﻿ 基于EMI与ANN技术的水下目标探测系统
 舰船科学技术  2017, Vol. 39 Issue (12): 86-90 PDF

Underwater target detection based on EMI and ANN
ZHANG Jun, LI Jun-nan, ZHANG Ji-ming, ZHAO Yi, LU Jun-feng
Institute of Mechanical Engineering, Anhui University of Science and Technology, Huainan 232001, China
Abstract: The major way to detect the underwater target are acoustic detection and laser detection presently. The underwater detection system based on piezoelectric impedance and artificial neural network was given in this paper, the autonomous underwater vehicle which can detect the special frequency signal was designed, the major structure, motion mode, control mode were analyzed. The detection system based on BP neural network was built, combine the three ball positioning principle, the signal spatial position is more accurate.
Key words: EMI     ANN     target recognition     frequency detection     position detection
0 引　言

1 EMI与ANN模型的建立 1.1 EMI技术

EMI技术是利用压电智能材料的压电特性，通过分析PZT与结构的耦合特性得出耦合结构电阻抗与机械阻抗之间的关系，由测量电阻抗间接测得机械阻抗[4]

1.1.1 机械阻抗
 图 1 单自由度机械系统 Fig. 1 Single freedom mechanical system

 $m\ddot X + C\dot X + KX = f(t)\text{。}$ (1)

 $( - {\omega ^2}m + j\omega C + K)X = F\text{，}$ (2)

$A = {\omega ^2}X$ $V = j\omega X$ ，则由以上3式可得到：

 ${Z_D} = \frac{F}{X} = K - {\omega ^2}m + j\omega C\text{，}$ (3)

 ${Z_v} = \frac{F}{V} = C + j\omega m + \frac{K}{{j\omega }}\text{，}$ (4)
 ${Z_A} = \frac{F}{A} = m - \frac{K}{{{\omega ^2}}} + \frac{C}{{j\omega }}\text{。}$ (5)

1.1.2 机电耦合

 图 2 压电陶瓷与结构的机电耦合模型 Fig. 2 Electromechanical coupling model of PZT-structure

 ${v^2}\frac{{{\partial ^2}u(x,t)}}{{\partial {x^2}}} = \frac{{{\partial ^2}u(x,t)}}{{\partial {t^2}}}\text{。}$ (6)

PZT与构造物的耦合的协调性、平衡性可描述为：

 ${\left. {{T_1}} \right|_{x = {l_{PZT}}}} = - j\omega \frac{{{{\left. {{Z_S}\overline u } \right|}_{x = {l_{PZT}}}}}}{{{w_{PZT}}{h_{PZT}}}}{e^{j\omega t}}\text{。}$ (7)

 $x = \frac{{{d_{31}}\overline {{E_3}} }}{{k\cos (k{l_{PZT}}) + j\omega \frac{{\overline s _{11}^E}}{{{w_{PZT}}{h_{PZT}}}}\sin (k{l_{PZT}})}}\sin (kx){e^{j\omega t}}\text{。}$ (8)

 ${v_{PZT}} = {\dot x_{PZT}} = j\omega \frac{{{d_{31}}\overline {{E_3}} }}{{k\cos (k{l_{PZT}})}}\sin (k{l_{PZT}}){e^{j\omega t}}\text{。}$ (9)

 ${x_f} = {l_{PZT}}{d_{31}}{E_3} = {l_{PZT}}{d_{31}}\overline {{E_3}} {e^{j\omega t}}\text{。}$ (10)

 ${f_{PZT}} = \overline {{K_{PZT}}} {x_f} = \overline {{K_{PZT}}} {l_{PZT}}{d_{31}}\overline {{E_3}} {e^{jwt}}\text{。}$ (11)

 ${Z_{PZT}} = \frac{{{f_{PZT}}}}{{{v_{PZT}}}} = \frac{{\overline {{K_{PZT}}} {l_{PZT}}{d_{31}}\overline {{E_3}} {e^{j\omega t}}}}{{j\omega \frac{{{d_{31}}\overline {{E_3}} }}{{k\cos (k{l_{PZT}})}}\sin (k{l_{PZT}}){e^{j\omega t}}}} = \frac{{\overline {{K_{PZT}}} k{l_{PZT}}}}{{j\omega \tan (k{l_{PZT}})}}\text{。}$ (12)

 ${\left. x \right|_{x = {l_{PZT}}}} = \frac{{{Z_{PZT}}\tan (k{l_{PZT}}){d_{31}}\overline {{E_3}} }}{{({Z_{PZT}} + {Z_S})k}}{e^{j\omega t}}\text{，}$ (13)

