多版塊接縫式混凝土鋪面之結構反應分析與驗證

Structural Response Analysis and Verification of Multi-Slab Jointed Concrete Pavements

李育昇(Yu-Sheng Lee)

指導教授 ： 李英豪

淡江大學/工學院/土木工程學系碩士班/碩士(2008年)

https://doi.org/10.6846/TKU.2008.00510

摘要

論文提要內容： 接縫式混凝土的結構反應的判定是設計及結構估算的上不可或缺的一個步驟。各種使用有限元素法程式ISLAB2000 所建構的類神經網路模式被鋪面設計導覽手冊(NCHRP 1-37A)介紹用來估算臨界應力及撓度。然而使用其原先所使用的設定參數來建立資料庫及進行類神經網路的訓練及測試的話需要耗費龐大的時間及精力。因此，本研究主要目的為發展使用因次分析以及當代迴歸技術的多版塊接縫式混凝土鋪面結構反應的預測程序。 首先為了證明ILSAB2000 程式所得結果之正確性，與知名的ILLI-SLAB有限元素程式使用相同參數、條件來進行一系列的分析，最後由結果得知差異不大。而為了擴展現有邊緣應力預測模式的適用性，本研究多考慮了版尺寸效應、輪軸載重型式、版塊相鄰及溫差效應所造成的影響。並對建立多個小資料庫再個別使用投影追逐迴歸法進行預測後再進行疊加的方式及直接建立大資料然後使用小區域迴歸來直接預測的方式進行比較，以期對舊有預測模式進行改善。 最後為了增加預測的準確度，本研究把許多複雜的預測模式整合成一直接使用迴歸技術的預測模式，如此，可以避免以往使用疊加原理時，可能因為多個預測模式相乘時，每個預測模式的估算誤差也相乘而導致應力估算結果之誤差變大。此外，本研究最後採用I-Miner 軟體來整合，以便使用者更容易進行邊緣應力的預測，並經由幾個實例的運算來證明本應力分析程序的有效性。

關鍵字

接縫式混凝土 ； 剛性鋪面 ； 荷重傳遞 ； 溫差 ； 有限元素法 ； 因次分析 ； 迴歸分析

並列摘要

Abstract:
Determination of critical structural responses in terms of stresses and deflections in a jointed concrete pavement is essential to mechanistic-based design and structural
evaluation procedures. Various Artificial Neural Networks (ANN) models developed based on the ISLAB2000 finite element (F.E.) model have been utilized to compute critical stresses and deflections in the recommended Mechanistic-Empirical Pavement Design Guide (NCHRP 1-37A). Tremendous amount of time and efforts are often required in generating the databases and conducting network training and testing using original pertinent input
parameters. Thus, the main objective of this study is to develop alternative prediction models for critical structural responses of multi-slab jointed concrete pavements using the principles of dimensional analysis and modern regression techniques.
The applicability of ISLAB2000 program was first validated through a series of factorial F.E. runs by reproducing almost identical results using the well-known ILLI-SLAB F.E. program. To expand the applicability of existing edge stress prediction models, various factors including the effects of finite slab sizes, gear configurations,adjacent slabs, and thermal curling were considered. Many series of F.E. factorial runs over a wide range of pavement designs under edge loading conditions have been conducted.
Accordingly, various prediction models were separately developed using Projection Pursuit Regression (PPR) together with spline regression as well as local regression (LOESS) techniques. Favorable agreements were achieved when comparing the newly developed versus the existing edge stress prediction models.
To increase the prediction accuracy, more complicated models have also been developed for direct prediction using local regression techniques. In such an approach,
the principle of superposition is no longer adopted to avoid possible magnification effects of prediction errors when multiplying two or more prediction models together.
Furthermore, the proposed approach has been implemented in the Insightful Miner software program to facilitate critical edge stress prediction. Several case studies were
provided to illustrate and validate the proposed critical edge stress analysis procedure.

並列關鍵字

Jointed Concrete Pavement ； Rigid Pavement ； Load Transfer ； Thermal Curling ； Finite Element Method ； Dimensional Analysis ； Regression Analysis

參考文獻

3. Brill, D.R., “Development of Advanced Computational Models for Airport Pavement Design,” DOT/FAA/AR-97/47., 1998.

Transportation Research Board, National Research Council, Washington, D.C., 1994, pp. 101-113.

-Based Backcalculation Program for Composite Pavements,” Transportation Research Record 1570, 1997, pp. 143-150.

10. Ceylan, H., E. Tutumluer, and E. J. Barenberg, “Artificial Neural Networks as Design Tools in Concrete Airfield Pavements Design.”

11. Ceylan, H., E. Tutumluer, and E. J. Barenberg, “Analysis of Concrete Airfield Slabs Serving the Boeing 777 Aircraft Using Artificial Neural Networks.” Journal of the Transportation Research Board, Transportation Research Record 1684, 1999, pp. 110-117.