基於強化型Morlet轉換、解調變頻譜、多尺度熵、多頻帶頻譜熵與決策樹之齒輪箱異常診斷系統

Abstract

產業應用上，齒輪箱扮演著重要的角色；典型的齒輪異常，包含了：磨損、輪尺斷裂、動不平衡、缺乏潤滑等，嚴重的甚至會發生齒輪本身崩壞的情形。當齒輪出現故障，振動訊號可能被激發出異常的振動特性；因此，可藉由對振動訊號的分析，利用不同的訊號處理方法，達成齒輪箱的異常診斷。本論文提出一齒輪箱異常診斷系統，用以辨識齒輪箱的異常狀態情形。首先，使用解調變頻譜、影像熵、多尺度熵和多頻帶頻譜熵抽取出異常狀態之特徵；接著，利用抽取出之特徵建立一決策樹模型。本論文所使用的齒輪箱實驗資料來源，是工業技術研究院機械與系統研究所智慧系統技術組監控系統技術部所建置之齒輪箱實驗平台，並由作者親自進行所有的實驗以收集本論文所需之實驗資料。實驗結果顯示，訓練出的決策樹模型，對於測試使用的資料之異常診斷，具有高度的準確性。Gearboxes play an important role in industrial applications. Typical faults of gears include pitting, chipping, imbalance, loss-of-lubrication and more seriously, crack. When a gear has a fault, the vibration signal may carry the signature of the fault in the gears. Therefore, fault detection of the gearbox is possible by analyzing the vibration signal by different signal processing algorithms. In this dissertation, we propose a gearbox fault diagnosis system to distinguish different fault types of the gearbox. Firstly, signatures of the gear faults were extracted by the demodulation spectrum, image entropy, multi-scale entropy (MSE) and multiband spectral entropy (MBSE). Secondly, these extracted signatures were used to build a decision tree (DT) based model. In our simulations, the vibration signal datasets of gearbox from Industrial Technology Research Institute (ITRI) are utilized. In experimental results, the trained DT models have shown high accuracy of fault detection and fault classification on the test set.