High-Voltage Generator Fault Diagnosis Using Decision Forest Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional fault diagnosis methods for high-voltage generators are limited in accurately locating and explaining faults, failing to provide early warnings, which reduces the timeliness and accuracy of diagnosis.
Innovation Solution
A fault diagnosis method and system that collects and preprocesses data to generate decision tables, reduces attributes, builds decision tree and forest models, and determines fault types and levels, enabling early warning and improved diagnosis accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional threshold-based signal processing is used for fault diagnosis, then the diagnosis process is simple and fast, but the accuracy and timeliness of fault diagnosis deteriorates because faults cannot be located or warned early
Solution Approach 1:
The patent segments the fault diagnosis process into multiple stages: data collection, data preprocessing, decision table generation, attribute reduction, decision tree model building, and fault diagnosis. This segmentation allows complex analysis to be performed systematically while maintaining manageable complexity through modular processing steps.
Solution Approach 2:
The patent performs preliminary actions by collecting and preprocessing operational data before faults occur, building decision tree models in advance based on historical data. This enables early warning and accurate fault location before serious failures happen, improving diagnosis accuracy without requiring complex real-time analysis during fault occurrence.
2Loss of time
If traditional fixed protection strategy is used, then the protection action is simple and rapid, but the timeliness and accuracy of fault diagnosis deteriorates because faults are only detected after serious damage occurs
Solution Approach 1:
The system performs preliminary data collection and model building during normal operation, creating a knowledge base of fault patterns in advance. When faults occur, the pre-built decision tree models enable rapid and accurate diagnosis, reducing detection time while improving reliability through comprehensive fault coverage.
Solution Approach 2:
The patent implements feedback mechanisms where diagnostic results and fault information are continuously fed back to refine the decision tree models. This iterative improvement enhances the reliability of fault diagnosis over time while maintaining rapid response capability through the structured decision-making framework.
3Loss of information
If traditional signal processing without learning mechanism is used, then the system structure is simple, but the ability to explain diagnosis results and provide early warning deteriorates
Solution Approach 1:
The patent segments information processing into distinct modules: data collection, preprocessing, decision table generation, attribute reduction, and model building. This segmentation preserves fault location and reason information through structured processing while managing complexity through systematic organization of data flow and analysis steps.
Solution Approach 2:
The decision tree models act as intermediaries between raw operational data and fault diagnosis conclusions. These models process and interpret complex data patterns, providing explainable fault location and reason information while bridging the gap between simple data collection and comprehensive diagnostic output.
Data Source
AI summary
The present disclosure relate to a fault diagnosis method, including: collecting at least one set of original data of the high voltage generator during operation, each condition attribute in a condition attribute set corresponding to one piece of original data; pre-processing the original data to obtain pre-processed data and generating a decision table; performing an attribute reduction based on the decision table to obtain a reduced condition attribute set and a reduced decision table; generating a decision tree model according to the reduced decision table; generating a decision forest model according to at least one decision tree model; obtaining a fault type and a fault level according to the decision forest model; determining an acceptable level of the fault according to the fault level and a frequency of occurrence of the same type of fault, and sending the fault type and the acceptable level to the display terminal.