跳到主要內容

臺灣博碩士論文加值系統

(44.223.39.67) 您好!臺灣時間:2024/05/26 14:42
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

: 
twitterline
研究生:蔡旭龍
研究生(外文):Hsu-Lung Tsai
論文名稱:以混合式基因演算法訓練之類神經診斷系統
論文名稱(外文):The Diagnostic System by incorporating a Genetic Algorithm and ANN
指導教授:黎文龍黎文龍引用關係
口試委員:李福星曾百由
口試日期:2005-07-20
學位類別:碩士
校院名稱:國立臺北科技大學
系所名稱:機電整合研究所
學門:工程學門
學類:機械工程學類
論文種類:學術論文
論文出版年:2005
畢業學年度:93
語文別:中文
論文頁數:81
中文關鍵詞:頻譜訊號結構阻尼訊號基因演算法類神經網路
外文關鍵詞:frequency spectrumstructure dampingGenetic AlgorithmANN
相關次數:
  • 被引用被引用:2
  • 點閱點閱:188
  • 評分評分:
  • 下載下載:7
  • 收藏至我的研究室書目清單書目收藏:3
本研究是以類神經網路為基本架構的診斷系統,並以管線結構振動訊號來當作診斷輸入,而輸入的診斷特徵包括:(1)時域訊號的高階統計動差、(2)經快速傅立葉轉換後之頻域訊號及(3)結構阻尼訊號,當作萃取特徵的方式。本研究所建立的診斷系統,是從振動訊號擷取與訊號前處理,乃至於管線結構的診斷為一完整的系統,並不須外加其他輔助軟體。研究過程中,在診斷訊號擷取的部分,乃是針對凸緣螺栓鬆脫的個數進行實驗驗證,擷取不同狀態下結構的振動訊號,當作診斷系統之輸入訊號,並用上述的特徵輸入網路進行訓練。基本上,訓練是以倒傳遞類神經網路為訓練機制,但於多種症狀訊號混和時,訓練的學習收斂情況較差,因此,本研究也另加入基因演算法來輔佐訓練,經測試後發現:基因演算法對於混合多種型態的範例有較佳的訓練結果。
此外,本研究是屬於分類診斷的問題,故在回想過程中發現,單一的基因演算機制可能有過度學習的情形,甚至可能比倒傳遞神經網路訓練還嚴重,本研究將兩者結合,以基因演算為先找尋較好的初始值,再代入倒傳遞網路訓練,故融合了兩者優點的效果。研究驗證的結果發現,以此混合方式診斷系統可增加診斷系統之判別能力,且對訓練後之管線診斷系統也有相當助益,此點已經由本研究之線上即時診斷實驗中,被進一步證實了。
A diagnostic System based on the ANN (artificial neural network) and vibration signal inputs is developed for a piping system in this study. We utilize three methods to extract the features of the piping system. They are (1) the statistic moments of higher orders, (2) the harmonic peaks of the frequency spectrum, and (3) the structure damping signals of the vibration signal. The diagnostic system is a standalone system without any other auxiliary software. The major functions of the diagnostic system include generating vibration symptom signals, pre-processing the signals, diagnosing the piping system. In this study, the experiment focuses on the looseness of the flanged joints in which they always exist due to the unbalanced vibration from motors or pumps. Taking the advantages of this operation signals, the current study take them as the ANN inputs so that the ANN can be verified. On the whole, this study used the back propagation network (BPN) for the main structure of ANN. However, it has been found that the BPN alone tends to raise the problem of slow convergence during the ANN training process. And, it even worse that the BPN may not be able to converge if a poor learning rate is set. For this reason, a genetic algorithm (GA) is purposely added to alleviate the short comings of the BPN. The present study has substantiated that the added GA works much better than that of the BPN alone.