 $\overline {{D_3}} = \frac{{{Z_{PZT}}d_{31}^2\overline c _{11}^E}}{{{{\rm{Z}}_{{\rm{PZT}}}} + {Z_{PZT}}}}\frac{{\cos (kx)}}{{\cos (k{l_{PZT}})}} + (\overline \varepsilon _{33}^{\rm T} - d_{31}^2\overline c _{11}^E)\overline {{E_3}} \text{。}$ (14)

$I = \bar I{e^{j\omega t}}$ ，则有电流为：

 $\begin{split}&\overline I = \iint {\overline {{D_3}} dxdy = j\omega \overline {{E_3}} {w_{PZT}}{l_{PZT}}} (\frac{{d_{31}^2\overline c _{11}^E{Z_{PZT}}}}{{{Z_S} + {Z_{PZT}}}}\frac{{\tan (k{l_{PZT}})}}{{k{l_{PZT}}}} +\\ &\;\;\;\;\;\;\;\;\;\;\;\;\overline \varepsilon _{33}^T - d_{31}^2\overline c _{11}^E)\text{。}\end{split}$ (15)

 ${Z_{PZT}} = - j\frac{{{h_{PZT}}}}{{\omega {w_{PZT}}{l_{PZT}}}}{[\overline \varepsilon _{33}^{\rm T} - \frac{{{Z_{PZT}}}}{{{{\rm{Z}}_{{\rm{PZT}}}} + {Z_{PZT}}}}d_{31}^2\overline c _{11}^E]^{ - 1}}\text{。}$ (16)
1.2 ANN技术

ANN技术即人工神经网络技术，神经网络是一种使用数目庞大的简单计算单元间的相互连接，这些简单计算单元就是“神经元”或者“处理单元”。

 图 3 神经元模型 Fig. 3 Neural model

BP神经网络（Back Propagation）是一种按照逆向传播算法训练的多层前馈神经网络，具有非线性特性、大量并行分布结构以及学习和归纳能力，对于无法用明确的模型表示的数据是一种很好的处理方法，可以实现对于目标位置的识别。该神经网络结构有2个显著特点：1）信号向前传播；2）误差反向传播。

2 基于EMI和ANN的水下目标探测系统

2.1 水下潜航器结构

 图 4 潜航器结构图 Fig. 4 Autonomous underwater vehicle structure

2.2 水下目标探测原理

 图 5 某物体声振FRF结果和压电阻抗法所得阻抗图形 Fig. 5 The FRF result of vibration noise and the resistance result of piezoelectric impedance
2.3 基于EMI和ANN的水下目标探测实验

1）将潜航器b、潜航器c、潜航器d上的阻抗分析仪的扫频范围设定为39.9～40.1 kHz，将40 kHz的水下移动潜体a放入目标海域，通过测量船基站向潜航器发出扫频指令，扫频后把数据传回测量船基站进行分析，当潜航器都能扫频到稳定的40 kHz的水下移动潜体a时，说明潜航器b，c，d都进入到水下移动潜体a所在附近海域；

2）基于ANN的水下目标识别数据库

 图 6 探测超声信标位置神经网络分析图 Fig. 6 The analysis chart of underwater target detection

3）确定水下移动潜体a的位置

 图 7 空间三球定位原理 Fig. 7 Positioning principle of the three ball

 ${\left( {x - {x_1}} \right)^2} + {\left( {y - {y_1}} \right)^2} + {\left( {z - {z_0}} \right)^2} = {L_1}^2\text{，}$ (17)
 ${\left( {x - {x_2}} \right)^2} + {\left( {y - {y_2}} \right)^2} + {\left( {z - {z_0}} \right)^2} = {L_2}^2\text{，}$ (18)
 ${\left( {x - {x_3}} \right)^2} + {\left( {y - {y_3}} \right)^2} + {\left( {z - {z_0}} \right)^2} = {L_3}^2\text{。}$ (19)

