本論文對於現今的機台AI判別提出了新的架構，該架構基於機台的振動變化量之上，並稱此機台振動變化量為衰減係數，透過計算機台之衰減係數，我們可以判斷出機台現在所處的狀態與行為，並且依照所屬的狀態與行為的不同，來計算機台健康程度。由於衰減係數的計算，是透過振動的變化量，而非振動的純量總值，因此該計算模型還可套用在不同的機台上。 本論文透過模擬與實際的數據來驗證以上之理論，從模擬取樣率、衰減率、振動週期之改變來實驗單一變數的情況，對衰減係數的影響，並套用在實際的馬達的運作上，在該馬達運作行程上分為五段，每段行程有著轉速與負載的差異，而透過收取該馬達的原始振動數據，很難以肉眼分辨其差異，但該數據透過衰減係數計算後，其五段不同行為之行程，卻是肉眼可見。因此本論文不僅提供一機台振動狀態判別的方法，透過該方法更能使振動數據具有更高的可讀性與意義，而其極限為振動週期介於2~180秒且感測器取樣週期不可超過750ms之任意機台。 ;This paper proposes a new framework for the AI discrimination of present machine, which is based on the vibration variation of the machine known as decay coefficient. By calculating the machine’s decay coefficient, we can determine the current state and behavior of the machine, and calculate the machine’s condition according to the different state and behavior. Since the calculation of the decay coefficient is based on the variation of the vibration rather than the total value of the vibration, the calculation model can be applied to different machines.
The aforementioned theory is verified by simulated and actual data. The effect of a single variable on the decay coefficient is tested by simulating the change of sampling rate, decay rate and vibration period, and applied to the actual motor operation. However, after the data is calculated by its decay coefficient, the five stages of travel with different behaviors are visible to the naked eye. Therefore, this paper not only provides a method to identify the vibration status of the machine, but also makes the vibration data more readable and meaningful through this method. As such, the limit of the vibration period between 2~180 seconds and the sensor sampling period will not exceed 750ms for any machine.
