Automatic Tag Selection for Industrial Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial control systems generate overwhelming volumes of data from sensors and control elements, making it difficult to analyze and manage effectively, particularly in identifying interrelated components and suggesting relevant data tags for analysis and anomaly detection.
Innovation Solution
A computer-implemented method using a server system that calculates time-weighted averages, performs linear regression, and normalizes residual values to automatically associate tags, allowing for the identification of closely related tags and reducing data overload by suggesting relevant tags for analysis and anomaly detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If all sensor data is distributed to user devices for analysis, then complete data availability is achieved, but network resources are unnecessarily consumed and system performance deteriorates
Solution Approach 1:
The system extracts and identifies only the most relevant tags and data points that are critical for anomaly detection, separating them from the overwhelming volume of irrelevant data. This extraction process allows the system to distribute only essential information to user devices, maintaining data availability for critical parameters while significantly reducing network resource consumption by filtering out redundant data.
Solution Approach 2:
The system applies local quality by providing different data sets to different user devices based on their specific needs and contexts. Each user receives tailored data relevant to their role and monitoring requirements, rather than distributing all raw data universally. This approach ensures each user has access to necessary information while minimizing overall network traffic.
2Reliability
If thousands of sensor data tags are analyzed manually, then comprehensive monitoring is achieved, but analysis time and operational complexity increase significantly
Solution Approach 1:
The system performs self-service by automatically identifying relationships between tags, detecting anomalies, and generating alerts without requiring manual analysis of thousands of data points. The automated anomaly detection engine continuously monitors data streams, applies detection algorithms, and identifies issues independently, ensuring comprehensive monitoring while eliminating the time burden of manual analysis.
Solution Approach 2:
The system performs preliminary action by pre-processing and pre-analyzing data streams in real-time, identifying patterns and relationships before presenting information to users. This preliminary analysis includes automatic tag association, data validation, and anomaly pre-detection, which prepares the data in advance and reduces the time required for final analysis and decision-making.
3Measurement precision
If manual tag association is performed to identify interrelated components, then relationship accuracy is achieved, but system complexity and operational burden increase
Solution Approach 1:
The system replaces the mechanical process of manual tag association with automated computational algorithms. The anomaly detection engine uses data-driven methods to automatically identify relationships between tags based on their interconnections and correlations, eliminating the need for manual configuration while maintaining or improving relationship detection accuracy through objective, consistent analysis.
Solution Approach 2:
The system introduces an intermediary automated tag association layer that mediates between raw sensor data and user interpretation. This intermediary component automatically establishes and maintains tag relationships based on data patterns and system knowledge, providing accurate associations without requiring direct manual intervention, thereby reducing operational complexity while preserving precision.
Data Source
AI summary
A server system can operate to function as an automatic association of tags defining a system within a process. The operations include accessing tags and associated signals including a plurality of data values over time indicative of a physical property, behavior or measurement of a component of the process. For each signal, calculating a time-weighted average over a specific time period, selecting a specific number of different day periods sampled from the tags and signals, and for each tag, calculating a slope and intercept by calculating a linear regression of plurality of signals over the specific time period. Further, calculating a residual value of each data value of the signals over the specific time period, and calculating a normalized value of each residual value, and then calculating the absolute value of the dot product of the normalized residual value and the residual value of a subsequent number of tags.