Alarm Data Root Cause Detection Using Neural Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Telecommunication networks face challenges in isolating the root cause of faults from large volumes of alarm data, as alarm data often indicates issues in linked components rather than the faulty component itself, leading to delayed problem resolution.
Innovation Solution
A method and system utilizing a generator unit and discriminator unit within a recurrent neural network (RNN) to iteratively modify and refine alarm data, with a policy unit weighting feedback to distinguish between true and false alarms, enabling the detection of patterns and root causes in alarm data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional alarm data processing methods are used to monitor network elements, then alarm data can be collected and displayed, but the root cause of faults cannot be effectively isolated from large volumes of alarm data
Solution Approach 1:
The patent implements feedback loops where alarm data is continuously analyzed, processed through neural networks, and the results are fed back to improve future alarm data processing. The system learns from historical alarm patterns and continuously refines its ability to identify root causes, creating a self-improving diagnostic system that overcomes the limitation of traditional static analysis methods.
Solution Approach 2:
The patent introduces an intermediary processing layer consisting of neural networks and data correlation modules that sit between the raw alarm data and the final fault diagnosis. This intermediary layer correlates alarm data across multiple network elements, identifies patterns, and extracts meaningful information, thereby preventing loss of root cause information that would occur with direct traditional analysis methods.
2Reliability
If alarm data from all network elements is collected and analyzed in real-time, then comprehensive monitoring is achieved, but the complexity of processing and analyzing the data increases significantly
Solution Approach 1:
The patent segments the complex alarm data processing task into multiple specialized modules: data collection modules for different network elements, correlation modules for pattern recognition, neural network modules for advanced analysis, and visualization modules. This segmentation allows each component to handle specific aspects of the data, reducing overall system complexity while maintaining comprehensive monitoring coverage.
Solution Approach 2:
The patent creates a universal processing framework that handles multiple types of network elements (base stations, transmission links, terminals) and multiple alarm data formats through a single integrated system. The neural networks and correlation algorithms are designed to be element-agnostic, processing alarm data from any source through the same pipeline, thereby reducing complexity compared to having separate specialized systems for each network element type.
3Quantity of substance
If traditional alarm correlation techniques are used to reduce alarm data volume, then some redundancy is eliminated, but the techniques cannot effectively distinguish between true alarms and false alarms
Solution Approach 1:
The patent replaces traditional mechanical rule-based alarm correlation techniques with neural network-based intelligent analysis. Instead of using predefined correlation rules and thresholds, the system employs machine learning models that automatically learn patterns in alarm data, enabling more accurate distinction between true alarms and false alarms while maintaining effective data volume reduction.
Solution Approach 2:
The patent dynamically adjusts analysis parameters and thresholds based on learned patterns from historical data rather than using fixed parameters. The neural networks adapt their sensitivity and correlation criteria based on the specific context and type of alarm, allowing the system to maintain high accuracy in distinguishing true from false alarms across different scenarios while still reducing data volume effectively.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method, comprising receiving, by a generator unit (202), real alarm data (201) comprising alarm events and network parameters, from a network element; generating, by the generator unit (202), modified alarm data (203) by selecting and modifying at least one network parameter of the real alarm data (201); training the generator unit (202) by iterating through the following steps: evaluating, by a discriminator unit (204), the modified alarm data (203) against the real alarm data (201); based on the evaluation results, generating, by the discriminator unit (204), feedback to be provided to the generator unit (202); and generating, by the generator unit (202) based on the feedback new modified alarm data; repeating the iterative steps using the new modified alarm data as modified alarm data (203), wherein the modified alarm data (203) approaches the real alarm data (201) at the discriminator.