t时刻的交点 $P\left( {x,y,z} \right)$ ，即水下移动潜体a的位置坐标为：

 $x = \frac{{(L_2^2 - L_1^2)({y_1} - {y_3}) - (L_3^2 - L_1^2)({y_1} - {y_2})}}{{2\left[ {({x_1} - {x_2})({y_1} - {y_3}) - ({x_1} - {x_3})({y_1} - {y_2})} \right]}}\text{，}$ (20)
 $y = \frac{{(L_2^2 - L_1^2)({x_1} - {x_3}) - (L_3^2 - L_1^2)({x_1} - {x_2})}}{{2\left[ {({x_1} - {x_3})({y_1} - {y_2}) - ({x_1} - {x_2})({y_1} - {y_3})} \right]}}\text{，}$ (21)
 $z = - \sqrt {{\rm{L}}_1^2 - {{\left( {{\rm{x}} - {{\rm{x}}_1}} \right)}^2} - {{\left( {{\rm{y}} - {{\rm{y}}_1}} \right)}^2}} + {z_0}\text{。}$ (22)

t2时刻的水下移动潜体a的位置坐标 ${P_2}\left( {x,y,z} \right)$

t3时刻的水下移动潜体a的位置坐标 ${P_3}\left( {x,y,z} \right)$

tn的水下移动潜体a的位置坐标 ${P_n}\left( {x,y,z} \right)$

3 结　语

1）基于压电阻抗法探测声波信号时，信号源与探测器间的距离与实验所测得阻抗值直接相关，距离越小，阻抗值越大；信号源的发射频率与探测器检测到的频率一致。该实验直接表明使用压电阻抗法在水下检测发射固定频率声波信号的目标可行。

2）使用BP神经网络训练完成后的网络在验证样本测试下基本与训练样本保持一致；结合三球定位原理，使用BP神经网络建立的用于定位探测到的水下固频信号的神经网络系统，实现了对目标信号空间位置更精确的定位。

 [1] 赵安邦, 董海防, 孙国仓. 一种应用于水下目标探测的激光声系统[J]. 声学技术, 2009, 28 (1): 14–16. ZHAO An-bang, DONG Hai-Fang, SUN Guo-Cang. A laser-acoustic system for underwater target detection[J]. TECHNICAL ACOUSTICS, 2009, 28 (1): 14–16. [2] 林伟, 苑秉成, 张坚. 基于滤波与BP神经网络的水下探测信号检测算法[J]. 海军工程大学学报, 2012 (03): 57–60. LIN Wei, FAN Bing-Cheng, ZHANG Jian. Signal detection algorithm of underwater detection based on band-pass filter and BP neural network[J]. Journal of Naval University of Engineering, 2012 (03): 57–60. [3] 孙岚, 李厚朴, 边少锋, 李宏武, 周超烨, 等. 基于重力梯度的潜艇探测方法研究[J]. 海洋测绘, 2010 (02): 24–27+45. SUN Lan, LI Hou-Pu, BIAN Shao-Feng, LI Hong-Wei, Zhou Chao-Ye, et al. Study on the Submarine Detection Method Based on the Gravity Gradient[J]. HYDROGRAPHIC SURVEYING AND CHARTING, 2010 (02): 24–27+45. DOI: 10.3969/j.issn.1671-3044.2010.02.007 [4] 陶娟. 基于EMI与ANN技术的构造物健康诊断研究[D]. 淮南: 安徽理工大学, 2013. [5] 田军, 余亚军, 蔡龙. 针对持续性跟踪无人艇的探测技术[J]. 舰船科学技术, 2013, 35 (4): 98–101. TIAN Jun, YU Ya-Long, CAI Long. Research on detection technology on ACTUV[J]. Ship Science and Technology, 2013, 35 (4): 98–101. [6] 苏镇. 一种基于超声波的潜艇数据传输系统[J]. 舰船电子工程, 2012, 32 (12): 142–143. SU Zhen. Data Transmission System for Submarine Based on Ultrasonic Waves[J]. Ship Electronic Engineering, 2012, 32 (12): 142–143. DOI: 10.3969/j.issn.1627-9730.2012.12.049 [7] 王永浩, 陆铭华, 林晓峰, 章阳, 高劼, 等. 基于水声探测条件下潜射反舰导弹攻击中远程目标可行性研究[J]. 飞航导弹, 2013 (2): 49–51. WANG Yong-Hao, LU Ming-Hua, LIN Xiao-Feng, ZHANG Yang, GAO Jie, et al. Feasibility study of underwater launched anti-ship missile attacking medium and long-range targets based on underwater acoustic detection condition[J]. Aerodynamic Missile Journal, 2013 (2): 49–51. [8] 郭榜增. 基于EMI与ANN技术的声波信号探测机理研究[D]. 淮南: 安徽理工大学, 2015. [9] 汪天博. 基于EMI与ANN技术水下固定频率信号探测研究[D]. 淮南: 安徽理工大学, 2015.