Related Experiment Video
Updated: Sep 19, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.9K
Circular saw blade wear status prediction based on generative adversarial network and CNN-LSTM model
Chao Zeng1, Chengchao Wang2, Xueqin Xiong3
1School of Nuclear Science and Technology, University of South China, Hengyang, China.
Plos One
|June 18, 2025
Summary
This study introduces a new method for predicting circular saw blade wear using generative adversarial networks (GAN) and CNN-LSTM models. The approach significantly improves accuracy in complex conditions, enhancing safety and efficiency in spent fuel assembly cutting.
Area of Science:
- Mechanical Engineering
- Artificial Intelligence
- Materials Science
Background:
- Accurate monitoring of circular saw blade wear is crucial for efficient and safe cutting of spent fuel assemblies.
- Current prediction models struggle with complex working conditions due to data limitations and model robustness issues.
Purpose of the Study:
- To develop an accurate and robust method for predicting circular saw blade wear in challenging environments.
- To address the challenges of data acquisition, complex feature extraction, and model reliability in wear prediction.
Main Methods:
- A novel approach combining generative adversarial networks (GAN) and CNN-LSTM models for wear prediction.
- Data preprocessing including overlapping sampling, wavelet denoising, and normalization.
- GAN, optimized by Pearson correlation coefficient (PCC), augmented data to 300 samples per fault state (2100 total).
Main Results:
- The proposed CNN-LSTM model achieved 100% accuracy in identifying circular saw blade wear status.
- This performance surpasses traditional models like Long Short-Term Memory (LSTM) at 86.2% and Radial Basis Function Neural Networks (RBFNNs) at 94.9%.
Conclusions:
- The study successfully overcomes the limitation of small sample sizes in wear prediction datasets.
- The developed method offers an efficient and accurate solution for circular saw blade wear identification in complex conditions, vital for improving nuclear fuel reprocessing safety and efficiency.
