Visual Parameterization of Density-Based Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current anomaly detection methods in industrial plants, particularly those using density-based clustering algorithms, face challenges in parameterization, requiring time-consuming and user-dependent adjustments to distinguish normal from abnormal behavior effectively, and often fail to account for domain-specific knowledge.
Innovation Solution
A method that maps sensor data points into a pixel space, allowing for rapid visualization and simulation of clustering operations using high-performance hardware, enabling quick parameter adjustments and intuitive understanding of clustering results without deep technical knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a density-based clustering algorithm is used for anomaly detection, then the selectivity in distinguishing normal from abnormal behavior is improved, but the time required for parameterization and calculation increases significantly
Solution Approach 1:
The patent creates a visual copy of the clustering process by mapping sensor data points to pixel positions and representing cluster formations as visual patterns. This visual representation allows users to understand and adjust parameters without performing actual clustering calculations, dramatically reducing parameterization time while maintaining detection accuracy.
Solution Approach 2:
The patent introduces a visual intermediary layer between the clustering algorithm and the user. Instead of directly interacting with complex algorithmic parameters, users interact with visual representations that mediate the parameter adjustment process, making it intuitive and rapid.
2Productivity
If automated parameter determination methods are used, then the parameterization process is speeded up, but the results only approximate the optimum and do not account for domain-specific knowledge
Solution Approach 1:
The patent implements a feedback mechanism where the visual representation of cluster formations provides immediate feedback to the user about the effect of parameter changes. Users can iteratively adjust parameters based on visual feedback until the desired cluster formation is achieved, ensuring both speed and accuracy.
Solution Approach 2:
The visual representation system allows users to independently assess and adjust parameters based on their domain knowledge and the visual feedback provided by the system, without relying on automated determination methods that may not account for specific application requirements.
3Measurement precision
If manual parameter adjustment is performed to achieve optimal clustering results, then the anomaly detection accuracy is improved, but the calculation time increases from seconds to hours
Solution Approach 1:
The patent creates a visual copy of the data space where cluster formations can be observed directly as visual patterns. This allows rapid assessment of clustering results and parameter effectiveness without performing multiple time-consuming calculation cycles, achieving both accuracy and speed.
Solution Approach 2:
The patent replaces the mechanical calculation process with a visual representation system. Instead of iteratively running clustering algorithms to assess parameter effectiveness, the system visually represents the effects of parameters, allowing instantaneous assessment and adjustment.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A method for parameterizing an anomaly detection method, which takes a multiplicity of sensor data points as a basis for performing a density-based cluster method, comprising a) mapping (S10) each sensor data point (SP1, SP2, SP3) in a data space into a pixel data point (PP1, PP2, PP3) in a pixel space, b) reproducing (S11) at least one operation of the density-based cluster method in the data space by means of at least one pixel operation in the pixel space, c) receiving (S12) at least one parameter value for each parameter of the density-based cluster method, d) applying (S13) the at least one pixel operation in accordance with the parameter values to the pixel data points (PP1, PP2, PP3), e) outputting (S14) a cluster result in visual form in the pixel space, and f) providing (S16) the received parameter values for the anomaly detection method, and an assistance apparatus (80) for parameterizing an anomaly detection apparatus (90) that performs the anomaly detection method.