此篇研究，我們發展一套針對於風力發電機葉片的自動化檢測設備。藉由標準麥克風(1/4”130D20, PCB Piezo-tronics)擷取正常健康狀態下風機葉片運轉時產生的風噪訊號，我們建立了三種風速下的標準線。而後改由低成本微機電麥克風進行收音，確認其可行性後，藉由鋪設高防水防塵係數之聲學篩網/膜以抵抗惡劣環境條件及減少風噪突波的影響。確認擷取系統架構後，我們將演算法從PC環境執行MATLAB®重構為嵌入式系統版本，可於ADLINK Technology®所生產之獨立式乙太網資料擷取器(MCM-204)執行。藉由短時傅立葉轉換(STFT)分析、對照與健康狀況下的時頻圖，我們得到以下結果：1.葉片受損下所得到之時頻圖與健康狀態下比較，在時序上有明顯的強度差別；2.利用微機電麥克風與MCM-204所組成之自動化系統架構與標準麥克風系統架構擷取之訊號在MATLAB®上分析，我們得到相似且一致的結果。藉由上述兩個結果，對於整體未來自動化風機葉片健康度檢測上，我們提出了成本相對低廉與掛載式系統架構，將可在未來進行全自動化檢測。

In this study, we built up a total solution for blade surface diagnosis on a wind turbine. By capturing sound/noise from healthy blades through a conventional condenser microphone (1/4”130D20, PCB Piezotronics), we constructed standard characteristic curves for three wind speed intervals. Then, we used a MEMS microphone to replace the condenser microphone for signal capturing. After verifying the functionality of the microphone, the device was covered with membrane/mesh to eliminate wind noise/throb during sound recording and provide protection from dust and water. The membranes/meshes were specially chosen with high IP rating to resist the severe environment. Thereafter, we constructed an automatic diagnosis system ported from PC with MATLAB software to single ADLINK Technology® MCM-204 standalone Ethernet DAQ device. Analyzing and comparing the time-frequency diagram through Short-Time Fourier Transform (STFT) method and a self-developed algorithm on MATLAB software shows the intensity changing in the time-frequency diagram compared to the standard data (recorded by condenser microphone), whereas the spectrogram from the data recorded by MEMS microphone shows a similar characteristic pattern. These results show the possibility of constructing an unmanned, automatic analyzing/monitoring system for turbine monitoring with low component cost.