水面艦艇聲納對來襲魚雷類型識別技術研究

Identifical Technology of Incoming Torpedo for Surface Ship's Sonar

林澄貴 ； 吳柏賢

《海軍學術雙月刊》 48卷6期 (2014/12) Pp. 59-76

https://doi.org/10.6237/NPJ.201412_48(6).0004

摘要

一、艦船與魚雷在水面下的輻射噪音源可分為機械噪音、螺旋槳噪音與水動力噪音，其中螺旋槳噪音所產生的低頻線譜是識別水中目標的重要特徵，在水中傳播時具有較高的穩定性與較小的能量損失，可提高對魚雷目標的偵測距離。魚雷預警系統中的目標識別器僅將被偵測歸類為魚雷或非魚雷兩類型。魚雷航速較快且螺旋槳的軸頻也較高，透過LOFAR與DEMON頻譜分析，提取魚雷目標軸頻就可以有效判斷是否為魚雷。本文亦介紹其他常用的聲紋特徵分析與辨識方法，高階統計量分析、倒傳遞類神經網路以及高斯混合模型等，並建議可採用專家系統來降低對魚雷樣本數量的依賴。因此，未來應加強對低頻弱信號檢測及魚雷目標特徵識別等技術的精進，期以提升魚雷識別器的辨識功能。二、魚雷識別器除了以目標聲紋特徵而確定為魚雷或非魚雷證據外，還需要再以魚雷預警聲納偵測追蹤到的目標運動軌跡以及作戰海洋環境、戰場態勢等證據，做數據融合來判定何種類型的魚雷，經由戰術反應決策模組做戰術評估後，再擇一有效魚雷反制器材對抗來襲魚雷予以軟殺或硬殺。

關鍵字

魚雷目標辨識 ； 魚雷預警 ； 高階統計量 ； 倒傳遞類神經網路 ； 高斯混合模型 ； 作戰海洋環境 ； 戰場態勢 ； 戰術反應決策

並列摘要

1. The noise sources of warship or torpedo under water can be separated into mechanical noise, propeller noise and waterpower noise. The line spectrums in low frequency which are produced by the propeller noise is an important characteristic to identify the underwater target, because the detection distance can be enhanced by low frequency which has lower transmission loss. The target which is torpedo or non-torpedo can be identified and classified according to use torpedo early warning system. The shaft frequency of target is an useful information which can be analyzed by using LOFAR and DEMON to identify the torpedo, because the speed of torpedo is fast and the shaft frequency of propeller is high. In this paper, the higher-order statistics analysis, back-propagation neural network and gaussian mixed model are introduced to identify the torpedo signal, and the quantities of torpedo data can be reduced on torpedo identification by using the professional system. In the further, the low frequency signal detection and the characteristic of torpedo identification must be reinforced to enhance the performance of torpedo identification system. 2. But the torpedo identification system just can identify the target signal which is torpedo or non- torpedo, the type of torpedo is decided by using torpedo alert sonar which can detect the locus of target, battle ocean sourrounding and battlefield state. The result is used to choose an effective tactics to resist the torpedo after the tactical reaction decision.

並列關鍵字

Torpedo Identification ； Torpedo Alert ； Higher-Order Statistics ； Back Propagation Artifical Neural Network ； Gaussian Mixed Model ； Battle Ocean Sourrounding ； Battlefield State ； Tactical Reaction Decision