﻿ 基于模糊层次分析的舰船故障数据定位挖掘算法
 舰船科学技术  2022, Vol. 44 Issue (15): 149-152    DOI: 10.3404/j.issn.1672-7649.2022.15.031 PDF

1. 西南大学 计算机与信息科学学院，重庆 400715;
2. 重庆对外经贸学院大数据与智能工程学院，重庆 401520

Mining algorithm of ship fault data location based on fuzzy analytic hierarchy process
LI Chuan1,2
1. School of Computer and Information Science, Southwest University, Chongqing 400715, China;
2. School of Big Data and Intelligent Engineering, Chongqing University of International Business and Economics, Chongqing 401520, China
Abstract: An algorithm of ship fault data location and mining based on fuzzy analytic hierarchy process is proposed to solve the problem of ship fault data location and mining. After the fault state of the ship system is diagnosed by using the quantitative method of the ship system state based on fuzzy analytic hierarchy process, the dimension of the fault data is constrained in the same dimension by the dimension reduction method of the fault data based on the principal component analysis method, so as to improve the operability of the fault data. After the dimensionality reduction, the ship fault data is input into the deep learning network to realize the fault data type identification, so as to complete the fault data location mining. The experimental results show that after the application of the proposed algorithm, the dimensions of ship fault data are consistent after dimensionality reduction, the recognition of ship fault state is accurate, and the fault data location mining results are in line with the reality. The application performance can meet the needs of ship system operation and maintenance, and can be used in the task of ship fault data location mining.
Key words: fuzzy analytic hierarchy process     warship     fault     data     positioning excavation     data dimensionality reduction
0 引　言

1 舰船故障数据定位挖掘算法 1.1 基于模糊层次分析的舰船系统状态量化方法

1)分析舰船系统运行状态量化因素的重要性

2)设计判断矩阵。假如所有舰船系统运行状态量化因素的集合是 $Y = \left\{ {{y_1},{y_2},\cdots,{y_m}} \right\}$ ，结合表1标准，将全部因素执行两两对比，设计舰船系统运行状态的判断矩阵 $D$

3)运算权重 ${\beta _i}$ ，使用判断矩阵 $D$ ，运算其特征根最大值 ${\delta _{\max }}$ ，然后运算 $D$ 针对 ${\delta _{\max }}$ 的特征向量 $\varepsilon = ( {a_1}, {a_2},\cdots,{a_m} )$ ，并将其执行归一化处理，所获取 ${a_j}$ 就是每个舰船系统运行状态量化因素的权值。

 ${D_R} = {D_l}/{S_I}。$ (1)

 ${A} = g\left( BS \right)$ (2)

1.2 基于主元分析方法的故障数据降维方法

 $F = \sum\limits_{j = 1}^m {y_j^{\rm{T}}{y_j}}。$ (3)

 $\eta W = FW ，$ (4)

 ${Y_p} = {y_j}\eta W 。$ (5)

1.3 基于深度学习的舰船故障数据定位挖掘模型

 $\left\{ {\begin{array}{*{20}{l}} {{X_j} = g\left( {{Y_j}} \right)}，\\ {{Y_j} = \left[ {{y_1},{y_2},\cdots,{y_n}} \right]} ，\\ {{X_j} = \left[ {{x_1},{x_2},\cdots,{x_m}} \right]} 。\end{array}} \right.$ (6)

 图 1 深度学习网络结构 Fig. 1 Deep learning network structure

 ${k_t} = \vartheta \left( {{Y_t} + {k_{t - 1}} + c} \right) 。$ (7)

$t$ 时间段中隐藏层的输出 ${\sigma _t}$ 和最后预测输出Xt为：

 ${\sigma _t} = {k_t} + b ，$ (8)
 ${X_t} = \vartheta {\sigma _t} ，$ (9)

$t$ 时间段第 $j \in M$ 个故障数据样本的目标输出xt，预测输出Xt之间差值设成损失函数 ${\xi _t}$ ，实现模型的量化处理。则

 ${\xi _t} = \sum\limits_{j = 1}^M {{{\left( {X_t^j - x_t^j} \right)}^2}} 。$ (10)

