臺灣博碩士論文加值系統

近年來隨著工業4.0技術進步迅速，國內各機械加工廠之自動化技術發展已十分成熟，各加工廠的設備維護成本也隨之提高，在加工過程中，加工參數為影響產品的精度與刀具磨耗的重要參數，以往只依靠經驗值進行加工參數的選用。類神經網路預測技術是以數值訓練的方式，以非線性的方式，找出加工參數與表面粗糙度及刀具磨耗的關係，使系統自身擁有預測的能力，藉此獲得較好的切削條件。有鑑於此，本研究擬開發一套簡易且快速之表面粗糙度分析與刀具磨耗預測系統，找出刀具磨耗較小同時表面粗糙度良好之加工參數。研究主要分為三個部分，其一是透過倒傳遞類神經網路建立出加工參數與刀具磨耗之間的關係；其二是藉由轉移函數、資料前處理、學習效率與慣量項參數分析來優化倒傳遞類神經網路，提升其學習效率與降低學習中的振盪現象；其三是以LabVIEW開發一套表面粗糙度與刀具磨耗預測系統，其功能包含類神經網路的學習、表面粗糙度與刀具磨耗的預測以及參數最佳化。理論分析方面，本研究採用類神經網路中的倒傳遞類神經網路，倒傳遞神經網路的部分，將加工參數中的切削速度、進給速度與材料總移除量作為學習網路中的輸入值，以刀具磨耗作為學習網路中的輸出值，透過倒傳遞網路可找出非線性關係的特性，找出加工參數對應輸出值的關係式，進而預測刀具磨耗，同時以切削試驗進行表面粗糙度與刀腹磨耗的分析。本研究利用切削試驗針對捨棄式車刀片車削304不鏽鋼圓棒進行加工參數的規劃與表面粗糙度量測及刀具磨耗檢驗。最後本研究將預測出的切削結果與實驗切削進行比較，以驗證學習模型和類神經網路預測系統之準確度。

In recent years, with the rapid progress of industrial 4.0 technology, the development of automation technology in domestic machinery processing plants is very mature, the cost of equipment maintenance at various processing plants is also increasing. During processing, processing parameters are important parameters that affect the accuracy of the product and tool wear, previously relying only on empirical values for the selection of processing parameters. Neural network prediction technology is based on numerical training, in a non-linear manner, the relationship between machining parameters, surface accuracy, and tool wear is found, so that the system itself has the ability to predict, thereby obtaining better cutting conditions. In view of this, this study intends to develop a set of simple and fast surface accuracy and tool wear prediction systems to find out the machining parameters with less tool wear and good surface accuracy. The research is mainly divided into three parts. One is to establish the relationship between processing parameters, surface accuracy, and tool wear through inverse transfer neural networks. The second is to optimize backpropagation neural networks through transfer function, data preprocessing, learning efficiency, and inertial-parameter analysis to improve learning efficiency and reduce oscillations in learning. The third is to develop a set of surface accuracy and tool wear prediction systems based on LabVIEW. The functions include neural network learning, surface accuracy and tool wear prediction and parameter optimization. Theoretical analysis, this study uses a backpropagation neural network in the neural network. Inverting the part of the neural network, taking the cutting speed, feed rate, and total material removal in the processing parameters as the input value in the learning network. Surface accuracy and tool wear are used as the output values in the learning network. Through the inverted transmission network, the characteristics of the nonlinear relationship can be found, and the relationship between the output parameters of the processing parameters can be found, and the surface accuracy and tool wear can be predicted. In this study, the cutting test was used to plan the machining parameters, surface accuracy measurement and tool wear test of the turning 304 stainless steel round bar. Finally, this study compares the predicted cutting results with the experimental cutting to verify the accuracy of the learning model and the neural network prediction system.

摘要...............................i
Abstract..........................ii
誌謝.............................iii
目錄..............................iv
表目錄.............................vi
圖目錄............................vii
符號說明...........................ix
第 一 章 緒論.......................1
1.1 研究背景與動機..................1
1.2 研究目的.......................2
1.3 論文架構.......................2
第 二 章 文獻回顧..................3
2.1 類神經網路優化相關研究..........3
2.2 刀具壽命分析探討...............8
第 三 章 研究架構與方法............15
3.1 研究架構與流程................15
3.2 硬體及系統架構................16
3.2.1 實驗材料與刀具、刀把規格.....16
3.2.2 實驗機台介紹................16
3.3 類神經網路系統................17
3.3.1 類神經網路基本概論...........17
3.3.2 類神經網路導論..............19
3.3.3 感知機網路..................22
3.4 倒傳遞類神經網路..............23
3.4.1 倒傳遞類神經網路基本概論.....23
3.4.2 監督式網路演算法............23
3.4.3 學習過程...................27
3.4.4 學習參數...................29
3.5 切削試驗與刀具壽命............36
3.5.1 刀具壽命定義...............36
3.5.2 加工參數的影響.............37
3.5.3 刀具磨耗量測...............38
3.5.4 車床切削參數...............38
第 四 章 實驗結果................39
4.1 切削實驗....................39
4.1.2 刀具磨耗切削結果...........40
4.2 刀具壽命預測系統建置..........47
4.2.1 倒傳遞類神經網路學習功能....47
4.2.2 倒傳遞類神經網路預測功能....48
4.3 倒傳遞類神經網路參數分析......50
4.3.1 隱藏層神經元數目...........50
4.3.2 隱藏層數目................52
4.3.3 轉換函數..................55
4.3.4 正規化....................56
4.3.5 慣量項....................57
第 五 章 結論與未來展望..........58
參考文獻........................59
Extended Abstract..............61

