應用匹配濾波器與聲學經驗公式模擬偵測分析船舶運行低頻噪音

Abstract

本研究以2022年彰芳外海離岸風場施工期間的三艘工作船（電纜鋪設船Maersk Connector、離岸工作支援船Onyx、多功能鋪纜船Topaz Installer）為對象，利用水下收音機陣列於點位UN1蒐集長達一個月的水下聲音資料，並透過自動識別系統AIS（Automatic Identification System）定位船舶實際通過時間與距離。研究方法包含四大步驟：首先，進行音檔的頻譜圖分析與特徵區段擷取；其次，透過希爾伯特轉換進行包絡解調，強化螺旋槳通過頻率（Blade-Pass Frequency, BPF）訊號的辨識度；再者，設計帶通濾波器過濾背景噪音並正規化訊號；最後以修改後的模擬音源作為模板，進行匹配濾波偵測，評估其在不同距離與背景環境下的有效性。 聲源模擬部分，參考Wittekind及Takinacı文獻所提經驗公式，結合線性相位、傅立葉調變與振幅調變技術，成功生成具時間特徵之模擬聲音訊號。透過匹配濾波器進行實測比對後發現，10公里內為穩定高效的偵測距離區，超過20公里後偵測性能逐漸下降，30公里則接近失效門檻。此外，同一艘船於日夜不同背景噪音條件下的偵測表現亦有顯著差異。 本研究整合聲學模擬與AIS輔助資訊，並透過實地量測資料驗證模擬結果，不僅展現匹配濾波器於水下聲紋偵測的高準確性，也提供日後海洋噪音監測系統設計與風場環評中重要的技術參考與基礎資料。

Three work vessels—Maersk Connector, Onyx, and Topaz Installer—operating near the Changfang offshore wind farm during construction in 2022 were selected as case vessels. One-month continuous underwater acoustic recordings were obtained using the position UN1 hydrophone array, while AIS data provided precise passage times and distances. The methodology comprises four main steps: spectrogram analysis and extraction of characteristic segments; envelope demodulation via the Hilbert transform to enhance blade-pass frequency （BPF） recognition; design of a band-pass filter to suppress background noise and normalize signal amplitude; and matched-filter detection using the processed simulated source as the template, with performance evaluated across varying ranges and ambient conditions. Source simulation was based on empirical formulas proposed by Wittekind and Takinacı, combining linear-phase filtering, Fourier-based modulation, and amplitude modulation to create time-characterized synthetic signals. Field validation revealed stable, highly efficient detection within 10 km, progressively declining performance beyond 20 km, and near-threshold detection at 30 km. Significant day-night differences in detection efficacy were also observed. By integrating acoustic simulation with AIS-aided field measurements, this work demonstrates the high accuracy of matched-filter techniques for underwater ship-signal detection and offers critical technical insights for future marine noise monitoring system design and environmental impact assessments in offshore wind projects.