Switch Open-Circuit Fault Diagnosis Using DC and RMS Current Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for diagnosing open-circuit faults in switches require high-specification computers and prolonged training times, and are susceptible to human error in pattern classification.
Innovation Solution
A system and method using an artificial neural network model that inputs DC components and RMS values of three-phase current, with a simplified calculation method, to diagnose open-circuit faults during regenerative motor operation, reducing the need for high-specification computers and training time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If two artificial neural networks are used with DC components and total harmonic distortion as input values, then open-circuit fault diagnosis accuracy is improved, but computer specification requirements increase and training time is prolonged
Solution Approach 1:
The patent extracts only the essential features (DC components and RMS values of three-phase current) needed for fault diagnosis, eliminating unnecessary processing steps and intermediate networks. This reduces the computational burden while maintaining diagnostic accuracy, directly resolving the contradiction between diagnosis precision and computer specification requirements
Solution Approach 2:
Instead of using complex two-network architectures with multiple processing stages, the patent inverts the approach by using a single simplified neural network with directly calculated input features. This inversion reduces system complexity and training time while preserving the ability to accurately diagnose open-circuit faults in switches
2Measurement precision
If two artificial neural networks are used with DC components and total harmonic distortion as input values, then open-circuit fault diagnosis accuracy is improved, but training time is prolonged
Solution Approach 1:
The patent extracts only the essential features (DC components and RMS values of three-phase current) needed for fault diagnosis, eliminating unnecessary processing steps and intermediate networks. This reduces the computational burden while maintaining diagnostic accuracy, directly resolving the contradiction between diagnosis precision and training time
Solution Approach 2:
The patent uses a simplified single-network approach with selective feature extraction (DC components and RMS values only), avoiding the excessive computation of two full neural networks. This partial action approach maintains sufficient diagnostic capability while significantly reducing training time
3Ease of operation
If manual classification of current distortion patterns is performed, then fault diagnosis can be conducted, but human error increases
Solution Approach 1:
The patent implements an automated neural network system that performs fault diagnosis independently without human intervention. The system automatically processes current measurements, calculates DC components and RMS values, and identifies faulty switches, eliminating human error while maintaining ease of operation through self-service automation
Solution Approach 2:
The patent replaces the manual mechanical classification process with an automated computational neural network system. This substitution eliminates human error in pattern recognition while maintaining operational simplicity through automated processing of electrical signals
Data Source
Figure 1
Figure 2(a)~2(d)
Figure 3(a)~3(d)
AI summary
Disclosed is a system for diagnosing open-circuit faults in switches using an artificial neural network model. In the switch open-circuit fault diagnosis system, data input to an input layer of the artificial neural network model includes DC components and RMS values of three-phase current.