Massive MIMO Alarm Threshold Calibration Using KPI-Driven ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems use fixed alarm thresholds for hardware and antenna calibration in Massive MIMO radios, leading to false alarms and unnecessary radio decommissioning due to non-critical faults, which are not adequately addressed by current methods.
Innovation Solution
Implement a machine learning-based fault classifier model that utilizes network-level KPI data and radio log data to dynamically adjust alarm thresholds based on the impact on network performance, distinguishing between critical and non-critical faults using features like faulty antenna branches and beamforming weights.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If fixed alarm thresholds are used for hardware and antenna calibration faults, then alarm generation is simple and deterministic, but false alarms increase and system reliability decreases due to non-critical faults being treated as critical
Solution Approach 1:
The patent applies dynamics by transforming the static fixed alarm threshold into a dynamic adaptive threshold that changes based on real-time network conditions and KPI degradation levels. The ML model continuously learns from historical data and adjusts thresholds to match actual system behavior, making the alarm system adaptable rather than rigid.
Solution Approach 2:
The patent changes the parameter of alarm threshold from a fixed value to a variable determined by multiple factors including KPI degradation levels, fault patterns, and network conditions. The ML model outputs optimized threshold values that are continuously updated based on learned patterns, effectively changing the threshold parameter dynamically.
2Reliability
If fixed alarm thresholds are used, then the alarm system is easy to operate and maintain, but unnecessary radio decommissioning occurs due to false alarms
Solution Approach 1:
The patent implements self-service by enabling the alarm system to automatically optimize its own thresholds through the ML model without requiring manual intervention. The system autonomously learns from historical data, identifies patterns, and adjusts thresholds independently, reducing the need for operator involvement in threshold management.
Solution Approach 2:
The patent applies feedback by using actual alarm outcomes and KPI degradation data to continuously refine and optimize alarm thresholds. The ML model receives feedback from the system's operational data and adjusts future threshold decisions based on learned patterns, creating a closed-loop optimization system.
3Measurement precision
If fixed alarm thresholds are used, then alarm processing is straightforward, but the ability to distinguish between critical and non-critical faults is insufficient
Solution Approach 1:
The patent applies segmentation by dividing faults into different categories based on their criticality levels. The ML model analyzes multiple features and segments faults into critical, non-critical, and conditional categories, allowing differentiated handling and response strategies for each segment.
Solution Approach 2:
The patent adds another dimension to fault analysis by incorporating multiple features beyond just the number of faulty branches. The ML model considers beamforming weights, KPI degradation levels, and other parameters, transforming the one-dimensional threshold check into a multi-dimensional assessment.
Data Source
AI summary
Systems and methods are disclosed that relate to Key Performance Indicator (KPI)-driven alarm threshold optimization using machine learning. In one embodiment, a computer-implemented method comprises obtaining network-level KPI data for a network and radio log data for one or more radio systems in the network. The KPI data comprises KPI values for one or more network-level KPIs, and the radio log data comprises a number of faulty or uncalibrated antenna branches in the radio system and cell or user beamforming weights. The method further comprises pre-processing the KPI data and the radio log data, labeling the pre-processed KPI data as degraded or non-degraded, and training, with the labeled KPI data and the pre-processed radio log data, a fault classifier Machine Learning (ML) model to output a value(s) that represent a probability that the KPI(s) will be degraded for a given input feature set.