Anomaly Detection Filtering for Battery Metric Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current anomaly detection systems in industrial and retail facilities face challenges in efficiently identifying and filtering anomalies from data collected by devices such as mobile computing devices, printers, and scanners, particularly in battery performance metrics like cycle count, temperature, and discharge rate.
Innovation Solution
A data model training system that receives and updates data periodically, applies predefined filters to identified anomalies, and records only reportable anomalies, using methods like Isolation Forest models and incremental data analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If anomaly detection systems analyze all collected data from devices, then detection accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent segments the data analysis process into two distinct phases: a training phase using a larger first time period of historical data to build the data model, and a detection phase using a smaller second time period of recent data to identify anomalies. This segmentation allows the system to achieve high detection accuracy while reducing the processing time for real-time anomaly identification.
Solution Approach 2:
The system performs preliminary action by training the data model in advance using the first time period of data before actual anomaly detection begins. This pre-training step prepares the model to quickly and accurately detect anomalies in the second time period without requiring extensive processing during real-time operation.
2Reliability
If the system identifies all potential anomalies, then detection completeness improves, but false positives increase
Solution Approach 1:
The patent introduces predefined filters as intermediary components between anomaly detection and final reporting. These filters act as mediators that evaluate detected anomalies against established criteria and thresholds, allowing the system to maintain high detection completeness while reducing false positives by filtering out spurious anomalies before they are reported.
3Quantity of substance
If the system processes cumulative metrics, then data completeness improves, but anomaly detection accuracy deteriorates
Solution Approach 1:
The system dynamically adapts its processing approach based on the type of metric. For cumulative metrics, it processes the raw cumulative values to maintain data completeness. For non-cumulative metrics, it calculates delta values (changes from previous measurements) to improve anomaly detection accuracy. This dynamic processing strategy allows the system to optimize for the appropriate metric depending on the specific measurement type.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods, apparatuses, and computer program products for anomaly detection are provided. For example, a computer-implemented method includes receiving data corresponding to a plurality of metrics from a plurality of devices, inputting at least a first portion of the received data covering a first time period into a data model to train the data model, inputting at least a second portion of the received data covering a second time period into the trained data model, identifying from the data model one or more anomalies corresponding to one or more of the plurality of metrics of one or more of the plurality of devices, determining if a predefined filter exists for the one or more of the plurality of metrics corresponding to each of the one or more anomalies identified, and applying the predefined filter to determine if the corresponding identified anomaly will be recorded as a reportable anomaly.