Related Experiment Video
Updated: May 8, 2025
Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks
Published on: March 2, 2015
Reliability evaluation of rolling bearings based on generative adversarial network sample enhancement and maximum
Fannian Meng1, Liujie Wang2, Hao Li2
1Henan Key Laboratory of Intelligent Manufacturing of Mechanical Equipment, Zhengzhou University of Light Industry, Zhengzhou, 450002, China. wangljzzuli@gmail.com.
This study introduces a novel bearing reliability evaluation method using generative adversarial networks (GANs) for data enhancement and the maximum entropy method for analysis, especially effective with limited vibration data. The approach accurately predicts bearing degradation stages and remaining life, outperforming existing models.
Area of Science:
- Mechanical Engineering
- Reliability Engineering
- Data Science
Background:
- Extracting vibration data for rolling bearing reliability evaluation under real-world conditions is challenging.
- Limited sample sizes often hinder traditional reliability analysis methods.
Purpose of the Study:
- To propose an effective bearing reliability evaluation method for scenarios with few vibration samples.
- To enhance data samples using generative adversarial networks (GANs) and establish a reliability model via the maximum entropy principle and Poisson process.
Main Methods:
- Generative adversarial network (GAN) based data sample enhancement for limited datasets.
- Reliability analysis model integrating maximum entropy principle and Poisson process.
- Evaluation based on reliability variation frequency, speed, and acceleration.
Main Results:
- Reliability variation frequency exhibits a nonlinear growth trend, identifying distinct running-in stages (initial, stable, intense).
- Reliability variation speed helps pinpoint the start times of these stages.
- A preliminary relationship between reliability variation acceleration and remaining bearing life was established.
Conclusions:
- The proposed method effectively evaluates bearing reliability with fewer samples and no preprocessing.
- It offers higher accuracy compared to existing models, demonstrated on the XJTU-SY dataset.
- This work provides a valuable addition to existing bearing reliability analysis techniques.
