本篇論文提出「以有效的模糊C-平均值演算法(Fuzzy C-Means, FCM)分析馬達電流訊號與辨識馬達的品質類別」。本論文是將FCM演算法應用於辨識馬達的品質類別為例說明。辨識過程簡述如下:(i)首先,從已知品質類別的樣本馬達中,取得各種品質類別的輸入訊號,並分別計算在各種不同品質類別之主要特徵點的特徵值向量;(ii)其次,執行FCM演算法,並計算在所有不同品質類別中,每一個品質類別之最終的群體中心值;(iii)接著,計算下列兩者之間的歐氏距離,它們分別是:待辨識馬達之主要特徵點的特徵值向量,及每一種品質類別之最終的群體中心值;(iv)最後,依據計算得到的歐氏距離,決定該待辨識馬達之品質類別是屬於何種類別。若是屬於“品質類別-z”,條件是在所有的品質類別中,該待辨識馬達的主要特徵值向量,它與“品質類別-z”之最終的群體中心值,兩者之間的歐氏距離是最小的值。本篇論文經過多次的測試,證實以模糊C-平均值演算法應用於分析電流訊號與辨識馬達的品質類別,是一個有效且實用的方法。
This dissertation proposes a novel Fuzzy C-Means (FCM) algorithm for determining the motors’ quality types by analyzing their current waveforms. The determining process of motors’ quality types are listed as follows: (i) Obtaining input current signals of sample motors with well-known distinct classes and then calculating their feature values of qualitative features; (ii) Computing centroid values of each quality class; (iii) Computing Euclidean distances of their feature values and centroid values of classes; and (iv) Determining motors’ quality types using obtained Euclidean distances. A motor belongs to a class-z if it and class-z has the minimum Euclidean distance. Experimental results show that the proposed FSM algorithm is an effective quality determining method for motors.