Facility Abnormality Detection Using PCA on Time-Series Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality monitoring techniques for hydraulic screw-down devices in steel rolling processes are insufficient in detecting non-linear characteristics and require extensive manpower to set appropriate values for each facility, leading to inefficient and inaccurate abnormality detection.
Innovation Solution
An abnormality determination apparatus and method that uses time-series signal clipping, normal vector registration, and principal component analysis to detect abnormalities in facilities with non-linear characteristics, employing a decision tree-based trigger condition decision model for universal and high-accuracy detection without individual facility monitoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If reference value and actual value monitoring is used for servo valve spool movement, then abnormality detection is performed, but the amount of information for monitoring is insufficient
Solution Approach 1:
The patent transitions from monitoring only spool movement amount (1D) to constructing M-dimensional vectors that include spool movement amount, excitation current, and oil column cylinder position (17). This dimensional expansion provides comprehensive monitoring information for accurate abnormality detection in non-linear hydraulic systems.
2Reliability
If appropriate values are set for proportional relationship and saturation state in excitation current and oil column cylinder position monitoring, then abnormality monitoring is performed, but it requires extensive manpower for each facility
Solution Approach 1:
The system performs self-service by automatically determining appropriate values for monitoring parameters through data collection and analysis during normal operation. The abnormality determination apparatus autonomously sets the proportional relationship threshold and saturation state values without requiring manual configuration for each facility, reducing manpower requirements while maintaining reliable abnormality monitoring.
3Measurement precision
If individual facility monitoring with manual parameter setting is performed, then accurate abnormality detection is achieved, but the process becomes complex and time-consuming
Solution Approach 1:
The patent implements preliminary action by collecting and analyzing operational data during the normal operation phase to pre-determine monitoring parameters and thresholds. This preliminary data collection and analysis enables the system to be ready for immediate abnormality detection without requiring time-consuming manual parameter setting when abnormalities occur, reducing loss of time while maintaining detection accuracy.
Data Source
AI summary
An abnormality determination apparatus: performs, during normal operation of the facility, K times of clipping from time-series signals indicating an operation state of the facility; sets M types as types of the time-series signals clipped by the time-series signal clipping unit, constructs an M-dimensional vector, and registers the constructed vector as a normal vector; sets an abnormality determination flag as a first type when a maximum value of correlation between variables is less than a predetermined value, sets an abnormality determination flag as a second type when the maximum value is the predetermined value or more, and performs, when the flag is of the second type, a principal component analysis on a registered normal vector group to calculate a transform coefficient of a principal component and registers each of the calculated transform coefficients as an abnormality determination model; and determines an abnormality of the facility.