﻿ 基于大数据的舰船故障趋势估计方法
 舰船科学技术  2022, Vol. 44 Issue (16): 159-162    DOI: 10.3404/j.issn.1672-7649.2022.16.034 PDF

Research on ship fault trend estimation method based on big data
JIANG Yu-ting
Information Engineering Institute, Jiangsu Maritime Institute, Nanjing 211170, China
Abstract: The ship fault trend estimation method based on big data is studied to improve the effect of fault trend estimation. Using the improved attention mechanism, the multivariable time series of the historical operation data of each ship equipment is extracted. In the autoregressive prediction model, the multivariable time series is input and the ship fault trend estimation is output. In the wavelet neural network, the multivariable time series is input, the ship fault trend estimation value is output, and the parameters of the wavelet neural network are adjusted by the gradient correction method to reduce the fault trend estimation error. The average value of the estimated fault trend is calculated by the autoregressive prediction model and the wavelet neural network, which is used as the final fault trend estimation result. Experiments show that this method can accurately estimate the ship fault trend. When different faults occur, the method can still accurately estimate the fault trend. The red pool information criterion value of ship equipment fault trend estimation is low, which has better fault trend estimation effect.
Key words: big data     ship failure     trend estimation     attention mechanism     autoregressive prediction     the neural network
0 引　言

1 舰船故障趋势估计方法

1.1 舰船历史数据的注意力处理

1） 计算t时刻，第i个舰船设备历史运行数据时间序列的重要性 $z_t^i$ ，公式如下：

 $z_t^i = v_m^{\rm{T}}\tanh \left[ {\left( {{W_m}{h_{t - 1}};{W_m}{s_{t - 1}}} \right) + {Q_m}{x^i}} \right]。$ (1)

2）通过 $z_t^i$ 获取舰船设备历史运行数据的多变量时间序列 $\hat X = \left\{ {{{\hat x}_1},{{\hat x}_2}, \cdots ,{{\hat x}_T}} \right\}$ ，公式如下：

 $\begin{split} {\hat x_t} =& \Big( soft\max \left( {z_t^1} \right)x_t^1,soft\max \left( {z_t^2} \right)x_t^2, \cdots ,\\ & soft\max \left( {z_t^n} \right)x_t^n \Big)^{\rm{T}}。\end{split}$ (2)
1.2 基于自回归预测的舰船故障趋势估计方法

p阶自回归预测模型 $AR\left( p \right)$ 的定义为：

 $AR\left( p \right) = {u_0} + {u_1}{\hat x_{t - 1}} + {u_2}{\hat x_{t - 2}} + \cdots + {u_p}{\hat x_{t - p}} + {\varepsilon _t}。$ (3)

t-1时刻， ${\hat x_t}$ 的估计值是 ${x'_t}$ ，公式如下：

 ${x'_t} = {u_0} + {u_1}{\hat x_{t - 1}} + {u_2}{\hat x_{t - 2}} + \cdots + {u_p}{\hat x_{t - p}} + {\varepsilon _t}。$ (4)

 ${\boldsymbol{{{X}}}}' = {\boldsymbol{{{A}}}}{\boldsymbol{\varPhi}} + R。$ (5)

 $\hat \varPhi = \frac{{{A^{{\rm{T}}'}}X'}}{{{A^{{\rm{T}}'}}A}}。$ (6)

1.3 基于小波神经网络的舰船故障趋势估计方法

 ${f_{\hat x}}\left( {a,\tau } \right) = \frac{{\displaystyle\int_{ - \infty }^\infty {{{\hat x}_t}\frac{{\phi t - \phi \tau }}{a}{\rm{d}}t} }}{{\sqrt a }}。$ (7)

 $g\left( j \right) = {r_j}\left( {\frac{{\displaystyle\sum\limits_{t = 1}^T {\sum\limits_{j = 1}^N {{\omega _{tj}}{{\hat x}_t} - {b_j}} } }}{{{q_j}}}} \right),j = 1,2, \cdots ,N$ (8)

