隨著電動車技術快速發展，齒輪箱之振動與噪音控制已成為NVH (Noise， Vibration and Harshness)性能開發的重要課題。相較於傳統內燃機車輛，電動車缺乏引擎背景噪音的遮蔽效果，使齒輪箱振動更易被感知，並且車輛馬達具有寬廣轉速範圍與快速扭矩響應，造成振動訊號呈現明顯非穩態特性，增加故障特徵擷取之困難。本研究以合作廠商提供之電動車用二級減速齒輪箱為研究對象，建立適用於變轉速工況之振動量測與故障特徵分析流程。首先於輸入軸、中間軸及輸出軸鄰近位置配置三組三軸加速規，量測齒輪箱升速、穩速及減速過程之振動訊號，並以RMS評估整體振動能量以及以峰值因數評估局部衝擊特性。為降低變轉速造成之頻率漂移現象，利用輸出軸轉速訊號進行角度重採樣與階次分析。針對齒輪撞傷所造成之週期性衝擊與振幅調變特徵，進一步利用包絡分析，並以各軸轉階次及其倍頻振幅作為故障特徵指標。為建立健康狀態之比較基準，本研究針對30筆健康齒輪箱重複量測資料進行平均值、標準差及變異係數分析，並將同一量測位置三方向階次振幅以和平方根(Root Sum Square， RSS)計算。最後以輸入軸、中間軸及輸出軸齒輪撞傷案例進行驗證。研究結果顯示，RMS與峰值因數受量測方向影響，故障特徵未呈現一致增加；齒輪第一嚙合階次與第二嚙合階次振幅超過門檻比例分別僅為22.2%與0%，顯示局部齒輪撞傷不一定造成整體嚙合振幅增加；相較之下，輸入軸、中間軸及輸出軸故障案例於包絡階次振幅超過門檻比例分別為86.7%、100%及100%，遠高於健康門檻。因此包絡階次分析能較一致地提取齒輪撞傷所造成之軸轉週期調變特徵，並可作為本測試條件下齒輪箱異常篩檢與故障軸位置辨識之依據。;With the rapid development of electric vehicle technology， vibration and noise control in gearboxes has become an important issue in NVH (Noise， Vibration and Harshness) performance development. Compared with conventional internal combustion engine vehicles， electric vehicles lack the masking effect of engine background noise， making gearbox vibration more readily perceptible. In addition， electric motors operate over a wide speed range and exhibit rapid torque response， resulting in pronounced non-stationary characteristics in the measured vibration signals and making fault feature extraction more challenging. This study investigates a two-stage reduction gearbox for electric vehicles provided by an industrial partner and establishes a vibration measurement and fault feature analysis procedure suitable for variable-speed operating conditions. Three triaxial accelerometers were installed near the input shaft， intermediate shaft， and output shaft to measure gearbox vibration during acceleration， steady-speed operation， and deceleration. The root mean square (RMS) was used to evaluate the overall vibration energy， while the crest factor was employed to characterize local impulsive behavior. To reduce the frequency drift caused by variable-speed operation， angular resampling and order analysis were performed using the output-shaft speed signal. For the periodic impacts and amplitude modulation characteristics induced by gear impact damage， envelope analysis was further applied， and the shaft rotational orders and their harmonic amplitudes were adopted as fault feature indicators. To establish a healthy-state reference for comparison， 30 repeated measurements of the healthy gearbox were statistically analyzed using the mean， standard deviation， and coefficient of variation. The three directional order amplitudes at each measurement location were further combined using the root sum square (RSS). Finally， fault cases involving gear impact damage on the input shaft， intermediate shaft， and output shaft were used for validation. The results show that RMS and crest factor are affected by measurement direction， and the fault features do not exhibit a consistent increasing trend. The proportions of amplitudes exceeding the healthy threshold were only 22.2% and 0% for the first and second gear-meshing orders， respectively， indicating that local gear impact damage does not necessarily increase the overall meshing vibration amplitude. In contrast， the proportions of envelope-order amplitudes exceeding the threshold were 86.7%， 100%， and 100% for the input-shaft， intermediate-shaft， and output-shaft fault cases， respectively. Therefore， envelope order analysis provides a more consistent means of extracting the shaft-periodic modulation characteristics induced by gear impact damage and can serve as a basis for gearbox abnormality screening and fault-shaft localization under the test conditions investigated in this study.
