Research on Fault Prediction of Reactor Power Measurement Circuit Based on Relevance Vector Machine
-
摘要: 为了提高核测量装置的保障性和维修性，本文以反应堆功率测量放大电路为对象，通过基于量子粒子群优化算法的多核相关向量机模型对电路的典型故障进行预测。从功率测量放大电路的脉冲响应信号中，用小波包分解方法提取特征信息，将计算所得到的特征与电路正常状态特征之间的欧氏距离作为电路故障程度指标，选用多核相关向量机建立电路故障预测模型，并分析了相关向量机模型核函数种类、参数优化算法对于模型预测效果的影响，研究结果表明采用量子粒子群算法优化的多核相关向量机模型对于电路未来运行状态的预测精度较优，能够准确预测电路故障程度的变化规律。Abstract: In order to improve the supportability and maintainability of the nuclear measuring device, taking the reactor power measurement amplifier circuit as the object, this paper predicts the typical faults of the circuit through the multi-kernel relevance vector machine model based on quantum particle swarm optimization algorithm. From the pulse response signal of the power measurement amplifier circuit, the feature information is extracted by the wavelet packet decomposition method, and the Euclidean distance between the feature and the normal state feature of the circuit is calculated as the fault degree indicator of the circuit. Multi-kernel relevance vector machine is selected to establish circuit fault prediction model. The influence of kernel function type and parameter optimization algorithm of relevance vector machine model on the prediction effect of the model is analyzed. The research results show that the multi-kernel relevance vector machine model optimized by quantum particle swarm algorithm has better prediction accuracy for the future running state of the circuit, and can accurately predict the changing law of the fault degree of the circuit.
-
Key words:
- Wavelet packet /
- Feature extraction /
- RVM /
- QPSO /
- Fault prediction
-
图 4 多核RVM模型在训练集上的输出结果
Figure 4. Output Results of Multi-kernel RVM Models on the Training Set
图 5 多核RVM模型在验证集上的输出结果
Figure 5. Output Results of Multi-kernel RVM Models on the Validation Set
表 1 关键元件故障阈值
Table 1. Fault Threshold of Key Component
元件名 标称值 容差/% 故障阈值 R3 10 kΩ 5 15 kΩ R7 10 kΩ 5 15 kΩ C3 1 µF 5 0.5 µF C4 1 µF 5 0.5 µF 
下载: 导出CSV
表 2 关键元件故障阈值特征距离
Table 2. Feature Distance of Fault Threshold of Key Component
元件名 故障阈值 特征距离阈值 R3 15 kΩ 1609 R7 15 kΩ 1142 C3 0.5 µF 71.7 C4 0.5 µF 343.2
下载: 导出CSV
表 3 多核RVM预测绝对误差
Table 3. Absolute Errors of Multi-kernel RVM Prediction
元件名 特征距离阈值 特征距离阈值绝对误差 PSO QPSO R3 1609 68.5929 27.8732 R7 1142 88.8613 25.0870 C3 343.2 7.7569 7.7914 C4 71.7 3.8918 1.5365
下载: 导出CSV
-
[1] ZHAO S, CHEN S W, YANG F, et al. A composite failure precursor for condition monitoring and remaining useful life prediction of discrete power devices[J]. IEEE Transactions on Industrial Informatics, 2021, 17(1): 688-698. doi: 10.1109/TII.2020.2991454 [2] 吕克洪,程先哲,李华康,等. 电子设备故障预测与健康管理技术发展新动态[J]. 航空学报,2019, 40(11): 13-24. [3] 孟晨,杨华晖,王成,等. 数据驱动的武器系统电子元部件级故障诊断研究综述[J]. 系统工程与电子技术,2021, 43(2): 574-583. doi: 10.12305/j.issn.1001-506X.2021.02.33 [4] LONG Y W, LUO H W, ZHI Y, et al. Remaining useful life estimation of solder joints using an ARMA model optimized by genetic algorithm[C]//2018 19th International Conference on Electronic Packaging Technology. Shanghai: IEEE, 2018: 1108-1111. [5] 郝春燕,魏海坤,张侃健,等. 基于神经网络的模拟电路故障诊断方法[J]. 工业控制计算机,2018, 31(6): 63-65. doi: 10.3969/j.issn.1001-182X.2018.06.028 [6] 闫理跃,王厚军,刘震. 一种结合噪声辅助技术和现场数据的模拟电路实时可靠度预测方法[J]. 兵工学报,2019, 40(2): 326-333. [7] RATHNAPRIYA S, MANIKANDAN V. Remaining useful life prediction of analog circuit using improved unscented particle filter[J]. Journal of Electronic Testing, 2020, 36(2): 169-181. doi: 10.1007/s10836-020-05870-9 [8] TIPPING M E, SMOLA A. Sparse Bayesian learning and the relevance vector machine[J]. Journal of Machine Learning Research, 2001, 1(3): 211-244. [9] BISHOP C M. Pattern recognition and machine learning[M]. New York: Springer, 2006: 345-346. [10] WANG J J, FENG G J, YI C, et al. Electronic circuit fault diagnosis based on an ACO-RVM classifier with a new nonlinearly mixed kernel function[C]//2019 Prognostics and System Health Management Conference. Paris: IEEE, 2019: 332-337. [11] SUN J, XU W B, FENG B. A global search strategy of quantum-behaved particle swarm optimization[C]//IEEE Conference on Cybernetics and Intelligent Systems, 2004. Singapore: IEEE, 2004: 111-116. [12] PRISACARU A, GROMALA P, HAN B, et al. Degradation estimation and prediction of electronic packages using data-driven approach[J]. IEEE Transactions on Industrial Electronics, 2022, 69(3): 2996-3006. doi: 10.1109/TIE.2021.3068681 [13] HANNAMAN G W, WILKINSON C D. Evaluating the effects of aging on electronic instrument and control circuit boards and components in nuclear power plants: 1011709[R]. Washington, DC: U. S. Department of Energy, 2005. [14] LIU Z B, LIU T M, HAN J W, et al. Signal model-based fault coding for diagnostics and prognostics of analog electronic circuits[J]. IEEE Transactions on Industrial Electronics, 2017, 64(1): 605-614. doi: 10.1109/TIE.2016.2599142 [15] 刘立芳,杨海霞,齐小刚. 基于线性判别分析的时频域特征提取算法[J]. 系统工程与电子技术,2019, 41(10): 2184-2190. doi: 10.3969/j.issn.1001-506X.2019.10.05 [16] 时艳玲,刘子鹏,张学良,等. 基于EMD能量占比的海面漂浮小目标特征检测[J]. 系统工程与电子技术,2021, 43(2): 300-310. doi: 10.12305/j.issn.1001-506X.2021.02.03 [17] GROSSMANN A, MORLET J. Decomposition of hardy functions into square integrable wavelets of constant shape[J]. SIAM Journal on Mathematical Analysis, 1984, 15(4): 723-736. doi: 10.1137/0515056 [18] 肖迎群,何怡刚. 基于小波分形分析和脊波网络的模拟电路故障诊断方法[J]. 电工技术学报,2011, 26(11): 105-114. -
图(6) / 表(3)
计量
- 文章访问数: 190
- HTML全文浏览量: 46
- PDF下载量: 28
- 被引次数: 0
