簡易檢索 / 詳目顯示

回結果列表
研究生： 陳曦
Hsi Chen
論文名稱： 分散式計算與代理模型於大地工程反算分析之應用初探
A Preliminary Study of Geotechnical Back Analysis Using Distributed Computing and Surrogate Modeling
指導教授： 謝佑明
Yo‐Ming Hsieh
口試委員: 陳鴻銘
楊元森
學位類別： 碩士
Master
系所名稱： 工程學院 - 營建工程系
Department of Civil and Construction Engineering
論文出版年： 2018
畢業學年度： 106
語文別： 中文
論文頁數： 192
中文關鍵詞： 反算分析 、代理模型 、克利金法 、分散式計算
外文關鍵詞： Back Analysis, Surrogate Model, Kriging Interpolation, Distributed System
相關次數： 點閱：282  下載：14 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報
  •  上一筆

    • 在大地工程反算分析中，真實模型之計算耗時不菲，整體運算時間冗長。過去研究提出以計算代價較低、行為與真實模型接近的代理模型(Surrogate Model) 進行反算以降低運算時間。典型代理模型之建立使用模式為先訓練、再驗證、最後取代真實模型。但無論是在訓練階段或是驗證階段，都需要大量動用真實模型，導致準備代理模型的過程非常冗長。本研究使用克利金法建構新的代理模型運作機制，並結合分散式運算技術以加快整體反算分析之過程。
    本研究複合粒子群演算法與梯度下降法作為反算分析中的最佳化演算法，並實作克利金法做為代理模型。本研究所提出的代理模型建構機制創新利用克利金變異數來檢驗代理模型之信心水準。當代理模型顯示出一定的信心水準時，將用代理模型取代真實模型來計算結果。為了進一步加快整體之效率，本研究還使用了Node.js搭配Rabbit Message Queue，實作了一套容易擴展運算效能的分散式系統。透過將真實模型地計算交給眾多的運算節點，來進一步地降低整體反算分析時間。經過實測，本研究所開發的演算法與純粒子群演算法比較起來，有著更快的收斂速度以及更少的真實模型使用次數。

    Back-analysis in geotechnical engineering is time consuming due to long calculation time of real models. Past researches proposed the use of surrogate models which approximate real models with lower computational cost in back-analysis to reduce computation time. The typical pattern of surrogate modeling includes three steps: first the training, second the validation and the last, replacing real model. However, both training and validation phases require many evaluations of the real model. As a result, the process of constructing the surrogate model needs a long time to complete. This research developed a new strategy of surrogate-based modeling by using Kriging interpolation and employed distributed computing technique to speed up the process of back-analysis.
    This research combines particle swarm optimization (PSO) algorithm and gradient descent method to be the optimization algorithm in back analysis. The Kriging method is used as the surrogate model. The innovation of the proposed surrogate modeling technique uses the Kriging variance to estimate the confidence level of the Kriging surrogate model. When the surrogate model has acceptable level of confidence, the surrogate model will replace the real model to calculate the result, and vice versa. In order to further accelerate the overall efficiency, this research uses Node.js with Rabbit Message Queue to implement a scalable back-analysis system that evaluations of the real model are done on computing nodes, resulting in reduced overall computational time. It is demonstrated that the proposed algorithm converges faster than generic PSO, and real model evaluations is reduced by the surrogate model.