1)将降维后的舰船系统故障数据集合Yj输入深度学习网络；

2)深度学习网络将降维后的舰船系统故障数据集的故障种类编码后，训练基于深度学习的舰船故障数据定位挖掘模型，训练精度符合需求后，进入舰船系统故障数据定位挖掘的测试阶段，识别故障数据类型，便可定位故障数据来源，以此得到舰船故障数据定位挖掘结果

2 实验结果与分析

 图 2 舰船电力系统结构示意图 Fig. 2 Structure diagram of ship power system

 图 3 降维前故障数据维度信息 Fig. 3 Dimension information of fault data before dimension reduction

 图 4 降维后故障数据维度信息 Fig. 4 Dimension information of fault data after dimension reduction

 图 5 舰船电站故障数据定位挖掘结果 Fig. 5 Mining results of fault data location of ship power station
3 结　语

 [1] 苏文明, 柴凯. 基于VMD和RBF的舰船管路泄漏识别和定位[J]. 船舶工程, 2020, 42(10): 105-112. SU Wen-ming, CHAI Kai. Warship Pipeline Leakage Identification and Location Based on VMD and RBF[J]. Ship Engineering, 2020, 42(10): 105-112. DOI:10.13788/j.cnki.cbgc.2020.10.19 [2] 邹永久, 杜太利, 姜兴家, 等. 基于AEC模型的船舶燃油供给系统状态评估[J]. 中国舰船研究, 2021, 16(1): 175-180. ZOU Yong-jiu, DU Tai-li, JIANG Xing-jia, et al. Evaluation of marine fuel supply system state based on AEC model[J]. Chinese Journal of Ship Research, 2021, 16(1): 175-180. DOI:10.19693/j.issn.1673-3185.01779 [3] 仲国强, 贾宝柱, 肖峰, 等. 基于深度信念网络的船舶柴油机智能故障诊断[J]. 中国舰船研究, 2020, 15(3): 136-142+184. ZHONG Guo-qiang, JIA Bao-zhu, XIAO Feng, et al. Intelligent fault diagnosis of marine diesel engine based on deep belief network[J]. Chinese Journal of Ship Research, 2020, 15(3): 136-142+184. [4] 李军, 刘杰, 杨梓辉. 基于三角模糊层次分析法的高原库区施工船舶安全预警系统的设计[J]. 上海海事大学学报, 2021, 42(1): 94-99+106. [5] 边荣正, 张鉴, 周亮, 等. 面向复杂多流形高维数据的t-SNE降维方法[J]. 计算机辅助设计与图形学学报, 2021, 33(11): 1746-1754. [6] 常书源, 赵荣珍, 陈博, 等. 基于边界判别多流形分析的故障数据集降维方法[J]. 振动与冲击, 2021, 40(23): 120-126. [7] 魏世超, 李歆, 张宜弛, 等. 基于E-t-SNE的混合属性数据降维可视化方法[J]. 计算机工程与应用, 2020, 56(6): 66-72. [8] 江粼, 房小兆, 滕少华. 基于全局-局部保持投影的稀疏降维方法[J]. 江西师范大学学报(自然科学版), 2021, 45(1): 46-54. [9] 李海林, 梁叶. 基于关键形态特征的多元时间序列降维方法[J]. 控制与决策, 2020, 35(3): 629-636. [10] 刘文博, 梁盛楠, 余泉, 等. 基于伪F统计量的属性特征降维方法研究[J]. 东北师大学报(自然科学版), 2020, 52(1): 43-49. [11] 姜帅全, 朱志宇. 基于单频法的船舶电网绝缘监测与故障定位系统[J]. 船舶工程, 2021, 43(2): 95-102+157. [12] 陶珩, 魏海峰, 张懿, 等. 环网-辐射形船舶混合配电网络的Levy风驱动故障定位[J]. 江苏科技大学学报(自然科学版), 2020, 34(1): 54-61. TAO Heng, WEI Hai-feng, ZHANG Yi, et al. Levy wind drive fault diagnosis of ring network-radial ship hybrid distribution network[J]. Journal of Jiangsu University of Science and Technology (Natural Science Edition), 2020, 34(1): 54-61.