[1]中華民國行政院, “智慧機械產業推動方案規劃”.
[2]Peter Zelinski, ” Cutting Costs with Cutting Tools: Instead of Life or Price, Look to Capability”, Related Suppliers: Sandvik Coromant Inc,Modern Machine Shop, 10/15/2003.
[3]S. Kurada, C. Bradley,” A review of machine vision sensors for tool condition monitoring”, Computer in Industry, 34 (1997), pp. 55-72.
[4]M. Castejon, E. Alegre, J. Barreiro, L.K. Hernandez,” On-line tool wear monitoring using geometric descriptors from digital images”, International Journal of Machine Tools and Manufacture, 47 (2007), pp. 1847-1853.
[5]馬政剛,“類神經網路誤差估計與學習速率之探討”, 逢甲大學-應用數學系碩士班, 2013.
[6]Wen-Tsao Pan,” The Study On The Optimizing Predictability Of The Neural Networks: Case Study Of Finance Holding Stocks”, 遠東學報第二十四卷.
[7]Ming Tan, Gaohong He, Xiangcun Li, Yuanfa Liu, Chunxu Dong, Jinghai Feng,” Prediction of the effects of preparation conditions on pervaporation performances of polydimethylsiloxane(PDMS)/ceramic composite membranes by backpropagation neural network and genetic algorithm”，Procedia Computer Science, Volume 89, 22 March 2012, Pages 142-146
[8]Ming Tan, Gaohong He *, Fei Nie, Lingling Zhang, Liangping Hu,” Optimization of ultrafiltration membrane fabrication using backpropagation neural network and genetic algorithm”，Procedia Computer Science, Volume 45, Issue 1, January 2014, Pages 68-75.
[9]Han-Xiong Huang, Jiong-Cheng Li, Cheng-Long Xiao,” A proposed iteration optimization approach integrating backpropagation neural network with genetic algorithm”，Procedia Computer Science, Volume 45, Volume 42, Issue 1, January 2015, Pages 146-155.
[10]何佳祥,”應用基因演算法優化類神經網路於汙水廠興建成本預估之研究”, 國立雲林科技大學-環境與安全衛生工程系碩士班,2003.
[11]W.Grzesik, P. Nieslony, W. Habrat, J. Sieniawski, P. Laskowski, “Investigation of tool wear in the turning of Inconel 718 superalloy in terms of process performance and productivity enhancement”, Procedia Computer Science, Volume 118, February 2018, Pages 337-346.
[12]MuhammadRizal, Jaharah A.Ghania, Mohd ZakiNuawia and Che Hassan Che Harona “Online tool wear prediction system in the turning process using an adaptive neuro-fuzzy inference system,” Procedia Computer Science, Volume 13, Issue 4, April 2013, Pages 1960-1968.
[13]T.Mikołajczyk, K.Nowicki, A.Bustillo, D.Yu Pimenov,” Predicting tool life in turning operations using neural networks and image processing” Procedia Computer Science, Volume 104, 1 May 2018, Pages 503-513.
[14]Dongdong Kong, Yongjie Chen, Ning Li,” Gaussian process regression for tool wear prediction” Procedia Computer Science, Volume 104, 1 May 2018, Pages 556-574.
[15]Samik Dutta, Surjya K.Pal, RanjanSen, “On-machine tool prediction of flank wear from machined surface images using texture analyses and support vector regression”, Procedia Computer Science, Volume 43, January 2016, Pages 34-42.
[16]I.P. Okokpujie, O.S. Ohunakin, C.A. Bolu, K.O. Okokpujie, “Experimental data-set for prediction of tool wear during turning of Al-1061 alloy by high speed steel cutting tools”, Procedia Computer Science, Volume 18, June 2018, Pages 1196-1203.
[17]勝傑工業股份有限公司, http://www.ecoca.com/tw_index.asp.
[18]葉怡成，類神經網路模式應用與實作，儒林圖書有限公司，1993年1月初版，2000年4月七版.
[19]范揚志,”應用類神經網路與基因演算法於射出成形製程參數最佳化之研究”, 中華大學-科技管理研究所,2006.
[20]晉鳳山,”改良式倒傳遞類神經網路之研究”,國立台中師範學院教育測驗統計所 ,2002.
[21]蘇裕翔,”結合類神經網路與遺傳演算法評估地震尖峰地表加速度之研究”,國立屏東科技大學-土木工程系,2014.
[22]萬絢, 雷祖強, 陳達祺, ” 以樹狀倒傳遞類神經網路於田埂判釋研究”, 航測及遙測學刊第十六卷第一期,2011.03.
[23]ISO, “Tool Life Testing with Single-Point Turning Tools,”ISO, 5th draft proposal, ISO/TC 29/ WG 22 (secretariat 37), 91 (1972).