以資料庫回歸台灣混凝土收縮與潛變預測模型並應用於預力橋梁長期變位分析

Regression of Shrinkage and Creep Prediction Model of Concrete in Taiwan Based on Database Analysis and Application to Long-term Deflection Analysis of Prestressed Concrete Bridge

黃禾程(Ho-Cheng Huang)

指導教授 ： 廖文正

國立臺灣大學/工學院/土木工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202003020

摘要

進行混凝土結構物設計、興建與維護管理時，需考量強度、變形及耐久性以達到長期安全及服務性，然而實務上多未能準確考量收縮和潛變所造成的影響。國內外對於混凝土收縮與潛變已有長期的試驗與研究發展，且也提出多種收縮及潛變預測模型。台灣於2017年建立了「台灣混凝土潛變收縮資料庫」，並選用Model B4為基礎提出了本土化收縮潛變預測模型Model B4TW (2017)。
Model B4TW (2017) 雖然已針對台灣多項混凝土特性進行修正，包含高膠結材含量、低彈性模數與低粒料勁度等，但仍未考慮骨材含量與對預測模型中各項水泥相關參數進行本土化修正。為對前述進行修正並使TW資料庫更加完善，本研究在「台灣混凝土潛變收縮資料庫」中新增了多項欄位，包含細粒料量、粗粒料量、細粒料比重、粗粒料比重、爐石細度與飛灰種類等，期望提出更準確且更符合物理意義的預測模型Model B4TW (2020)。
本研究使用Python 建立分析方法並選用 Microsoft SQL (MSSQL) 做為資料庫管理系統，而非傳統之 Excel 或者 Access，其原因在於透過 SQL 語法之可攜性可輕易進行資料篩選與分析，且較適用於處理大量數據，效率遠勝傳統資料庫分析方法。近年來資料科學發展興盛，本研究透過使用多種機器學習演算法對台灣收縮與潛變數據進行回歸與預測，期望提出更準確的預測模型。為方便工程師以及各國學者在不需進行任何複雜操作以及程式碼處理的情況下使用本研究所提出的各項成果，本研究透過將資料庫雲端化，並使用ASP.net建置線上即時分析網頁S.C.D.T (Shrinkage and Creep Database in Taiwan)，使用者僅須在網頁上輸入所需參數便可以迅速得到各模型收縮與潛變預測成果。
造成預力混凝土橋梁長期變形的原因，主要可歸因於混凝土收縮與潛變所引致的預力損失，目前各國學者與規範所提出之收縮與潛變模型在預測台灣混凝土時皆呈現低估的現象。為探討並比較國內外各模型應用於橋梁長期變位分析之差異，本研究選用多種混凝土收縮潛變預測模型，包含ACI 209R-92、AASHTO LRFD 2014、CEM MC90、CEB MC10、GL2000、Model B3以及Model B4TW (2020)等，使用MIDAS Civil建立預力箱型橋梁模型進行分析。比較分析結果後發現，若直接套用國外預測模型，將嚴重低估橋梁長期變位與預力損失，導致使用年限高估。本研究期望發展出一套適合台灣混凝土預力橋梁長期變形的分析模式，使工程界能更準確掌握台灣混凝土橋梁長期變形量以及評估使用年限，利於安全監測及適時進行維護，以避免災害發生並延長混凝土結構物的生命週期。