 $y\left( k \right) = \sum\limits_{j = 1}^N {\sum\limits_{k = 1}^M {{\omega _{jk}}g\left( j \right)} } ,k = 1,2, \cdots ,M 。$ (9)

 $E = \sum\limits_{k = 1}^M {\hat y\left( k \right) - y\left( k \right)}。$ (10)

 $\omega \left( {j + 1} \right) = \omega \left( j \right) - \eta \frac{{\partial E}}{{\partial \omega \left( j \right)}}，$ (11)
 $q\left( {j + 1} \right) = q\left( j \right) - \eta \frac{{\partial E}}{{\partial q\left( j \right)}}，$ (12)
 $b\left( {j + 1} \right) = b\left( j \right) - \eta \frac{{\partial E}}{{\partial b\left( j \right)}}。$ (13)

 $O = \frac{{X' + Y}}{2}。$ (14)
2 实验结果分析

 图 1 舰船冷却水出口温度趋势估计效果 Fig. 1 Effect of temperature trend estimation of warship cooling water outlet

 图 2 舰船主柴油机故障趋势估计结果 Fig. 2 Fault trend estimation results of ship main diesel engine

3 结　语

 [1] 曾友渝, 谢强. 基于改进RNN和VAR的船舶设备故障预测方法[J]. 计算机科学, 2021, 48(6): 184-189. ZENG You-yu, XIE Qiang. Fault prediction method based on improved rnn and var for ship equipment[J]. Computer Science, 2021, 48(6): 184-189. DOI:10.11896/jsjkx.200700117 [2] 蒋佳炜, 胡以怀, 方云虎, 等. 船舶动力装置智能故障诊断技术的应用与展望[J]. 中国舰船研究, 2020, 15(1): 56-67. JIANG Jia-wei, HU Yi-huai, FANG Yun-hu, et al. Application and prospects of intelligent fault diagnosis technology for marine power system[J]. Chinese Journal of Ship Research, 2020, 15(1): 56-67. [3] 贾宝柱, 贾志涛, 余培文, 等. 数据驱动的船舶智能故障诊断方法[J]. 控制工程, 2019, 26(10): 1892-1898. JIA Bao-zhu, JIA Zhi-tao, YU Pei-wen, et al. Data-driven vessel smart fault diagnosis method[J]. Control Engineering of China, 2019, 26(10): 1892-1898. [4] 吴建波, 王春艳, 洪华军, 等. 基于极限学习机的船舶柴油机故障诊断[J]. 计算机工程与应用, 2019, 55(15): 147-152. WU Jian-bo, WANG Chun-yan, HONG Hua-jun, et al. Fault diagnosis of marine diesel engine based on extreme learning machine[J]. Computer Engineering and Applications, 2019, 55(15): 147-152. [5] 韩敏, 李锦冰, 许美玲, 等. 具有工作状态转换的EIIKF船舶柴油机故障预测[J]. 自动化学报, 2019, 45(5): 920-926. HAN Min, LI Jin-bing, XU Mei-ling, et al. Fault prognosis of marine diesel engine with working state transition based on EIIKF[J]. Acta Automatica Sinica, 2019, 45(5): 920-926. [6] 尚前明, 姜苗, 陈辉, 等. 基于PSO-ELM算法实现船舶发电机组故障识别[J]. 船舶工程, 2021, 43(1): 87-94. SHANG Qian-ming, JIANG Miao, CHEN Hui, et al. Fault identification of marine generator set based on PSO-ELM algorithm[J]. Ship Engineering, 2021, 43(1): 87-94. [7] 刘国强, 林叶锦, 张志政, 等. 基于粗糙集和优化DAG-SVM的船舶主机故障诊断方法[J]. 中国舰船研究, 2020, 15(1): 68-73. LIU Guo-qiang, LIN Ye-jin, ZHANG Zhi-zheng, et al. Main marine engine fault diagnosis method based on rough set theory and optimized DAG-SVM[J]. Chinese Journal of Ship Research, 2020, 15(1): 68-73.