摘要 本研究之目的，在於減低火箭發射時於駕駛室產生的暫態非穩態高分貝噪音;為方便評估控制效果，研究中使用一全比例壓克力模型模擬發射車駕駛室音場。研究中以FXLMS適應性演算法發展主動式噪音控制(Active Noise Control, ANC)系統。由於FXLMS型式之系統，其適應速度及系統穩定性取決於收斂因子的大小，而收斂因子之大小不容易決定，研究中使用可變因子演算法，使收斂因子能依據控制狀態自行調整改善系統的消音速度及穩定性。 此外，為達到更好的控制效果及應用方便性，改良市售電子式耳罩以取代8吋控制喇叭，並與所使用8吋喇叭的多通道ANC系統之消音效果作比較。考慮真實環境下第二路徑鑑別的準確性，將線上鑑別技術與耳罩做結合，並以穩態噪音(複頻噪音)及非穩態噪音(火箭噪音)驗證其效果。此外，對於不同麥克風擺放距離、及鑑別的取樣頻率，對於第二路徑鑑別所需階數之影響亦作詳細的討論。實驗結果顯示，使用可變因子演算法對穩態及非穩態噪音的消音速度有明顯改善，而第二路徑線上鑑別對於複頻噪音能提供更好的消音效果。 Abstract The main purpose of the study is to suppress the transient and nonstationary high sound-pressure-level (SPL) noise perceived by drivers inside the control chamber during rocket launching. A full-size acrylic model is built up for mimicking the sound field of the control room of a rocket-launching vehicle. For achieving the noise reduction, the adaptive algorithm called as FXLMS is applied to implement the active noise control (ANC) system. The convergence speed and stability of the FXLMS-type systems depend on the step-size values mainly. They are usually difficult to be decided for fast converging, and keeping the system stable. In the study, the variable step-size algorithms are applied to automatically select an appropriate step size varying with system status for the desired performance. In addition, a commercial electrical headset is modified for noise-cancellation purpose to obtain best noise reduction and convenience. The performance of the modified active headset is compared with the multi-channel ANC system, which employs 8-inch loudspeakers to generate anti-phase sound waveforms. Considering precisely modeling the secondary path for practical applications, an on-line modeling technique is combined with the ANC headset, and validated using both the complex harmonic noise and rocket-firing noise. Besides, the needed order of the secondary path for different distances and estimated frequencies is also described in detail. Experimental results show that the ANC system with the variable step-size algorithm performs well both in complex harmonic and transient high-SPL shock noise environments. Additionally, the on-line modeling method provides more noise attenuation for complex-harmonic noise.