Research on Sensor Fault Diagnosis of Nuclear Power Plant Based on Improved CWT-CNN
-
摘要: 核电厂传感器故障后果严重，而核电厂一回路系统和设备的固有复杂性为基于精确数学模型的传感器故障诊断带来了困难。本文提出了一种将深度学习算法与时频分析相结合的核电厂传感器智能故障诊断方法，将信号识别问题转化为图像识别问题。先利用连续小波变换（CWT）对核电厂典型传感器7种常见健康状态的时序信号进行处理，以生成捕捉故障信号特征的时频图；然后以预处理和标记的数据集训练经通道注意力机制（CA）改进的卷积神经网络（CNN）模型，提取时频图的细微图像特征，基于这些特征识别和隔离传感器故障。该方法不需建模和设计阈值，鲁棒性强，准确率达到97%以上，通过与长短期记忆（LSTM）神经网络、一维卷积神经网络（1D-CNN）等典型深度学习网络的诊断效果对比，验证了改进CWT-CNN的有效性和优越性。
-
关键词:
- 核电厂 /
- 传感器 /
- 故障诊断 /
- 时频图 /
- 卷积神经网络（CNN） /
- 通道注意力机制（CA）
Abstract: The consequences of sensor faults in nuclear power plants are serious, and the inherent complexity of primary circuit system and equipment in nuclear power plants brings difficulties to sensor fault diagnosis based on accurate mathematical models. In this paper, an intelligent sensor fault diagnosis method for nuclear power plant based on deep learning algorithm and time-frequency analysis is proposed, which transforms the signal recognition problem into image recognition problem. Firstly, the continuous wavelet transform (CWT) is used to process the time series signals of seven common health states of typical sensors in nuclear power plants to generate a time-frequency diagram that captures the characteristics of fault signals. Then, the convolutional neural network (CNN) model improved by channel attention mechanism (CA) is trained with pre-processed and labeled data sets, and the subtle image features of the time-frequency diagram are extracted. Based on these features, sensor faults are identified and isolated. This method does not need to model and design thresholds, and it has strong robustness and an accuracy rate of more than 97%. By comparing the diagnostic effects of typical deep learning networks such as long short-term memory network (LSTM) and one-dimensional convolutional neural network (1D-CNN), the effectiveness and superiority of the improved CWT-CNN are verified. -
图 6 CWT-CNN的测试集混淆矩阵
注：横坐标表示传感器7种真实的健康状况标签；纵坐标表示诊断的健康状况标签；每个橙色和绿色正方形框中2个数字分别表示样本数和准确率；最下行和最右列方框中的数字分别是诊断正确率和误判率的统计情况；右下角小方框中为算法诊断的平均准确率
Figure 6. Test Set Confusion Matrix of CWT-CNN
表 1 传感器健康状态标签
Table 1. Labels for Sensor Health Conditions
健康状态类型 恒偏差 漂移 冲击 工频噪声 精度下降 卡死 正常 标签 1 2 3 4 5 6 7 
下载: 导出CSV
表 2 网络参数
Table 2. Network Parameters
网络层 描述 输出特征图尺寸 输入层 输入数据 256×256×3 卷积层1（C1）+BN层 6个滤波器，大小为5×5 256×256×6 池化层1（P1） 滤波器大小为2×2 128×128×6 卷积层2（C2）+BN层 16个滤波器，大小为5×5 128×128×16 池化层2（P2） 滤波器大小为2×2 64×64×16 SE层（S1） 64×64×16 全连接层（F1） 128节点，Dropout=0.5 128×65536 全连接层（F2） 64节点，Dropout=0.5 64×128 全连接层（F3） 7节点 7×64 输出层 7×1 学习率 0.0001
下载: 导出CSV
表 3 不同方法的故障诊断效果
Table 3. Fault Diagnosis Efficacy of Different Methods
算法 测试集平均
诊断准确率/%平均F1
分数平均训
练时间/s平均测
