永磁馬達遠端監控及在線退磁診斷系統

Abstract

時至今日，永磁馬達在人們日常生活中的應用已經越來越廣泛。然而，在永磁馬達的使用過程中，可能發生幾種錯誤。退磁錯誤便是其中一種可能發生在永磁馬達的錯誤。退磁錯誤指的是永磁馬達中的永磁鐵顯著地失去磁性，其可能導致永磁馬達運轉不穩定。 本篇論文提出了一種用於永磁馬達監控以及線上退磁錯誤診斷的系統。此線上退磁錯誤診斷系統是基於反電動勢（BEMF）常數建立的。由於反電動勢常數與永磁馬達永磁鐵的磁性成正比的關係，因此其可以被用來估計永磁馬達的退磁情況。當今，亦有不少退磁診斷的方法，如有限元分析法（FEA）以及時頻分析法（TFA），但是以上所提到的方法很難用於在線實時分析。然而，根據永磁馬達的電氣模型計算反電動勢常數的方法可以被應用在馬達運轉的狀況下。 傳統的反電動勢常數估計的方法存在一些缺陷。其主要包括在永磁馬達低速運轉時，由於反電動勢較小所造成的反電動勢常數不準確估計，以及此方法無法應用在馬達加負載的情況下。針對這些缺陷，本文提出了使用最小平方估計法（LSE）建立回歸模型，用於消除負載以及摩擦等噪聲的影響。之後的實驗結果將會證明，無論在永磁馬達加負載或者低速運轉的情況下，建立的回歸模型都可以有效提升反電動勢常數估計的正確率。 永磁馬達監控系統可以為反電動勢常數的估計提供實時的數據支援。不僅如此，此永磁馬達監控系統結合了嵌入式板（ARM board），電腦（PC）以及手機，可以為用戶以及製造廠商提供遠端監控的數據，滿足他們的不同需求。 由於目前工業所使用的馬達變頻器缺少退磁診斷的功能，因此，本文所提出的系統可以在日後與馬達控制所需的變頻器結合，並用於永磁馬達的監控與退磁診斷。

Nowadays, Permanent-Magnet (PM) motors are utilized widely in human being’s daily life. However, several kinds of faults may occur in PM motors. Demagnetization fault is one of the faults existing in PM motors and it means that the magnetism of permanent magnets in PM motors decreases prominently. In this thesis, a system of PM motor monitoring and online demagnetization fault diagnosis is proposed. The online demagnetization fault diagnosis system is based on Back Electromotive Force (BEMF) constant which is proportional to the magnetism of permanent magnets. The BEMF constant is able to be estimated by means of the electric models of PM motors. In this way, the diagnosis of demagnetization fault can be applied in online condition while many methods such as Finite Element Analysis (FEA) and Time-Frequency Analysis (TFA) are difficult to be implemented when a PM motor is in service. The conventional approach of BEMF constant estimation has some drawbacks including bad performance when a PM motor is in low speed. Besides, the traditional way is not capable of being used when loading is added to a PM motor. Contraposing these deficiencies, the approach proposed in this thesis is going to take advantage of regression models to figure them out via Least Square Estimate (LSE). The regression models can eliminate the impact caused by loading and friction noise. The consequent experiments will prove that the regression models can improve the accuracy of estimating BEMF constant effectively when a PM motor is not only in loading condition but also in low speed. The motor monitoring system combined with ARM boards, PCs and mobile phones is able to offer users and manufacturers data remotely, which can meet their different kinds of needs. In addition, the system also provides the estimation of BEMF constant with the data of PM motor in real time. Owing to lacking the function of demagnetization fault diagnosis in most inverters presently, the system addressed in this thesis is likely to be associated with the inverters for PM motors’ monitoring and diagnosis in the future.