RRH Fault Prediction Using Operation Logs and ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In large-scale communication networks, existing methods for detecting faults in remote radio heads (RRHs) are inefficient, often resulting in false positives and negatives, leading to high costs and errors, and may invade user privacy by relying on data from user equipment (UEs).
Innovation Solution
A method and device that predict RRH faults based on operation information such as uptime, alarm type, and environmental data, independent of UE data, using a data analysis device with a processor and communication interface to issue notifications when a fault is predicted, employing models like artificial neural networks and support vector machines to analyze operation parameter logs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If technicians manually analyze alarms to detect faulty RRHs, then they can identify potential hardware faults, but the process is inefficient and results in high false positive and false negative rates
Solution Approach 1:
The patent replaces manual alarm analysis by technicians with an automated machine learning model that processes alarm data and operation parameters to predict RRH faults. The model uses algorithms to automatically detect patterns and predict failures, eliminating the need for human technicians to manually analyze alarms while significantly improving both detection accuracy and efficiency.
Solution Approach 2:
The patent introduces an intermediary prediction model that acts as a bridge between raw alarm data and fault detection decisions. This model processes alarm information, combines it with operation parameters, and generates predicted fault probabilities, serving as an intelligent intermediary that enhances both the precision and efficiency of fault detection.
2Reliability
If cell outage compensation is applied based on high-level KPIs, then network reliability is improved, but fundamental approaches to network security are neglected and existing nodes become saturated
Solution Approach 1:
The patent applies preliminary action by predicting RRH faults before they actually occur. The machine learning model analyzes current operation parameters and alarm data to forecast potential failures, enabling proactive maintenance and replacement decisions. This prevents the need for reactive cell outage compensation that saturates existing nodes, thereby improving reliability without overloading the network infrastructure.
3Loss of information
If UE data is used for fault detection, then more comprehensive information is available, but user privacy is invaded and security concerns arise
Solution Approach 1:
The patent extracts and uses only the necessary alarm data and operation parameters from RRHs for fault prediction, deliberately excluding UE data that would invade user privacy. The model is trained and operates using network-side information such as alarm logs, operation parameters, and environmental data, achieving effective fault detection without accessing sensitive user information.
Data Source
AI summary
Disclosed is a method for managing multiple remote radio heads (RRHs) in a communication network, the method including, in response to an alarm indicating that a value of an operation parameter of an RRH among the multiple RRHs is beyond a predetermined range, predicting whether the RRH is faulty, based on one or more pieces of operation information of the RRH which respectively correspond to one or more timestamps; and issuing a notification indicating that the RRH is faulty when it is predicted that the RRH is faulty, wherein the one or more pieces of operation information of the RRH comprise at least one of information related to an uptime of the RRH, information related to a type of the alarm, information related to a state of the RRH, and information related to an environment of the RRH.