Root Cause Estimation via Alarm Grouping and Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for troubleshooting network system failures require learning correlations between facilities in advance, making it impossible to identify the root cause of alarms generated from multiple devices without prior knowledge.
Innovation Solution
An information processing apparatus that receives and processes output signals from multiple devices to generate a ranking model by replacing selection results with combinations of output signals, allowing for the estimation of root causes without pre-learning correlations between devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If correlations of facilities are learned in advance using neural networks, then the accuracy of root cause identification is improved, but the system complexity and preprocessing requirements increase
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing the relationship between alarm groups and facilities in a lookup table during system setup. This allows the root cause identification to proceed without complex real-time neural network computations, thereby improving response speed while maintaining acceptable accuracy through the pre-established alarm-group-to-facility mappings.
Solution Approach 2:
The patent creates a simplified copy of the complex facility correlation structure by representing it as alarm groups with associated facility lists in a database. Instead of using the full complexity of neural network weights and activations, the system uses a tabular representation that captures the essential relationships, reducing computational complexity while preserving the core functionality of root cause identification.
2Reliability
If a large number of alarms are generated from multiple facilities, then comprehensive monitoring coverage is improved, but the difficulty of identifying the root cause alarm increases
Solution Approach 1:
The patent segments the large set of alarms into distinct alarm groups, where each group contains alarms that are correlated and likely share a common root cause. This segmentation transforms the overwhelming task of analyzing all alarms individually into a manageable process of identifying representative alarms from each group, thereby reducing detection difficulty while maintaining comprehensive monitoring coverage.
Solution Approach 2:
The patent introduces alarm groups as an intermediary layer between individual alarms and root cause identification. Instead of directly analyzing individual alarms to find root causes, the system first groups correlated alarms together, then selects representative alarms from each group. This intermediary structure simplifies the detection process by reducing the number of alarms that need to be analyzed in detail.
3Measurement precision
If correlations between facilities are learned in advance, then the model accuracy is improved, but the ability to handle new or unseen alarm patterns deteriorates
Solution Approach 1:
The patent implements dynamics by enabling the system to adapt to new alarm patterns through the selection of representative alarms from newly formed alarm groups. When new facilities or alarm types are introduced, they automatically form new groups or join existing groups, and the representative alarm selection process adapts to these changes without requiring retraining of complex models, thus maintaining both accuracy and adaptability.
Data Source
AI summary
An information processing apparatus includes: a reception module configured to receive selection of an output signal corresponding to a root cause from a group including a plurality of output signals, the plurality of output signals being obtained from a plurality of devices to be managed and being obtained in a predetermined period; a replacement module configured to replace a selection result received by the reception module with a combination of the selection results in two output signals, the selection result being corresponding to the group, the two output signals being selected from the plurality of output signals; and a model generation module configured to generate a ranking model of the plurality of output signals based on the selection results after the replacement.