關鍵字

混凝土 ； 卜作嵐材料 ； 收縮 ； 潛變 ； 資料庫 ； 機器學習 ； 預力混凝土橋 ； 長期變形 ； 預力損失

並列摘要

Under design, construction and maintenance of concrete structures, strength, deformation and durability must be taken into consideration in order to reach long-term safety and serviceability. However, the effect of shrinkage and creep is rarely considered in practice. Experiments and researches on shrinkage and creep of concrete have been developed over decades all around the world. Numerous shrinkage and creep prediction models have been proposed. In 2017, Shrinkage and Creep Database of Concrete in Taiwan (TW Database) was established. A localized prediction formula, Model B4TW (2017) was proposed to predict shrinkage and creep of concrete in Taiwan.
Model B4TW (2017) takes Model B4 as its base model and modifies with local characteristics of concrete in Taiwan, including high cementitious material content, low concrete elastic modulus and low aggregate stiffness. Even though Model B4TW (2017) already has considerable high prediction accuracy, modification on cement related coefficient and the effect of aggregate content were not yet considered. To modify on aforementioned and integrate TW database, this study adds several parameters, including coarse aggregate content, fine aggregate content, specific gravity of aggregate, slag type and fly ash type into TW database; looking forward to proposing the Model B4TW (2020) a prediction model with higher accuracy and satisfied physical principles..
This study uses Microsoft SQL as database management system and Python to establish all processing procedures. The reason why Microsoft SQL was selected rather than traditional software such as Excel and Access is that SQL is highly portable and more suitable when handling large datasets. By combining Microsoft SQL and Python, the efficiency of analysis can be greatly enhanced. In recent years, data science has gained popularity all over the globe. This research selected several machine learning algorithms to predict shrinkage and creep of concrete; expecting to propose a prediction model with even higher accuracy. In order to enable engineers and scholars to make use of the result proposed by this research without the need of complicated calculation and ability of coding, an online analysis webpage S.C.D.T (Shrinkage and Creep Database in Taiwan) was established using ASP.net. By inputting parameters required, users can easily and quickly obtain prediction results of shrinkage and creep calculated by each model.
The cause of long-term deflection of prestressed concrete bridges are mainly attributed by tendon loss induced by shrinkage and creep of concrete. All prediction models available at present underestimate shrinkage and creep strain of concrete in Taiwan. In order to investigate and compare long-term behavior of prestressed concrete bridge, this research applies multiple prediction models, including ACI 209R-92、AASHTO LRFD、CEM MC90、CEB MC10、Model B3 and Model B4TW (2020) to prestressed concrete box girder bridge model established by MIDAS Civil. Analysis results show that if foreign prediction model is directly applied in designing prestressed concrete bridge in Taiwan, maximum deflection of the bridge will be underestimated and service life will be overestimated. This research developed a long-term deformation analysis which enables engineers to accurately calculate long-term deflection, tendon loss and service life of prestressed concrete bridges in Taiwan. The Analysis results also enable engineers to monitor and provide retrofit when needed, extend the service life of structure during the design and maintenance phase to achieve sustainability.

並列關鍵字

Concrete ； Pozzolanic Material ； Shrinkage ； Creep ； Database ； Machine Learning ； Prestressed Concrete Bridge ； Long-term Deflection ； Tendon Loss

參考文獻

[1]AASHTO, 2014, “AASHTO LRFD Bridge Design Specifications. 7th ed,” American Association of State Highway and Transportation Officials (AASHTO), Washington DC

[2]ACI Committee 209, 1978,” Prediction of Creep, Shrinkage and Temperature Effects in Concrete Structures,” ACI, Detroit, pp.98.

[3]ACI Committee 209, 1982, “Prediction of Creep, Shrinkage and Temperature Effects in Concrete Structures,” Designing for Creep and Shrinkage in Concrete Structures, A Tribute to Adrian Pauw, SP-76, American Concrete Institute, Farmington Hills, MI, pp. 193-300.

[4]ACI Committee 209, 1992, “Prediction of Creep, Shrinkage, and Temperature Effects in Concrete Structures (ACI 209R-92),” American Concrete Institute, Farmington Hills, MI, 47 pp.

[5]Bažant, Z.P., and Baweja, S., 1995, “Creep and Shrinkage Prediction Model for Analysis and Design of Concrete Structures: Model B3,” Materials and Structures, V. 28, pp. 357-365, 415-430, 488-495.