Network Failure Localization via Spectrogram Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Network service providers face challenges in rapidly identifying and localizing equipment failures in network devices due to unreliable alarm data and noisy signals, which can lead to service outages and delayed restoration of services.
Innovation Solution
The technology processes status messages from network devices using a particle filter to detect anomalies and localize failures by transforming message counts into spectrograms, comparing them to normal operating conditions, and ranking devices based on similarity scores to determine the most likely cause of the failure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If connectionless communication methods are used to send status messages, then device complexity is reduced, but reliability of failure detection deteriorates
Solution Approach 1:
The system implements feedback by collecting status messages from multiple network devices and using particle filtering to continuously update the probability distribution of device states. This feedback loop allows the system to distinguish between temporary noise and actual failures by analyzing patterns in the feedback signals from multiple sources over time.
Solution Approach 2:
The patent introduces an intermediary analysis system that processes status messages between the network devices and the service provider. This intermediary layer applies particle filtering and spectrogram analysis to interpret the noisy connectionless messages, transforming unreliable individual signals into reliable failure detection without changing the simple communication protocol.
2Ease of operation
If traditional alarm-based methods are used to detect failures, then ease of operation is improved, but measurement precision deteriorates due to noisy and unreliable alarm signals
Solution Approach 1:
Instead of relying on a single alarm signal, the system collects excessive status messages from multiple devices and time points, then uses particle filtering to extract the relevant failure information. This partial action approach processes more data than traditionally needed, using statistical methods to isolate the precise failure signal from the noise.
Solution Approach 2:
The patent transforms the failure detection problem from analyzing individual alarm messages into a multi-dimensional spectrogram analysis. By converting status message patterns into frequency-domain representations, the system adds temporal and spectral dimensions to the analysis, enabling precise failure detection while maintaining ease of automated operation.
3Measurement precision
If comprehensive status message collection from all devices is performed, then measurement precision of failure localization is improved, but loss of time increases due to processing large data volumes
Solution Approach 1:
The system performs preliminary action by pre-computing and storing the relationship between device status patterns and potential failures. When a failure occurs, the particle filter uses these pre-established patterns to rapidly identify the failed device, avoiding the need to process all raw status messages from scratch and significantly reducing localization time.
Solution Approach 2:
The patent transforms the status message data into different parameter representations, specifically converting time-domain message counts into frequency-domain spectrograms. This parameter change enables more efficient processing by identifying failure patterns in the frequency domain, which can be analyzed faster than raw time-series data while maintaining high localization precision.
Data Source
AI summary
The disclosed technology is directed towards automatically detecting failure states and the cause of the failure. For a network, the technology collects status messages from equipment and customers into batches as they occur. The technology groups and aggregates messages, then transforms the aggregations to the frequency domain. Anomalies induce detectable changes in the particle distribution of a trained particle filter, from which an anomalous spectrogram is generated. The status messages of each device are iteratively removed from the larger set of messages, resulting in reduced subsets that are each aggregated, transformed into a modified spectrogram and compared against the anomalous spectrogram to obtain a distance score. The distance score for each device is used to rank the devices with respect to being the cause of the failure.