﻿ RBF神经网络在船舶模拟电路故障诊断中的应用
 舰船科学技术  2024, Vol. 46 Issue (10): 182-185    DOI: 10.3404/j.issn.1672-7649.2024.10.033 PDF
RBF神经网络在船舶模拟电路故障诊断中的应用

Application of RBF neural network in fault diagnosis of ship analog circuits
HUO Yan-fei
Applied Technology College, Dalian Ocean University, Dalian 116300, China
Abstract: A fault diagnosis method for ship analog circuits based on RBF neural network is proposed to address the complex interaction of components in ship analog circuits, the difficulty of highlighting fault signals in a large number of normal signals, and the difficulty of extracting and identifying fault features. The fault feature extraction method for ship simulation circuits based on wavelet packets captures the energy variation characteristics of the circuit frequency band through wavelet decomposition and reconstruction; Using state transition algorithm to optimize the fault diagnosis model of RBF neural network, optimizing the parameters of RBF neural network, constructing an RBF neural network model for diagnosing circuit faults, learning the relationship between extracted fault features and types, and diagnosing the new input ship analog circuit output signal fault type. The experimental test results show that after effectively capturing the energy changes in the frequency band of ship analog circuit faults, this method has not shown significant errors in the diagnosis of various ship analog circuit faults.
Key words: RBF neural network     ship analog circuit     fault diagnosis     state transition algorithm
0 引　言

RBF神经网络具有强大的逼近任意非线性函数的能力，模拟电路中故障往往表现为复杂的非线性特征，而RBF神经网络可以通过学习训练[5]，逼近非线性故障模式，实现对模拟电路故障的有效诊断，且RBF神经网络的参数调整可能更为直观和简单。

1 船舶模拟电路故障诊断方法 1.1 基于小波包的船舶模拟电路故障特征提取方法

 $V_{j,i}^m\left( b \right) = {2^{\frac{j}{2}}}{V_m}\left( {{2^j}b - i} \right)。$ (1)

 ${F_{4j}} = \sum\limits_{j = 1}^m {{{\left| {s_j^m\left( b \right)} \right|}^2}}。$ (3)

 ${F'_4} = {\left( {\sum\limits_{j = 0}^m {{{\left| {{F_{4j}}} \right|}^2}} } \right)^{\frac{1}{2}}} 。$ (4)

 ${d_j} = \frac{{{F_{4j}}}}{{{{F'}_4}}}。$ (5)

 $D = \left\{ {{d_0},{d_1},...,{d_m}} \right\}。$ (6)
1.2 基于状态转移算法优化RBF神经网络的故障诊断模型 1.2.1 用于诊断电路故障的RBF神经网络

 图 1 诊断电路故障的RBF神经网络拓扑结构 Fig. 1 Topological structure of RBF neural network for diagnosing circuit faults

 ${S_j}\left( D \right) = \exp \left[ { - \frac{{{{\left( {D - {o_j}} \right)}^2}}}{{2\beta _j^2}}} \right]。$ (7)

 ${X_i}\left( D \right) = \sum\limits_{i = 1}^q {{\varpi _{ji}}{S_j}\left( D \right)}。$ (8)

 \left\{ {\begin{aligned} & {{G_{t + 1}} = {\phi _t}{G_t} + {\varphi _t}{v_t}}，\\ & {{W_{t + 1}} = f\left( {{G_{t + 1}}} \right)} 。\end{aligned}} \right. (9)

 $G\left( {t + 1} \right) = G\left( t \right) + \tau {\psi _e}G\left( t \right)，$ (10)

 $G\left( {t + 1} \right) = G\left( t \right) + \upsilon {\psi _e}\frac{{G\left( t \right) - G\left( {t - 1} \right)}}{{{{\left\| {G\left( t \right) - G\left( {t - 1} \right)} \right\|}_2}}} 。$ (11)

 $G\left( {t + 1} \right) = {\lambda _n} + \sigma \frac{1}{{n{{\left\| {G\left( t \right)} \right\|}_2}}}G\left( t \right)。$ (12)

 $G\left( {t + 1} \right) = G\left( t \right) + \varsigma {\psi _e}。$ (13)

2 结果分析 2.1 实验内容设计

2.2 舰船模拟电路故障诊断实验数据与分析

 图 2 舰船模拟电路X6故障特征提取结果 Fig. 2 Extraction results of fault characteristics for ship simulation circuit X6

 图 3 舰船模拟电路X0~X6故障诊断结果 Fig. 3 Fault diagnosis results of ship analog circuits X0~X6
3 结　语

 [1] 朱敏, 许爱强, 许晴, 等. 基于改进多层核超限学习机的模拟电路故障诊断[J]. 兵工学报, 2021, 42(2): 356-369. ZHU Min, XU Ai-qiang, XU Qing, et al. Fault diagnosis of analog circuits based on improved multilayer kernel extreme learning machine[J]. Acta Armamentarii, 2021, 42(2): 356-369. [2] 梁志义, 郑伟, 徐庆, 等. 基于近场扫描和相似性度量的电路板故障区域检测[J]. 电子测量与仪器学报, 2023, 37(3): 111-120. LIANG Zhi-yi, ZHENG Wei, XU Qing, et al. Circuit board fault area detection based on near-field scanning and similarity measure[J]. Journal of Electronic Measurement and Instrumentation, 2023, 37(3): 111-120. [3] 杨东儒, 魏建文, 林雄威, 等. 基于自注意力机制的深度学习模拟电路故障诊断[J]. 仪器仪表学报, 2023, 44(3): 128-136. YANG Dong-ru, WEI Jian-wen, LIN Xiong-wei, et al. A fault diagnosis algorithm for analog circuits based on self-attention mechanism deep learning[J]. Chinese Journal of Scientific Instrument, 2023, 44(3): 128-136. [4] CHEN L, KHAN U S, KHATTAK M K, et al. An effective approach based on nonlinear spectrum and improved convolution neural network for analog circuit fault diagnosis[J]. Review of Scientific Instruments, 2023, 94(5): 1-14. [5] 周海勇, 张晓松, 贺振杰, 等. 船舶液压设备双缸同步液压回路设计[J]. 船舶工程, 2021, 43(3): 72-75. ZHOU Hai-yong, ZHANG Xiao-song, HE Zhen-jie, et al. Design of double cylinder synchronous hydraulic circuit of marine hydraulic equipment[J]. Ship Engineering, 2021, 43(3): 72-75. [6] 吴希杰, 周方明. 船舶控制设备用微连接器激光软钎焊电路设计[J]. 船海工程, 2021, 50(1): 64-67. WU Xi-jie, ZHOU Fang-ming. Design of laser soldering circuit for micro connector of marine control equipment[J]. Ship & Ocean Engineering, 2021, 50(1): 64-67.