Related Experiment Video
Updated: Sep 5, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.2K
An Intelligent Fault Diagnosis Based on Adversarial Generating Module and Semi-supervised Convolutional Neural
1School of Computer Science, Yangtze University, Jingzhou 430023, China.
Computational Intelligence and Neuroscience
|July 5, 2022
Summary
This study introduces a novel semi-supervised intelligent fault diagnosis framework using a generative adversarial network (GAN) and semi-supervised convolutional neural network (SSCNN). The approach effectively diagnoses machinery faults with high accuracy using limited labeled data.
Area of Science:
- Mechanical Engineering
- Artificial Intelligence
- Machine Learning
Background:
- Machinery monitoring data often suffers from high labeling costs and a scarcity of typical failure samples.
- Intelligent fault diagnosis requires robust methods that can leverage abundant unlabeled data.
Purpose of the Study:
- To develop a semi-supervised intelligent fault diagnosis framework addressing limitations of labeled data.
- To enhance fault diagnosis accuracy by utilizing both labeled and unlabeled data effectively.
Main Methods:
- A novel framework combining a generative adversarial module (GAN) with a semi-supervised convolutional neural network (SSCNN).
- Implementation of manifold-regularization-based fuzzy clustering discrimination (MRFCD) for unlabeled data utilization.
- Construction of pseudo-labels and objective function estimation for unlabeled data within the CNN framework.
Main Results:
- The proposed GAN-MRFCD-SSCNN framework achieved 96.2% accuracy in diagnosing faults.
- The framework demonstrated superior performance compared to classical semi-supervised diagnostic models.
- Effective utilization of vibrational signals from main reducers for validation.
Conclusions:
- The developed semi-supervised fault diagnosis framework is effective and accurate, even with limited labeled samples.
- The integration of GAN and MRFCD-SSCNN offers a promising solution for intelligent machinery monitoring.
- This approach significantly reduces the reliance on extensively labeled datasets for fault diagnosis.
