Please use this identifier to cite or link to this item:
標題: 地質材料參數最佳化及其於地滑行為分析之應用
The Optimization of Geomaterial Parameters and Its Application in the Analysis of Landslide Behavior
作者: 賴丞昶
Lai, Cheng-Chang
關鍵字: parameter optimization
artificial neural networks
出版社: 水土保持學系所
引用: [01] 「地質敏感區災害潛勢評估與監測-重大山崩災害潛勢地區災害模擬與監測(1/4)成果報告書」,經濟部中央地質調查所,中華民國九十七年六月。 [02] 「地質敏感區災害潛勢評估與監測-重大山崩災害潛勢地區災害模擬與監測(2/4)成果報告」,經濟部中央地質調查所,中華民國九十七年十二月。 [03] 「台14線88K至91K 地滑地治理調查規劃工程-成果報告書」,行政院農業委員會水土保持局第三工程所,中華民國九十五年十月。 [04] 「廬山地滑監測及後續治理規劃-期末報告書」,行政院農業委員會水土保持局第三工程所,中華民國九十七年一月。 [05] 葉怡成,「應用類神經網路」,儒林圖書,台北,(2002)。 [06] 張志增、高永濤、張曉平,「邊坡岩體力學參數反分析方法」,北京科技大學學報,第28卷,第12期,Dec.2006。 [07] 汪能君、梧松,「單樁靜載試驗的位移反分析研究」,重慶建築大學學報,第29卷,第2期,Apr.2007。 [08] 李端有、甘孝清,「滑坡體力學參數反分析研究」,長江科學院院報,第22卷,第6期,Dec.2005。 [09] 李端有、李迪、馬山水,「三峽永久船閘開挖邊坡岩體力學參數反分析」,長江科學院院報,1998(2),第10-13頁。 [10] Calvello M, Leonardo Cascini and Giuseppe Sorbino, “A numerical procedure for predicting rainfall-induced movements of active landslides along pre-existing slip surfaces, ” International Journal for Numerical and Analytical Methods in Geomechanics (in press). [11] Calvello M, Finno RJ. “Selecting parameters to optimize in model calibration by inverse analysis, ” Computers and Geotechnics 2004; 31(5):411–425. [12] Finno RJ, Calvello M., “Supported excavations: the observational method and inverse modeling, ” ASCE Journal of Geotechnical and Environmental Engineering 2005;131(7):826–836. [13] Karanagh K, Clough R W., “Finite Element Application in the Characterization of Elastic Solids DJ, ” Solids Structures, 1971 (7): 11-13. [14] Keidser A, Rosjberg D. “A comparison of four inverse approaches to groundwater flow and transport parameter identification, “ Water Resources Research 1991;27(9):2219–2232. [15] K.M. Neaupane, S.H. Achet, “Use of backpropagation neural network for landslide monitoring: a case study in the higher Himalaya, ” Engineering Geology, 74 (2004) 213–226. [16] Ou CY, Tang YG. “Soil parameter determination for deep excavation analysis by optimization, ” Journal of the Chinese Institute of Engineers 1994; 17(5):671–688. [17] Poeter EP, Hill MC. “Documentation of UCODE, a computer code for universal inverse modeling, “ U.S. Geological Survey Water-Resources Investigations Report 98-4080, 1998. [18] Poeter EP, Hill MC. “Inverse methods: a necessary next step in groundwater modeling, ” Ground Water 1997; 35(2):250–260. [19] Rosenblatt, F., “The Perceptron : A Probabilistic Model for Information Storage and Organization in the Brain, ” Psychological Review, Vol. 65, 1958, pp.386-408. [20] Rumelhart, D. E., Hinton, G. E., and Willams, R. J., “Learning Internal Representations by Error Propagation, ” Parallel Distributed Processing, Chapter 8, MIT Press,1986. [21] SangGi Hwang, Ivy F.Guevarra, ByongOk Yu, “Slope failure prediction using a decision tree: A case of engineered slopes in South Korea ” Engineering Geology, 104 (2009) 126–134. [22] Th.W.J. van Asch, L.P.H. Van Beek, T.A. Bogaard, “Problems in predicting the mobility of slow-moving landslides ” Engineering Geology, 91 (2007) 46–55.
