Alert Similarity Label Transfer for Sensor-Based Fault Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Comparing historical sensor data to current alerts for diagnosing anomalous device behavior is challenging due to changes over time, differences among assets, and variations in normal operating states, leading to inefficient and inconsistent troubleshooting processes.
Innovation Solution
A method and system that process feature data from multiple sensor devices to generate feature importance data, allowing for the identification of similar historical alerts based on relative importance values, thereby facilitating automated label transfer and improved diagnostic accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If historical sensor data is compared to current alerts for diagnosing anomalous device behavior, then diagnostic information can be obtained, but the accuracy and reliability of comparison deteriorates due to changes over time, differences among assets, and variations in normal operating states
Solution Approach 1:
The patent transforms raw sensor data into standardized feature representations that capture essential operational characteristics while being invariant to temporal changes, asset-specific variations, and normal operating state differences. This parameter transformation enables reliable comparison across different time periods and assets by focusing on discriminative features rather than raw values
Solution Approach 2:
The patent introduces an intermediary processing layer (feature extraction and representation module) between the raw historical sensor data and the comparison operation. This intermediary transforms diverse raw data into a standardized feature space where meaningful comparisons can be performed, acting as a mediator that reconciles differences in data sources
2Measurement precision
If manual comparison of historical alert data is performed by subject matter experts, then diagnostic accuracy can be maintained, but troubleshooting time and productivity are reduced
Solution Approach 1:
The patent implements automated alert similarity comparison and label transfer functionality that performs diagnostic assistance without requiring manual expert intervention. The system serves itself by automatically identifying similar historical alerts, comparing their features, and transferring diagnostic labels, thereby maintaining accuracy while dramatically improving troubleshooting speed
Solution Approach 2:
The patent replaces the manual mechanical process of expert comparison with an automated computational system. The automated system uses algorithmic feature comparison and similarity metrics to substitute for human expert analysis, maintaining diagnostic accuracy while eliminating the time cost of manual review
3Quantity of substance
If raw sensor data is directly compared without processing, then all original information is preserved, but the complexity of handling variations in data characteristics increases
Solution Approach 1:
The patent extracts essential diagnostic features from raw sensor data, separating the critical information needed for alert comparison from the extraneous variations. By taking out only the relevant feature components and discarding redundant information, the system reduces processing complexity while maintaining the quantity of diagnostically useful information
Data Source
AI summary
A method of identifying a historical alert that is similar to an alert associated with a detected deviation from an operational state of a device includes receiving feature data including time series data for multiple sensor devices associated with the device and receiving an alert indicator for the alert. The method includes processing a portion of the feature data that is within a temporal window associated with the alert indicator to generate feature importance data for the alert. The feature importance data includes values indicating relative importance of each of the sensor devices to the alert. The method also includes identifying one or more historical alerts that are most similar, based on the feature importance data and stored feature importance data, to the alert.