Abstract
由於近年來辦公室自動化與家庭電器設備之輕量小型化需求，相對應產品所內具之馬達均也朝小型迷你化設計。馬達小型化會造成轉動慣量之下降，將造成明顯之轉矩漣波，對於需在低轉速區工作之小型馬達，會造成轉速變動，影響產品之品質。本研究所探討分析之對象為用於錄放影機，隨身聽等的小型軸向氣隙無刷直流馬達。對於此種馬達而言，除了磁通與相電流波形匹配不良所造成之漣波外，尚有兩種漣波來源存在此類馬達中，一是霍爾元件位置之偏移，另一是相線圈位置之偏移，二者均能造成顯著之轉速變動。在第一部份研究工作中，先由建立一個軸向氣隙型無刷直流馬達之動態方程式開始，接著進行此種馬達之轉矩漣波分析，然後以電腦模擬證實上述二類偏移會對轉矩及轉速造成影響，最後提出一套鑑別此二漣波源的方法，以提供製程精度改善與設計控制器之參考。其次，由於此馬達系統的非線性特性與回授信號數目之限制，使用傳統控制器設計方法會有一些預期之困難存在，故在第二部份之研究中，我們提出一個使用回授誤差學習法則之控制系統架構。由於類神經網路特性之發揮，此一結合前饋類神經網路控制器與普通回授控制器之架構具有許多傳統控制器缺少之優點。電腦模擬顯示，不論馬達系統有或無元件偏移，所設計之馬達控制系統都能有效地壓制原本存在之轉速變動。Brushless DC (BLDC) motors have been prevalently used in thefields of office automation (OA), home appliances (HA), andportable products such as video cassette recorders (VCR) andcompact disc (CD) players. In recent years, the trend towardsminiaturization and weight reduction for those applications hasmade necessary miniaturization of motors and the correspondingdriver electronics devices. This implies that the motors willusually have greater steady-state speed variation since theirinertia are too small to suppress the torque fluctuation. Thisthesis focuses on the torque ripple analysis and velocitycontrol for a specific class of BLDC motors, i.e. axial-fieldBLDC motors. For those kinds of motors, there are two sourcesof torque ripples which are neglected by previous literature, i.e. bias of Hall sensors and bias of armature windings. Analysisand simulation of the torque production in such motors indicatethat both of the bias could induce significant torque ripples.Accordingly, methods based on patterns of the ripples areproposed here to do the identification of such ripple sources.The methods could supply assembly processes of the motors withcorrective information. Next, a control structure incorporatedneural networks is addressed, which is basically a feedback-error-learning scheme and has several advantages worthy ofnotice. Simulation results are presented, which verify thefeasibility of the controller. Also, this controller couldeffectively suppress the speed fluctuation whether or not thereis bias of sensors or windings.