基於電流特定邊頻特徵之永磁同步馬達軸承故障診斷

Abstract

滾珠軸承為電機機械的關鍵元件，其故障將導致工業設備停機，甚至在車用領域引發嚴重事故。傳統的軸承故障診斷多依賴加速規量測振動訊號，但此類感測器不僅具有安裝受限的問題，亦伴隨較高的感測成本。基於上述動機，本研究提出一種基於電流訊號分析（Motor Current Signature Analysis, MCSA）的軸承故障診斷方法，旨在以低成本之32位元微控制器上實現即時軸承診斷。研究中利用永磁同步馬達（Permanent Magnet Synchronous Motor, PMSM）之相電流訊號作為故障特徵來源，並針對故障電流訊號之訊噪比（SNR）偏低的問題，提出三項改良方法以提升 MCSA 診斷效能： (1)建立軸承故障電流之解析模型，用以評估馬達參數對 MCSA 診斷效能的敏感度。 (2)提出相電流同步技術以提升故障諧波之SNR，並透過「短路測試」量測感應電流，以避免PWM諧波對故障訊號的干擾。 (3)採用Goertzel演算法實現即時離散傅立葉轉換（DFT），使低成本微控制器能進行即時故障判別。 本研究於兩部參數不同的 PMSM上進行實驗驗證，所有診斷流程的演算邏輯僅需65kB記憶體，可有效實現於32位元微控制器中，並在10 kHz中斷服務例程（ISR）下僅佔用約2% 計算負載。結果證實所提出之方法可有效提升 MCSA 於低成本嵌入式系統上的即時軸承故障診斷能力。

Bearings are critical components inside electric motors. Bearing failures cause the equipment downtown in industry or sever accidents in automobile. Traditional bearing fault diagnostic methods rely on vibrational signature detection through accelerometers. These sensors suffer from installation limitations and considerable sensor cost. Under this motivation, this paper proposes a bearing fault diagnostic method based on motor current signature analysis (MCSA). The purpose is to implement this MCSA of bearing fault on real-time 32-bit microcontrollers at low cost. On the basis, phase current signals inside permanent magnet synchronous motor (PMSM) are utilized for bearing fault detection. Because fault reflected current signatures typically contains insufficient signal-to-noise ratio (SNR), three improvements are used to improve the MCSA-based bearing fault detection using existing microcontroller. 1)An analytical model for bearing fault reflected current is developed to evaluate the motor parameter sensitivity on MCSA performance. 2)A phase current synchronization algorithm is proposed to increase the SNR of fault reflected current harmonics. Besides, the motor short-circuit current measurement is performed to eliminate PWM harmonics on fault currents. 3)A real-time Goertzel-based discrete Fourier transform (DFT) is used for real-time fault detection on a low-cost microcontroller. The proposed MCSA fault detection is experimentally validated on two PMSMs with different parameters. All bearing diagnostic algorithms are implemented on a 32-bit microcontroller with only 65 kB memory size and 2% computation burden under 10kHz interrupt service routine (ISR).