Failure Estimation Support Device Using Correlation Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of IP interconnection between common carriers complicates failure isolation and analysis of SIP server abnormalities, increasing operational loads on maintenance personnel due to the variety of possible failures.
Innovation Solution
A failure estimation support device and method that calculates single and multiple failure occurrence probabilities, generates correlation rules based on these probabilities, and uses them as training data for machine learning to estimate failure causes, reducing the operational burden on maintenance personnel.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual failure analysis is performed by maintenance personnel, then failure cause estimation can be done, but operational load on maintenance personnel increases due to complexity of IP interconnection
Solution Approach 1:
The system enables automatic failure cause estimation through machine learning models that analyze alarm information and network configuration data independently, without requiring manual intervention from maintenance personnel. The learned model automatically processes failure data and generates cause estimates, making the system self-sufficient in handling failure analysis tasks.
Solution Approach 2:
The manual mechanical process of failure analysis by maintenance personnel is replaced with an automated information processing system using machine learning. The system substitutes human analytical work with computational algorithms that process alarm logs, network configurations, and failure data to automatically estimate failure causes.
2Measurement precision
If multiple single rules are combined for failure analysis, then failure isolation accuracy improves, but device complexity increases due to multiple correlation rules
Solution Approach 1:
The system performs preliminary actions by pre-calculating single failure occurrence probabilities for individual rules and pre-identifying correlated rule combinations before actual failure analysis is needed. These pre-computed probabilities and correlations are stored and ready for rapid retrieval during failure events, avoiding complex real-time calculations.
Solution Approach 2:
The system transforms complex multi-rule correlation analysis into simplified probability-based assessments. By converting rule correlations into numerical probability values and applying threshold filtering, the system maintains high failure isolation accuracy while reducing the complexity of rule evaluation and combination management.
3Measurement precision
If machine learning is performed with correlation rules, then failure estimation accuracy improves, but calculation time increases due to multiple probability calculations
Solution Approach 1:
The system performs preliminary calculations of single failure occurrence probabilities for all individual rules during normal operation or off-peak periods. These pre-computed probabilities are stored and reused during failure analysis, avoiding redundant calculations and significantly reducing the time required for real-time failure estimation while maintaining high accuracy.
Data Source
AI summary
A failure estimation support device 2 includes: a first calculation unit 21 configured to calculate single failure occurrence probabilities of individual single rules for each single rule relating to a single failure from past failure occurrence data; a second calculation unit 22 configured to calculate multiple failure occurrence probabilities that multiple failures occur in association for each combination of multiple single rules using the single failure occurrence probabilities; and a correlation rule generation unit 23 configured to select the combination of the multiple failure occurrence probabilities greater than a predetermined threshold among the multiple failure occurrence probabilities as a correlation rule, and the correlation rule is used as training data for making a learned model which estimates a failure factor perform machine learning together with the single rule.