摘要: 數值分析是利用電腦的快速運算及便利性的優點來縮短工程問題分析上的時間及解決複雜的問題。邊坡問題在進行數值分析時,須對需解決的問題建立地質模型、邊界條件和一些基本的參數設定,往往這些參數的設定是否具有代表性,仍是個疑問。參數的給定與數值模擬結果的準確性有密切的關係,以往藉由經驗及文獻等較為主觀的方式決定地質材料參數,其客觀度欠缺,如有較為客觀的方法找出具有代表性的參數,那將使數值分析的結果更具價值。本研究利用觀測值與UCODE程式及類神經網路的方法來校正參數,使參數達到最佳化,使模擬結果與觀測值更為接近。UCODE程式為一種統計原理的方式來修正參數的軟體,類神經網路則是在預測、分類、最佳化等方面上都有不錯的效果,其中的倒傳遞神經網路更適合最佳化的問題。 在分析中,除了利用上述兩種方式加以校正參數外,再利用人工目測調整參數的方式加以比較。在三軸排水壓縮試驗中,參數E50ref在經由人工目測調整及UCODE程式校正後,其參數值皆為向上修正,比起始參數值來的大上許多,唯有在類神經網路模擬這部分為向下修正些許;參數m之變動大都在0.9上下小幅變動;參數Φ之校正結果皆比起始參數值來的大,皆是向上修正,且在三軸試驗的分析中,發現Φ值對整體分析的影響性最大。在廬山地滑分析中,參數E50ref的變動無一規則,而參數C與Φ不論在何種方式的參數校正上,皆為向下修正,起始參數值有可能是有高估的情形,且同樣地發現參數Φ在廬山地滑的分析上,也是對分析結果影響最大的參數。由上可知在兩種案例的分析上,參數Φ是最具影響力的參數,且不管在三軸排水壓縮試驗或廬山地滑的分析中,經過UCODE程式及類神經網路修正後之參數,在數值分析上皆可改善分析的結果。 在廬山分析研究中,數值分析結果若要再進一步接近觀測值,建議對於地層分布情況或組成模式加以改善,相信對分析結果的準確度可改善不少。
Numerical analysis takes advantage of the fast calculation and convenience of computers in solving complicated engineering problems. It is down with the setup of a geological model, boundary conditions, and material parameters, numerical results closely depend on the material parameters, which are not easy to determine, however. While they are often subjectively decided by experience or from literature, objective methods should be used to find representative parameters to render more accurate numerical results. This study uses observation data, the program UCODE, and artificial neural networks (ANN) to obtain optimum parameters. The UCODE is based on statistics, and the artificial neural networks are capable of prediction, classification and optimization. In particular, the back-propagation network is suitable for the problems of optimization. In addition to the UCODE and ANN, parameters are adjusted artificially for comparison. In the analysis of the triaxial compression test, the value of the parameter E50ref is raised from the initial one after the calibration by artificial method and the UCODE, while it is dropped by the ANN method. The parameter m varies around 0.9. The parameter Φ is all raised from initial values after the calibration of the three methods, and it has the greatest influence on the analysis. In the analysis of the Lushan landslide, the values of the parameters C and Φ are all raised from initial values by the three methods. Similarly, the parameter Φ has the greatest influence on the numerical results. The UCODE and ANN methods show their ability to optimize the parameters and improve the numerical results in two cases of the triaxial compression test and the Lushan landslide. In the Lushan case, further improvement of the numerical results may require better geological and constitutive models.
其他識別: U0005-1708200917581600
Appears in Collections:水土保持學系



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