Network Element Alarm Modeling for Fault Impact Prioritization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing telecommunications network management systems struggle to accurately differentiate between faults that affect end-user experience and those that do not, leading to inefficiencies in fault prioritization and maintenance.
Innovation Solution
A machine learning-based prediction model is trained using alarm data from network elements to predict operational status, distinguishing between faults that impact data throughput and those that do not, utilizing supervised learning and algorithms like artificial neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional fault monitoring methods are used to detect all network element faults, then comprehensive fault detection is achieved, but the ability to differentiate between faults affecting end-user experience and those that do not deteriorates
Solution Approach 1:
The patent introduces alarm data as an intermediary indicator to bridge the gap between raw network element status and end-user experience impact. By analyzing alarm data patterns and correlations, the system can infer whether a fault affects user experience without directly measuring user experience metrics, thus improving differentiation accuracy while managing system complexity
Solution Approach 2:
The patent replaces traditional manual or rule-based fault prioritization methods with a machine learning model. This substitution enables the system to automatically learn complex patterns in alarm data and operational status, achieving high differentiation accuracy without requiring complex manual monitoring configurations
2Measurement precision
If machine learning models are used to predict operational status from alarm data, then fault prioritization accuracy improves, but the complexity of the monitoring system increases
Solution Approach 1:
The patent segments the fault monitoring problem into distinct components: alarm data collection, data preprocessing, model training, and prediction execution. This segmentation allows each component to be optimized independently, improving overall prediction accuracy while managing system complexity through modular architecture
Solution Approach 2:
The machine learning model is trained on historical alarm data and operational status to enable self-service prediction capabilities. Once trained, the model automatically predicts operational status without requiring continuous human intervention or complex real-time analysis, maintaining high accuracy while reducing operational complexity
3Productivity
If all network element faults are treated with equal priority, then comprehensive maintenance coverage is achieved, but maintenance efficiency and response time to critical issues deteriorates
Solution Approach 1:
The patent applies local quality by assigning different priority levels to different faults based on their predicted impact on operational status and end-user experience. Critical faults that affect user experience receive higher priority and faster response, while non-critical faults are handled with standard procedures, thus improving maintenance efficiency without compromising reliability for critical issues
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
A computer-implemented method of training a prediction model for determining an operational status of a telecommunications network element is provided. The method includes receiving data representative of a plurality of training operational statuses of a corresponding plurality of network elements, where each network element has output one or more alarms, and where, for each network element, the data representative of the training operational status of that network element includes an indication of the one or more alarms generated by that network element, such that each training operational status is associated with the one or more alarms generated by that network element. The method further includes forming, based on the received data, training data includes a plurality of training data records corresponding to the plurality of network elements, where each training data record includes information representative of: the one or more alarms output by the network element associated with that training data record, and the training operational status of the network element associated with that training data record. The computer-implemented method also includes training, using the training data, the prediction model for determining the operational status of a network element.