本論文主旨為發展並實現平滑維勒、Choi-Williams分佈與錐形時頻分佈等時頻分析演算法，以改善短時間傅立葉變換無法同時具有優良時間及頻率解析度的缺點，並能減少維勒分佈於時頻面上的干涉鬼影。研究中分別以模擬訊號探討個別時頻分析演算法的特性及優缺點；運用時頻分析法能夠精確解析時變或暫態非穩態訊號分量的特性，應用於傳動元件試驗平台的動態訊號分析上，以計算傳動元件的特徵轉速與時頻分析結果對照的方法，可得知試驗平台的振動與噪音的來源以及傳動元件運轉特性；對於電磁式磁粉煞車作動以致試驗平台停止運轉瞬間，對試驗平台產生衝擊，造成磁粉煞車端有上移的現象，以錐形時頻分析法可詮釋此現象。此外，研究中應用時頻分析於水下聲學訊號探討，透過水下麥克風(hydrophone)量測水下暫態非穩態訊號，包括水中、水面航具及海流等之音響，進行種類鑑別。 The study aims are developing and implementing time-frequency distribution (TFD) techniques, including the smoothed Wigner-Ville, Choi-Williams and Cone-kernel distributions, to improve time-frequency resolution that spectrogram cannot possess simultaneously and eliminate interferences resulting from the cross terms in a Wigner-Ville distribution as well. Several synthetic signals have been designed to investigate these TFDs’ properties. We utilize the excellences of these TFDs in analyzing the time-variant or transient signals to analyze the mechanical dynamic signal from test-rig and compare with the rotational speed of transmission components to get the cause of vibration and noise on test-rig and the characteristics of transmission components. Also, by analyzing dynamic signal on the electromagnetic powder brake (EPB), we can know that EPB generate an impulse on test-rig at the moment of full stop that make test-rig shake upward around EPB. In addition, we apply TFA to analyze t underwater acoustic signals by using hydrophone to measure transient and nonstationary signal including acoustics of vessels or ocean current to classify targets.