    論文摘要 I ABSTRACT III 致謝 V 目錄 VII 圖目錄 XI 表目錄 XIX 第一章 緒論 1 1.1 研究動機與目的 1 1.2 研究流程 4 1.3 論文架構 6 第二章 文獻回顧 7 2.1 反算分析於大地工程之應用 7 2.2 代理模型 7 第三章 研究方法與工具 11 3.1 粒子群演算法 11 3.2 克利金內插法 15 3.2.1 克利金法之基本形式 15 3.2.2 簡單克利金法 18 3.2.3 一般克利金法 21 3.2.4 通用克利金法 22 3.2.5 半變異圖與半變異數模型 24 3.2.6 克利金法之流程 27 3.3 梯度下降法 29 3.3.1 基本梯度下降法 30 3.3.2 共軛梯度下降法 30 3.4 複合演算法 32 3.5 Rabbit Message Queue 33 3.5.1 Message-Oriented Middleware 33 3.6 Node.js 35 3.7 Intel Math Kernel Library 35 第四章 系統分析與設計 37 4.1 使用案例圖 37 4.2 佈署圖 38 4.3 反算分析系統流程圖 41 4.4 類別圖 47 4.4.1 粒子群演算法 49 4.4.2 克利金法模組 50 4.4.3 梯度下降法模組 52 4.4.4 MessageMiddleware類別 53 第五章 克利金內插法之分析與驗證 54 5.1 內插驗證 54 5.1.1 二維函數內插之驗證 55 5.1.1.1 均勻點位驗證 55 5.1.1.2 隨機點位驗證 73 5.1.2 三維函數內插之驗證 90 5.1.2.1 均勻點位驗證 91 5.1.2.2 隨機點位驗證 105 5.1.3 誤差比較 119 5.2 克利金變異數分析 125 5.2.1 克利金變異數與誤差之關係 126 5.2.2 克利金變異數與誤差區間 131 5.3 取樣方法分析 132 5.3.1 LHS取樣法 133 5.3.2 MAS取樣法 137 5.4 小結 139 第六章 系統驗證 141 6.1 案例介紹 141 6.2 測試結果 146 6.3 分散式計算展示 157 6.4 小結 159 第七章 結論與建議 162 7.1 結論 162 7.2 建議 163 參考文獻 166

