透過您的圖書館登入
IP:54.175.5.131
  • 學位論文

監督式學習類神經網路於銑削斷刀即時監控之研究

An In-Process Monitoring Approach of Using Supervised Learning Neural Network to Detect Too Breakage in End Milling Operations

指導教授 : 黃博滄

摘要


近年來隨著電子產業的發達,電腦數值工具機(Computer Numerical Control, CNC)銑削加工的需求也隨之增大,再加上產品品質的要求,使得CNC銑削技術的提升成為各自動化產業注重的議題。而在CNC銑削加工時,刀具的磨損與斷裂都會明顯的導致加工品質的不穩定與生產效能降低,並且還會使昂貴的加工工件損壞、報銷,造成加工成本的提升,而CNC工具機本身並沒有自行探測刀具的能力,因此為了能夠確保生產品質的穩定與達到完全的自動化生產,本研究提出三種即時的斷刀監控決策系統,使得機器在運作的同時,能夠即時的探測到刀具的斷裂。這三種監控系統所使用的資料處理方法皆為監督式學習的類神經網路演算法,其分別為倒傳遞網路(Back-Propagation Network, BPN)、機率神經網路(Probabilistic Neural Network, PNN)及學習向量化網路(Learning Vector Quantization, LVQ)。此三個系統皆使用相同的力量感測器來量測切削力訊號,再以切削力與三個加工參數,分別為轉軸速度、進給率及切削深度作為網路的輸入因子,以及其相對應的目標參數來做網路的訓練,並於訓練完成後套用至CNC銑削工具機中來探討其判斷的精確性。最後在三種系統架構完成後,本研究再利用網路訓練速度較快的網路演算法,提出另一種判斷準確度較高的可調式斷刀即時監控系統,此系統可以在刀具被誤判為斷刀停機時,將誤判的參數資料丟回資料庫,重新訓練網路結構,使得此可調式預測系統能隨著樣本資料的增加,來提高網路預測之能力,以達到更 精準的斷刀預測。

並列摘要


In recent years, the needed of Computer Numerical Control (CNC) milling is creasing by the development of the electronic industry and plus the request of quality which makes the CNC milling technology progress become the issue that ever automatic industry focus on. When the CNC milling, the wear and breakage of the cutting tool obviously cause the unstable productive quality and decrease the produce efficacy, moreover, will damage and destroy the expensive tools and increase the manufacture cost. Since the CNC doesn’t have the ability of detecting the cutting tool itself, this research will mention about three decision-making systems of prompt detecting the tool breakage that makes the machine can detect the breakage of cutting tool immediately during its working. The data processing method of the three systems are all Supervised Learning Neural Network algorithm, there are Back-Propagation Network (BPN), Probabilistic Neural Network (PNN) and Learning Vector Quantization (LVQ). The systems all used the same force sensor to measure the cutting force signal, and the cutting force and machining parameters, such as spindle speed, feed rate, and depth of cut will be the input of network, and tool conditions will be the output to train the network. Then use the corresponding output to train the network, and will completely imitate in CNC afetr training to discuss the accuracy of the judgment. When the three systems finish the frame, this research will use the faster training Network and address the other prompt and relearning of tool breakage detecting system with higher accurate judgment. The system can put the ill-judged argument data back to the database when the cutting tool is shut down by ill-judged the breakage, and retrain the network structure to achieve more accurate tool breakage detecting by increasing the sample data of the adjustable detect system to raise the ability of the Network predicting.

參考文獻


[2] P. Prickett, C. Johns, An Overview of Approaches to End Milling Tool Monitoring, International Journal of Machine Tools and Manufacture, 39, 105-22, 1999.
[5] L. Dan, J. Mathew, Tool Wear and Failure Monitoring Techniques for Turning- a review, International journal of Machine Tools and Manufacture, 30(4), 579-598, 1990
[7] Y. Peng, Empirical Model Decomposition Based Time–Frequency Analysis for the Effective Detection of Tool Breakage, Journal of Manufacturing Science and Engineering, Transaction of the ASME, 128, 154-166, 2006.
[8] E. Govekar, J. Gradivek, I. Grabec, Analysis of Acoustic Emission Signals and Monitoring of Machining Processes, Ultrasonics, 38, 598-603, 2000.
[9] D. V. Hutton, F. Hu, Acoustic Emission Monitoring of Tool Wear in End-Milling Using Time-Domain Averaging, Journal of Manufacturing Science and Engineering, Transactions of the ASME, 121, 8-12, 1999.

被引用紀錄


李軍逸(2017)。CNC 工具機產效暨與節能同步優化設計與空刀率之智能監測研究〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840%2fcycu201700110
周聖軒(2013)。具雙向溝通介面之切削異常線上智慧監控與加工製程優化方法之研究〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840%2fcycu201400150
張黃傑(2012)。平面銑削之灰色即時可調式學習表面粗糙度預測系統開發〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840%2fcycu201300859
陳大芳(2012)。精密製造系統之智能化異常監控方法與應用〔博士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840%2fcycu201201040
陳久弘(2012)。運用灰關聯因子分析與類神經網路於銑削表面粗糙度即時預測系統〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840%2fcycu201200725

延伸閱讀