the varying system characteristics at different times while maintaining manageable model complexity through modular generation.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements dynamic model selection by automatically determining the current time period and selecting the appropriate pre-generated correlation model for failure detection. The system dynamically adapts to changing system characteristics by switching between different correlation models based on the current time context, rather than using a static single model for all situations.

Inventive Principle:
Principle #15Dynamics

2Reliability

If multiple correlation models are generated for different time periods, then the accuracy of failure detection improves, but the number of models to manage increases

Engineering Contradiction:
Improvereliability of system analysisVSAvoidinformation management overhead
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The patent incorporates automatic evaluation mechanisms that assess the accuracy and applicability of each correlation model based on actual system performance data. The system provides feedback on model performance and automatically selects the most appropriate model for the current situation, reducing the need for manual model management while maintaining high reliability in failure detection.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system performs self-service by automatically generating, evaluating, and selecting correlation models without requiring extensive manual intervention. The automated processes handle model generation from time series data, evaluate model fit to actual system behavior, and select appropriate models for failure detection, thereby reducing information management overhead despite having multiple models.

Inventive Principle:
Principle #25Self-service

3Measurement precision

If an administrator manually sets the modeling period based on system knowledge, then the correlation model can be accurate for that specific period, but the system becomes dependent on administrator availability and expertise

Engineering Contradiction:
Improveaccuracy of correlation modelVSAvoidease of model generation
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The patent enables the system to automatically determine appropriate time periods for generating correlation models by analyzing time series data and identifying patterns in system behavior. The system self-services by automatically segmenting time periods, generating relevant correlation models, and selecting appropriate models for failure detection without requiring administrator intervention or specialized knowledge about system operational patterns.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces the manual mechanical process of administrator-based model period selection with an automated computational system. The system uses algorithms to analyze time series data, identify meaningful time period segments, and generate correlation models automatically, substituting human expertise and manual operations with automated mechanical processes that are more consistent and scalable.

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

Data Source

PatentEP2613263B1Operations management device, operations management method, and program
Publication Date: 2020.01.08 NEC CORP
  • EP2613263B1 patent drawingFigure 1
  • EP2613263B1 patent drawingFigure 2
  • EP2613263B1 patent drawingFigure 3

AI summary

A correlation model which is appropriate for a system analysis can be generated with respect to each of fixed periods such as the dates on which the system analysis is performed, even if the information on modeling periods with respect to system characteristics is not available. A correlation model generation unit 102 generates a plurality of correlation models 122 each expressing correlations between different types of performance values in a predetermined period, which are stored in a performance information unit 111. A model setting unit 103 selects, from among the plurality of correlation models 122 generated by the correlation model generation unit 102, a basic model which is a correlation model 122 showing the highest fitting degree and one or more specific models which are correlation models 122 other than the basic model, on the basis of a fitting degree of each of the correlation models 122 for the performance information in the predetermined period, and sets time periods on which the basic model and the specific models are applied respectively to failure detection.