    Observational method. [cited 2018 1/24]; Available from: https://en.wikipedia.org/wiki/Observational_method_(geotechnics).
    2. Peck, R.B., Advantages and Limitations of the Observational Method in Applied Soil Mechanics. Géotechnique, 1969. 19(2): p. 171-187.
    3. Gioda, G. and G. Maier, Direct search solution of an inverse problem in elastoplasticity: Identification of cohesion, friction angle and in situ stress by pressure tunnel tests. International Journal for Numerical Methods in Engineering, 1980. 15(12): p. 1823-1848.
    4. Anandarajah, A. and D. Agarwal, Computer-aided calibration of a soil plasticity model. International Journal for Numerical and Analytical Methods in Geomechanics, 1991. 15(12): p. 835-856.
    5. Meier, J., et al., Inverse Parameter Identification Technique Using PSO Algorithm Applied to Geotechnical Modeling. Journal of Artificial Evolution and Applications, 2008. 2008.
    6. Zhu, C., H. Zhao, and M. Zhao, Back Analysis of Geomechanical Parameters in Underground Engineering Using Artificial Bee Colony. The Scientific World Journal, 2014. 2014: p. 13.
    7. Jin, Y.-F., et al., Selection of sand models and identification of parameters using an enhanced genetic algorithm. International Journal for Numerical and Analytical Methods in Geomechanics, 2016. 40(8): p. 1219-1240.
    8. Moreira, N., et al., Back analysis of geomechanical parameters in underground works using an Evolution Strategy algorithm. Tunnelling and Underground Space Technology, 2013. 33: p. 143-158.
    9. Papon, A., et al., Single-and multi-objective genetic algorithm optimization for identifying soil parameters. International Journal for Numerical and Analytical Methods in Geomechanics, 2012. 36(5): p. 597-618.
    10. Atkinson, K.E., An Introduction to Numerical Analysis. 2 ed. 1989: Wiley. 693.
    11. Goovaerts, P., Geostatistics for Natural Resources Evaluation. 1997: Oxford University Press. 483.
    12. Broomhead, D.S. and D. Lowe, Multivariable Functional Interpolation and Adaptive Networks. Complex Systems, 1988. 2: p. 321--355.
    13. Broad, D.R., G.C. Dandy, and H.R. Maier, Water distribution system optimization using metamodels. Journal of Water Resources Planning and Management, 2005. 131(3): p. 172-180.
    14. Lee, J. and J. Kim, Approximate optimization of high-speed train nose shape for reducing micropressure wave. Structural and Multidisciplinary Optimization, 2008. 35(1): p. 79-87.
    15. Forrester, A.I.J. and A.J. Keane, Recent advances in surrogate-based optimization. Progress in Aerospace Sciences, 2009. 45(1): p. 50-79.
    16. Braconnier, T., et al., Towards an adaptive POD/SVD surrogate model for aeronautic design. Computers & Fluids, 2011. 40(1): p. 195-209.
    17. Dolci, V. and R. Arina, Proper Orthogonal Decomposition as Surrogate Model for Aerodynamic Optimization. International Journal of Aerospace Engineering, 2016. 2016: p. 15.
    18. Pehlivanoglu, Y.V. and B. Yagiz, Aerodynamic design prediction using surrogate-based modeling in genetic algorithm architecture. Aerospace Science and Technology, 2012. 23(1): p. 479-491.
    19. You, K., A case study on the utilization of tunnel face mapping data for a back analysis based on artificial neural network. KSCE Journal of Civil Engineering, 2014. 18(3): p. 751-759.
    20. Pichler, B., R. Lackner, and H.A. Mang, Back analysis of model parameters in geotechnical engineering by means of soft computing. International Journal for Numerical Methods in Engineering, 2003. 57(14): p. 1943-1978.
    21. Obrzud, R.F., L. Vulliet, and A. Truty, Optimization framework for calibration of constitutive models enhanced by neural networks. International Journal for Numerical and Analytical Methods in Geomechanics, 2009. 33(1): p. 71-94.
    22. Wojciechowski, M., Application of artificial neural network in soil parameter identification for deep excavation numerical model. Computer Assisted Methods in Engineering and Science, 2017. 18(4): p. 303-311.
    23. Zhao, H.-b. and S. Yin, Geomechanical parameters identification by particle swarm optimization and support vector machine. Applied Mathematical Modelling, 2009. 33(10): p. 3997-4012.
    24. Yu, Y., B. Zhang, and H. Yuan, An intelligent displacement back-analysis method for earth-rockfill dams. Computers and Geotechnics, 2007. 34(6): p. 423-434.
    25. Khaledi, K., et al., Robust and reliable metamodels for mechanized tunnel simulations. Computers and Geotechnics, 2014. 61: p. 1-12.
    26. Kleijnen, J.P.C., Kriging metamodeling in simulation: A review. European Journal of Operational Research, 2009. 192(3): p. 707-716.
    27. Eason, J. and S. Cremaschi, Adaptive sequential sampling for surrogate model generation with artificial neural networks. Computers & Chemical Engineering, 2014. 68: p. 220-232.
    28. Liu, H., J. Cai, and Y.-S. Ong, An adaptive sampling approach for Kriging metamodeling by maximizing expected prediction error. Computers & Chemical Engineering, 2017. 106: p. 171-182.
    29. Kennedy, J. and R. Eberhart. Particle swarm optimization. in Neural Networks, 1995. Proceedings., IEEE International Conference on. 1995.
    30. Can someone please explain semivariance (for kriging) to me as simply as possible? [cited 2018 1/20]; Available from: https://www.reddit.com/r/askgis/comments/25a0r1/can_someone_please_explain_semivariance_for/.
    31. Modelling a semi-variogram. [cited 2018 1/29]; Available from: http://spatial-analyst.net/ILWIS/htm/ilwisapp/sec/semivar_models_sec.htm.
    32. Frederickson, B. An Interactive Tutorial on Numerical Optimization. [cited 2018 2/7]; Available from: http://www.benfrederickson.com/numerical-optimization/.
    33. Line search. [cited 2018 2/7]; Available from: https://en.wikipedia.org/wiki/Line_search.
    34. Developer Reference for Intel® Math Kernel Library 2018 - C. 2018 [cited 2018 1/25]; Available from: https://software.intel.com/en-us/mkl-developer-reference-c.
    35. Latin hypercube sampling. [cited 2018 1/20]; Available from: https://en.wikipedia.org/wiki/Latin_hypercube_sampling.
    36. Test functions for optimization. [cited 2018 1/21]; Available from: https://en.wikipedia.org/wiki/Test_functions_for_optimization.

    QR CODE