Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/91275
標題: Application of artificial neural network to determine the status of spindle bearing
應用類神經網路判斷銑削主軸之軸承狀態
作者: 陳志騏
Zhi-Chi Chen
關鍵字: artificial neural network;bearing;類神經網路;軸承
引用: 1.[TIMKEN, 2003] TIMKEN, Tapered Roller Bearing Damage Analysis, 2003 2.[Subrahmanyam, 1997] M. Subrahmanyam and C. Sujatha, Using neural networks for the diagnosis of localized defects in ball bearings, Tribology International Vol. 30, No. 10, pp. 739–752, 1997 3.[Blödt et al, 2008] Martin Blödt, Pierre Granjon, Bertrand Raison and Gilles Rostaing , Models for Bearing Damage Detection in Induction, Motors Using Stator Current Monitoring IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 55, NO. 4, APRIL 2008 4.[Gao, 2008] Robert X. Gao,Neural Networks for Machine Condition Monitoring and Fault Diagnosis,Department of Mechanical and Industrial Engineering, University of Massachusetts, 2008 5.[廖天志&許原昇,2004] 廖天志&許原昇,工具機主軸之鬆拉刀機構設計探討,精密機械研究發展中心,2004年5月25日,機械資訊月刊 568 期 6.[朱效賢,2005] 朱效賢,包絡譜分析於軸承故障診斷之探討暨工程應用,中央大學論文,2005年9月30日 7.[黃興杰,2010] 黃興杰,工具機主軸振動與檢測,2010年04月22日 8.[彭善謙,2004] 彭善謙,綜合振動信號於馬達故障診斷,中原大學論文,2004 9.[黃興杰,2010] 黃興杰,工具機主軸振動與檢測,財團法人精密機械研究發展中心,2010年04月22日,迴轉機械之振噪檢測研討會 10.[趙安民,2004] 趙安民, 馬達故障診斷之模糊類神經網路,中原大學論文,2004年7月
摘要: 
The machining performance of the spindle of a milling machine is highly related to its operation condition. Therefore it is important to monitor the operation status of the spindle. While the most important factor affecting spindle operation is the condition of its bearings, this research proposes a diagnostic method to improve accuracy for identifying bearing fault status of the milling spindles via measuring rotation signals of operations and establishing a decision model. The developed method is testified against several spindles from market and shows it can be used to identify the condition of spindle bearings.
In the study, the artificial neural network (ANN) is employed to build the decision model. The failure mode and effect and the feature corresponding to the failure of spindle bearings are analyzed at first. The items and signals to be measured are then designed accordingly. These signals are then collected from a lot of spindles, with some of them being damaged. Conditions and operation signals of bearings, dissembled from these spindles, are also measured. These measured signals are used to train the ANN for building the relationship among the bearing conditions and the measured signals. Two approaches, two-step and one-step approaches, are further conducted to compare the accuracy of the models. The result shows that the accuracies of the two approaches were 85% and 81% respectively but the one-step approach is more practical as it can be employed to industrial application directly. The major contribution of this research is the approach to build up an ANN model that can be used to identify the status of spindle bearings with fairly good accuracy. The model can be further employed for predictive monitoring of the spindle of a machine tool such that failure of the spindle can be forecasted to prevent against the loss of production due to spindle failure.

銑削主軸之加工性能與其運轉狀況息息相關,因此對於主軸運轉狀態的掌握至關重要,而影響主軸運轉性能最為重要的因素即為其軸承健康狀態,故本研究針對銑削加工用主軸,透過量測其運轉訊號,建立其軸承狀態的判斷法則,提出診斷分析方法,以提高軸承故障鑑別率,本研究結果並經驗證可用在實際主軸運轉時軸承健康狀態預測。
本研究應用類神經網路建構預測模型,首先分析主軸故障之要因,以及主軸軸承損壞對應之特徵,作為設計量測項目與訊號內容之依據,並透過返修之受損主軸,以儀器採集主軸運轉的關鍵特徵信號,之後進一步採集受測主軸之軸承運轉訊號,將此二種訊號用以訓練本研究之預測模型,研究進一步採用兩階段訓練之第一組模型以及一階段訓練之第二組模型進行比較,結果顯示兩種方法之預測準確度分別為85%與81%,唯第二組模型可透過訊號量測與處理即可預測結果,較為實用。本研究之主要貢獻在於以類神經的建立特徵訊號與軸承損壞現象間之關係,此模型可提供業界作為主軸運轉健康診斷及預警監控之用,掌握主軸軸承之健康狀態,在其失效前及早預警,避免主軸中途無預警損壞造成生產停工或工件與主軸毀損之損失。
URI: http://hdl.handle.net/11455/91275
Rights: 不同意授權瀏覽/列印電子全文服務
Appears in Collections:機械工程學系所

Files in This Item:
File Description SizeFormat Existing users please Login
nchu-104-5100061001-1.pdf4.43 MBAdobe PDFThis file is only available in the university internal network   
Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.