永磁同步馬達偏心診斷與滾珠軸承故障診斷

Abstract

機械故障為永磁同步馬達常見的故障種類，以傳動軸承偏心及軸承內部損壞兩者最為常見，這會使馬達運轉效率降低並有最終有運轉安全上的疑慮，因此馬達驅動器有必要透過運轉即時故障診斷提早發現問題，預約保養維修。 本論文首先提出用馬達相電流感測訊號，藉由電流區域極值代替傳統包絡線(envelope)計算的偏心故障診斷方法，目的是將演算法計算量降低，並且透過區域極值的先後順序判斷計算特徵頻率振幅，方便在微控制器上進行即時計算。同時因為區域極值的特徵頻率與轉速無關，此方法亦改善原先因頻率解析度的限制只能在定轉速診斷的問題。 本論文也針對軸承內部損傷，利用加速規量測振動訊號來進行分析改善，故障診斷是藉由時間同步平均(time synchronous averaging, TSA)方式降低包絡線上特定頻率振動的干擾，使在頻譜上軸承特徵頻率的振幅能更明顯，在分析上比較直觀。 為了驗證診斷的性能，診斷平台是使用100W馬達進行實驗測試，從實驗結果可以證實，所提出偏心診斷方法在電流區域極值與包絡線的特徵頻率振幅誤差在3.2%以內，且在微控制器上能在30us內計算完成，此外軸承損傷診斷實驗上也驗證用所提出方式可以有效診斷出軸承損傷，在有偏心影響下也能照樣分析，且能夠降低偏心振動在頻譜上的振幅。

Mechanical faults are common types of faults in permanent magnet synchronous motors, with eccentricity in the drive bearing and internal bearing damage being the most common. These faults can lead to a decrease in motor efficiency and eventually raise concerns about operational safety. Therefore, it is necessary for motor drives to employ real-time fault diagnosis to detect problems early and schedule maintenance and repairs. This thesis proposes a fault diagnosis method for eccentricity faults using motor phase current sensing signals. Instead of using the traditional envelope calculation, the method utilizes the local extreme values of the current in specific regions. The objective is to reduce the computational complexity of the algorithm and determine the computed feature frequency amplitudes based on the order of local extreme values. This facilitates real-time computation on microcontrollers. Additionally, since the characteristic frequencies of the local extreme values are independent of motor speed, this method overcomes the limitation of frequency resolution for diagnosing faults only at a fixed speed. The thesis also focuses on internal bearing damage and proposes an analysis improvement using acceleration measurements of vibration signals. The fault diagnosis utilizes time synchronous averaging to reduce the interference from vibration at specific frequencies on the envelope. This enhances the visibility of bearing characteristic frequencies in the spectrum and provides a more intuitive analysis. To verify the performance of the diagnosis, experiments were conducted on a 100W motor using a diagnostic platform. The results confirm that the proposed eccentricity diagnosis method achieves an error within 3.2% for the feature frequency amplitudes between the current local extreme values and the envelope. Furthermore, the computation can be completed within 30us on a microcontroller. In addition, the experiments for bearing damage diagnosis validate the effectiveness of the proposed method in detecting bearing faults, even under the influence of eccentricity, while reducing the amplitude of eccentric vibration in the spectrum.