This study belongs to the problem of system classification. In order to let the synthesizied GA works properly, the study uses the following steps : First, we used the GA to evolve the ANN’s weight matrices as to search the best initial values. Next, we put the weight matrices back to BPN training. Keep on running until the error level runs below the ending condition. The GA+BPN mode has been proved that it can incorporate the advantages of two traditions methods. The real-time experiments diagnosis results can further verify that the method is feasible and increasing the system’s diagnostic ability. Thus, the current system may be applied to the industries.
中文摘要.................................................i
英文摘要................................................ii
誌謝...................................................iii
目錄....................................................iv
表目錄..................................................vi
圖目錄.................................................vii
第一章 緒論.............................................1
1.1研究背景與動機...................................1
1.2研究方法與步驟...................................2
1.3研究範圍與限制........................................4
第二章 文獻探討.........................................7
2.1預知保養技術.....................................7
2.2類神經網路概論...................................8
2.3倒傳遞神經網路...................................9
2.3.1 訓練學習流程...................................12
2.3.2 訓練回想流程...................................15
2.3.3 倒傳遞神經網路參數...............................15
2.4基因演算法概論...............................16
2.4.1 基本運算子.......................17
2.4.2 基因演算法參數........................18
2.5基因演算法訓練類神經網路.....................20
2.5.1 基因元編碼..............................20
2.5.2 適應值的判別............................22
2.5.3 收尋終止條件............................22
2.5.4 排名法選取..............................23
2.5.5 實數值交配與突變........................25
第三章 診斷系統與振動訊號處理..........................27
3.1系統架構.....................................27
3.1.1 管件結構................................27
3.1.2 結構激振實驗...................29
3.2實驗之硬體設備...............................33
3.2.1 激振系統..............................33
3.2.2 量測系統..............................34
3.3實驗之軟體設備...............................36
3.3.1 訊號處理..............................36
3.3.2 特徵擷取..............................38
3.3.3 特徵輸入向量..........................47
3.4人機介面.....................................47
第四章
4.1以田口實驗法設計基因演算法參數...................52
4.1.1 田口法實驗結果...........................53
4.1.2 實驗驗證.........................................55
4.2 診斷測試....................................56
4.2.1 測試法則.............................56
4.2.2 測試結果.............................58
4.3 混合式演算法...............................60
4.3.1 連結方式.............................62
4.3.2 Pocket Algorithm.........................63
4.3.3 混合式演算法驗證.........................63
4.4 所有特徵組合實驗....................................64
4.5 線上即時診斷........................................67
第五章 結論與建議 ................................70
參考文獻..............................................72
附錄
A 振動複頻訊號產生器規格............................74
B 激振控制系統規格..................................77
[1] 力勝公司,http://leadsurf.myweb.hinet.net/振動分析通報。
[2] A. K. Pandy and M. Biswas, "Experimental Verification of Flexibility Difference Method for Locating Damage in Structures," Journal of Sound and Vibration, vol. 184, no. 2, 1995, pp.311-328.
[3] S. Salawu, "An Integrity Index Method for Structural Assessment of Engineering Structures Using Modal Testing, " INSIGHT, vol. 39, no. 1, 1997, pp.33-37.
[4] Wenlung Li and C.C. Tsaur, "An Energy Method to determine the Locations of Supports for Piping System under Dynamic Loadings," The 27th Conference on Theoretical and Applied Mechanics, Tainan, Taiwan, R.O.C., 2003, pp.1993-1200.
[5] Welung Li, "Continuous Evaluation of Viscous Damping by the Modulation of Steady-State Response, " Proceedings of the 20th National Conference on Mechanical Engineering The Chinese Society of Mechanical Engineers, Taipei, Taiwan, R.O.C., 2003, pp. 553-558.
[6] 陳健源,「類神經網路與其應用簡介」,檢測科技,十四卷,四期,1996,第229-237頁。
[7] M. Inamdar, "Development of an Artificial Neural Network to predict Springback in Air Vee Bending," The International Journal of Advanced Manufacturing Technology, vol. 16, 2000, pp.376-381.