试时间/s改进后CWT-CNN 97.54 0.98 52.94 0.38 未改进CWT-CNN 95.63 0.96 51.79 0.36 LSTM 93.88 0.94 62.72 0.84 1D-CNN 92.94 0.93 54.34 0.78 RBF 84.68 0.85 55.96 0.72
下载: 导出CSV
-
[1] YANG Z X, XU P, ZHANG B, et al. Nuclear power plant sensor signal reconstruction based on deep learning methods[J]. Annals of Nuclear Energy, 2022, 167: 108765. doi: 10.1016/j.anucene.2021.108765 [2] GAUTAM S, TAMBOLI P K, ROY K, et al. Sensors incipient fault detection and isolation of nuclear power plant using extended Kalman filter and Kullback–Leibler divergence[J]. ISA Transactions, 2019, 92: 180-190. doi: 10.1016/j.isatra.2019.02.011 [3] 刘红艳,麦艳红,孔繁镍,等. 利用自适应粒子滤波的传感器故障诊断识别[J]. 控制工程,2019, 26(7): 1425-1430. [4] YU Y, PENG M J, WANG H, et al. Improved PCA model for multiple fault detection, isolation and reconstruction of sensors in nuclear power plant[J]. Annals of Nuclear Energy, 2020, 148: 107662. doi: 10.1016/j.anucene.2020.107662 [5] ZHU S M, XIA H, ANNOR-NYARKO M, et al. A robust strategy for sensor fault detection in nuclear power plants based on principal component analysis[J]. Annals of Nuclear Energy, 2021, 164: 108621. doi: 10.1016/j.anucene.2021.108621 [6] MESSAI A, MELLIT A, ABDELLANI I, et al. On-line fault detection of a fuel rod temperature measurement sensor in a nuclear reactor core using ANNs[J]. Progress in Nuclear Energy, 2015, 79: 8-21. doi: 10.1016/j.pnucene.2014.10.013 [7] 尹子会,孟延辉,赵智龙,等. 声振信号构造CNN特征矩阵的断路器储能机构故障诊断方法[J]. 高压电器,2023, 59(9): 242-249. [8] 马立玲,郭建,汪首坤,等. 基于改进CNN-GRU网络的多源传感器故障诊断方法[J]. 北京理工大学学报,2021, 41(12): 1245-1252. [9] VERSTRAETE D, FERRADA A, LÓPEZ DROGUETT E, et al. Deep learning enabled fault diagnosis using time-frequency image analysis of rolling element bearings[J]. Shock and Vibration, 2017, 2017: 5067651. [10] ALASKAR H. Deep learning-based model architecture for time-frequency images analysis[J]. International Journal of Advanced Computer Science and Applications, 2018, 9(12): 486-494. [11] CHEN R X, HUANG X, YANG L X, et al. Intelligent fault diagnosis method of planetary gearboxes based on convolution neural network and discrete wavelet transform[J]. Computers in Industry, 2019, 106: 48-59. doi: 10.1016/j.compind.2018.11.003 [12] 郑攀,周建华,高素杰,等. 基于SPS与CNN的行星齿轮箱故障特征提取与诊断研究[J]. 机械传动,2022, 46(4): 73-79,147. [13] 郭庆华. CNN在单帧舰船回转运动图像特征提取中的应用[J]. 舰船科学技术,2021, 43(22): 190-192. [14] JING L Y, ZHAO M, LI P, et al. A convolutional neural network based feature learning and fault diagnosis method for the condition monitoring of gearbox[J]. Measurement, 2017, 111: 1-10. doi: 10.1016/j.measurement.2017.07.017 [15] 郭旭,骆勇鹏,王林堃,等. 基于CNN与DCGAN的结构振动监测传感器故障诊断及监测数据恢复[J]. 铁道科学与工程学报,2022, 19(11): 3383-3395. -
图(6) / 表(3)
计量
- 文章访问数: 22
- HTML全文浏览量: 9
- PDF下载量: 6
- 被引次数: 0
