基于宽度卷积神经网络的水轮机空化状态识别方法

刘忠; 郑佳稳; 邹淑云; 乔帅程; 谢文婷

doi:10.19805/j.cnki.jcspe.2025.240280

欢迎来到能源电力期刊网！

English Version 期刊登录

您当前的位置：

首页 >

文章列表页 >

基于宽度卷积神经网络的水轮机空化状态识别方法

动力设备与系统 | 更新时间：2026-10-10

- 基于宽度卷积神经网络的水轮机空化状态识别方法
- Hydraulic Turbine Cavitation State Recognition Method Based on Broad Convolutional Neural Network
- 动力工程学报 2025年45卷第7期页码：985-991
- 作者机构：
  
  长沙理工大学能源与动力工程学院,湖南,长沙,410114
- 作者简介：
  
  刘忠(1978-),男,湖北汉川人,教授,博士,研究方向为水力机械状态监测与故障诊断,E-mail:drliuzhong@csust.edu.cn
- 基金信息：
  
  国家自然科学基金资助项目(52079011);湖南省自然科学基金资助项目(2023JJ30032)
- DOI：10.19805/j.cnki.jcspe.2025.240280
  中图分类号：
- 网络首发：2025-07-14，
  
  纸质出版：2025
- 稿件说明：
移动端阅览
刘忠,郑佳稳,邹淑云,乔帅程,谢文婷. 基于宽度卷积神经网络的水轮机空化状态识别方法动力工程学报, 2025, 45(7): 985-991 https://doi.

org/10.19805/j.cnki.jcspe.2025.240280
刘忠,郑佳稳,邹淑云,乔帅程,谢文婷. 基于宽度卷积神经网络的水轮机空化状态识别方法动力工程学报, 2025, 45(7): 985-991 https://doi. DOI： 10.19805/j.cnki.jcspe.2025.240280.

org/10.19805/j.cnki.jcspe.2025.240280 DOI：

摘要

针对难以有效提取水轮机声发射(AE)信号故障特征

以及在训练过程中信号频谱故障分类不精确且耗时长的问题

建立了一种宽度卷积神经网络(BCNN)的水轮机空化状态识别方法。采用卷积神经网络(CNN)的卷积层和池化层代替宽度学习(BLS)的线性映射特征层

将得到的CNN特征节点作为增强层的输入

通过训练网络参数建立BCNN模型。利用BCNN模型对不同工况下的水轮机空化AE信号进行分类训练和测试。结果表明:所建立的BCNN模型能对水轮机4种空化状态进行准确快速的分类识别

其平均识别准确率达99.37%

平均耗时为44.16 s。

Abstract

In response to the difficulties in effectively extracting fault characteristics from acoustic emission (AE) signals of the hydraulic turbines

as well as the problems of imprecise and time-consuming fault classification in signal spectrum during the training process

a method for hydraulic turbine cavitation state recognition based on broad convolutional neural network (BCNN) was established. The linear mapping feature layers of broad learning system(BLS) were replaced by the convolutional layers and pooling layers of convolutional neural network (CNN)

the obtained CNN feature nodes were used as inputs for the enhancement layers. By training the network parameters

a BCNN model was built. The BCNN model was used to classify

train and test AE signals from the hydraulic turbine cavitation under different working conditions. Results show that the established BCNN model can accurately and quickly classify and recognize four different cavitation states of the hydraulic turbines

achieving average recognition accuracy rate of 99.37%

with average time consumption of 44.16 s.

关键词

Keywords

references

LUO Xianwu, JI Bin, TSUJIMOTO Y. A review of cavitation in hydraulic machinery[J]. Journal of Hydrodynamics, 2016, 28(3): 335-358.

ESCALER X, EGUSQUIZA E, FARHAT M, et al. Detection of cavitation in hydraulic turbines[J]. Mechanical Systems and Signal Processing, 2006, 20(4): 983-1007.

郑祥豪. 可逆式水泵水轮机运行状态监测与智能故障诊断研究[D]. 北京: 华北电力大学, 2022.

DAO Fang, ZENG Yun, QIAN Jing. Fault diagnosis of hydro-turbine via the incorporation of Bayesian algorithm optimized CNN-LSTM neural network[J]. Energy, 2024, 290: 130326.

刘忠, 李显伟, 邹淑云, 等. 基于混沌理论和CNN-OSVM的水轮机空化状态识别方法[J]. 动力工程学报, 2023, 43(11): 1454-1460. LIU Zhong, LI Xianwei, ZOU Shuyun, et al. State recognition method of hydraulic turbine cavitation based on chaos theory and CNN-OSVM[J]. Journal of Chinese Society of Power Engineering, 2023, 43(11): 1454-1460.

GAISSER L, KIRSCHNER O, RIEDELBAUCH S. Cavitation detection in hydraulic machinery by analyzing acoustic emissions under strong domain shifts using neural networks[J]. Physics of Fluids, 2023, 35(2): 027128.

ZHENG Xianghao, LI Hao, ZHANG Suqi, et al. Hydrodynamic feature extraction and intelligent identification of flow regimes in vaneless space of a pump turbine using improved empirical wavelet transform and Bayesian optimized convolutional neural network[J]. Energy, 2023, 282: 128705.

张苏祺, 李浩, 张宇宁, 等. 水泵水轮机复杂振动信号特征提取与智能识别[J]. 水力发电学报, 2023, 42(12): 70-78. ZHANG Suqi, LI Hao, ZHANG Yuning, et al. Feature extraction and intelligent recognition of complicated vibration signals of pump turbine[J]. Journal of Hydroelectric Engineering, 2023, 42(12): 70-78.

CHEN C L P, LIU Zhulin. Broad learning system: an effective and efficient incremental learning system without the need for deep architecture[J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 29(1): 10-24.

YU Wanke, ZHAO Chunhui. Broad convolutional neural network based industrial process fault diagnosis with incremental learning capability[J]. IEEE Transactions on Industrial Electronics, 2020, 67(6): 5081-5091.

PU Xiaokun, LI Chunguang. Online semisupervised broad learning system for industrial fault diagnosis[J]. IEEE Transactions on Industrial Informatics, 2021, 17(10): 6644-6654.

FENG Jian, YAO Yu, LU Senxiang, et al. Domain knowledge-based deep-broad learning framework for fault diagnosis[J]. IEEE Transactions on Industrial Electronics, 2021, 68(4): 3454-3464.

LIU Licheng, LIU Tingyun, CHEN C L P, et al. Modal-regression-based broad learning system for robust regression and classification[J]. IEEE Transactions on Neural Networks and Learning Systems, 2024, 35(9): 12344-12357.

刘忠, 邹淑云, 陈莹, 等. 混流式水轮机模型空化状态与声发射信号特征关系试验[J]. 动力工程学报, 2016, 36(12): 1017-1022. LIU Zhong, ZOU Shuyun, CHEN Ying, et al. Experiments on the relationship between cavitation status and acoustic emission signal features for a Francis turbine model[J]. Journal of Chinese Society of Power Engineering, 2016, 36(12): 1017-1022.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

煤-氨混燃锅炉建模及近邻传播-最小支持向量机多模态验证优化

分形特征对漂浮式风力机仿生平台动态响应的影响

基于WPHM和非支配性遗传算法的燃气轮机透平叶片多目标优化维护决策

考虑风光不确定性的风电光伏与燃气电站联合参与日前-实时市场的交易决策研究

基于多通道特征提取与深度交互融合网络的IES多元负荷短期预测