[8] 黎文龍、高志儒、邱垂霖,「多機環境下類神經網路診斷專家系統輸入特徵之研究」,中國機械工程學會第十七屆全國學術研討會論文集,高雄,2000,第629-636頁。
[9] Patrick H. Vieth, "Integrity-Verification Methods Support US Efforts in Pipeline Safty, " Oil & Gas Journal, vol. 100, no. 51, 2002, pp.52-64.
[10] 葉斯泉,預知保養-振動診斷分析,機械月刊,期別0266,1997
[11] 張顯盛,以類神經建立即時振動診斷系統之研究,碩士,國立國立臺北科技大學機電整合研究所,台北,1998。
[12] 邱垂霖,以聲音訊號為特徵輸入之類神經故障診斷系統,碩士,國立國立臺北科技大學機電整合研究所,台北,2001。
[13] 周光威,以類神經診斷管線凸緣鬆脫之研究,碩士,國立國立臺北科技大學機電整合研究所,台北,2004。
[14] 蘇木村、張孝德,機械學習:類神經網路、模糊系統以及基因演算法則,台北,全華,1999。
[15] Vitaly Schetinin,"A Neural Network Decision Tree for Learning Concepts from EEG Data," Advanced Study Institute on Neural Networks for Instrumentation, Measurement, and Related Industrial Applications: Study Cases Crema, Italy, 9-20 October 2001
[16] 廖鴻翰,以基因演算法建構類神經網路模型,碩士,大葉大學電機工程研究所,1998。
[17] 許智淵,基因演算法於類神經網路之應用,碩士,成大航太研究所,2000
[18] 楊武智,感測器與數位信號處理,台北,全華科技,1999
[19] 共和電業,共和電子計測器,日本,三聯科技,2002。
[20] 陳正義、何坤鑫、程啟正,Visual Basic程式設計與圖形監控應用,台北,滄海書局,2002。
[21] 金頓科技,振動複頻訊號產生器操作手冊,台北,金頓科技公司,2002。
[22] 台灣微軟開發工具研發小組,Visual Basic 6.0程式開發手冊,台北:松崗電腦圖書,1999,第1-1-17-39頁。
[23] 彭明柳,Visual Basic 中文專業版徹底研究,台北:博碩文化圖書,2000,第1-1-19-61頁。
[24] 王柏村,振動學,台北:全華科技圖書,2001,第1-1-6-43頁。
[25] 葉怡成,類神經網路模式應用與實作,台北,儒林圖書有限公司,2001,第1-1-4-51頁。
[26] Kenneth G. McConnell, Vibration Testing: Theory and Practice, New York: John Wiley & Sons, INC. 1998, pp596-634.
[27] 張顯盛、黎文龍,「輸入特徵對類神經網路診斷系統之鑑別影響」,大安高工學報,第十期,第148-157頁。
[28] 黎文龍,工程設計與分析,台北,東華書局,2000,第10-4-325頁。
[29] 田口玄一,田口統計解析法,台北,五南圖書,2003
[30] Whitley D. , Starkweather T. and Bogart C. ,"Genetic Algorithm and Neural Networks : Optimizing Connections and Connectivity, "Parallel Computing, Computing,14,1990,pp.347-361
[31] Kim K.J. and I. Han , " Genetic Algorithm approach to feature discretization in artificial neural network for the prediction of stock price index, "Expert system with application,Vol.19,pp.1250-132,2000.
[32] 周慶華,整合基因演算法及類神經網路於現貨開盤指數之預測-以新加坡交易所摩根台股指數期貨為例,碩士,輔仁大學,2001
[33] 徐英哲,混合基因演算法於流程型工廠排程研究-以TFT-LCD彩色濾光片廠為例,碩士,台北大學,2005
連結至畢業學校之論文網頁點我開啟連結
註: 此連結為研究生畢業學校所提供,不一定有電子全文可供下載,若連結有誤,請點選上方之〝勘誤回報〞功能,我們會盡快修正,謝謝!
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top