Unsupervised Anomaly Detection Across Machine Operating Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional anomaly detection methods in industrial systems fail to account for high-dimensional data scenarios, machine-specific variables, and operating regions, leading to inaccurate anomaly identification and potential machine downtime or failure.
Innovation Solution
An unsupervised anomaly detection system that leverages domain knowledge to subsample and cluster historic operating data, using Mahalanobis distance to calculate distances between nearest neighbors in a feature space, allowing for accurate anomaly detection by considering machine-specific parameters and operating regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional anomaly detection methods are used, then the system can operate with simple detection logic, but the detection accuracy deteriorates in high-dimensional data scenarios
Solution Approach 1:
The patent segments the high-dimensional data space by dividing it into multiple operating regions based on domain knowledge. Each region is characterized by specific parameter ranges and relationships, allowing the system to perform simpler local anomaly detection within each region rather than attempting complex global detection across all dimensions simultaneously.
Solution Approach 2:
The patent transforms the anomaly detection problem from direct high-dimensional space comparison to a two-stage process: first projecting data into operating region space (one dimension), then performing nearest neighbor comparison within selected regions (another dimension). This dimensional transformation reduces the effective complexity while preserving detection accuracy.
2Measurement precision
If domain knowledge is incorporated to create machine-specific solutions, then the detection accuracy improves, but the system adaptability deteriorates across different machines
Solution Approach 1:
The patent creates a universal framework where the same anomaly detection algorithm operates across different machines, but adapts to each machine's specific characteristics through domain knowledge parameters. The system can be configured with machine-specific parameter ranges, relationships, and operating regions while maintaining the same core detection logic, enabling both accuracy and adaptability.
Solution Approach 2:
The patent allows domain knowledge to be encoded as configurable parameters such as parameter ranges, relationships, and operating region definitions. These parameters can be adjusted for different machines without changing the underlying detection algorithm, enabling the system to adapt to machine-specific characteristics while maintaining universal applicability through parameter modification rather than structural changes.
3Reliability
If all historic operating data is used for analysis, then the detection comprehensiveness improves, but the data processing time deteriorates
Solution Approach 1:
The patent extracts and utilizes only the most relevant historical data by identifying and comparing test data points to their nearest neighbors within the same operating region. Rather than processing all historical data uniformly, the system extracts only the subset of data that is actually relevant for anomaly detection, significantly reducing processing time while maintaining detection comprehensiveness.
Solution Approach 2:
The patent applies partial action by performing anomaly detection only within selected operating regions rather than across the entire data space. The system identifies the nearest neighbor within the same operating region and performs comparison only there, rather than exhaustively comparing against all historical data points, achieving sufficient detection accuracy with reduced computational effort.
4Ease of manufacture
If static detection solutions are used, then the system implementation is simple, but the detection effectiveness deteriorates when machine variables change
Solution Approach 1:
The patent introduces dynamics by allowing the operating regions and their boundaries to be defined based on domain knowledge that captures the dynamic relationships between machine parameters. The system can accommodate changes in machine operating conditions by updating the parameter ranges and relationships that define operating regions, enabling the detection system to adapt to varying machine states while maintaining a relatively simple implementation framework.
Data Source
AI summary
Historic operating data for one or more parameters of a machine are obtained. The historic operating data is subsampled to generate a plurality of clusters based on domain knowledge for the machine, each cluster representing data points from the historic operating data that are associated with an operating region for the machine, the domain knowledge includes one or more model parameters associated with the machine. A model file is generated that includes the plurality of clusters of the data points and the one or more model parameters. Test data from the machine is received. A number of nearest neighbors is calculated from the plurality of clusters of the model file to the test data using an algorithm. A distance of the test data from the number of nearest neighbors is calculated. An action is executed based on comparing the distance to a threshold value.