以數據分析方法預測打擊樁於軸向載重下之承載力

Abstract

本研究旨在通過數據分析方法預測打擊樁在承受軸向載重時的承載能力表現。將收集來自全球各地的打擊樁資料和相關場址數據，並將這些真實數據存儲在一個新建的資料庫中。通過對這些數據的深入分析，建立了一個多變數機率分布模型，以預測打擊樁的極限承載力及其在軸向載重下的受力行為。 在文獻回顧階段，識別並收集了與打擊樁在軸向載重下承載能力相關的各種參數。在資料庫構建完成後，依據先前文獻中提出的經驗公式進行了初步檢核，以觀察數據趨勢並確認所收集數據的有效性。 該資料庫包含了多種與極限承載力相關的參數及現場試驗數據，包括SPT標準貫入試驗和CPT圓錐貫入試驗的結果。主要參數包括：樁的長度（L）、樁的直徑（B）、極限承載力（𝑞𝑢𝑙𝑡）、SPT-N值和CPT-qc值。 採用了階層式貝氏模型（HBM）及改良後的階層式貝氏模型(RHBM)，並結合Johnson分布系統、吉布斯抽樣和貝氏分析，來研究不同土壤參數之間的相關性並期望得到打擊樁資料庫中場址統計特性。同時利用目標場址的有限已知數據推估未知數據的分布。結果表明，隨著已知信息的增加，階層式貝氏模型能有效降低對打擊樁承載力預測的不確定性，若進一步應用於可靠性設計，可顯著提高精準度並降低成本。

This study aims to predict the bearing capacity performance of driven piles under axial loading using data analysis methods. Data on driven piles and relevant site conditions will be collected from around the world and stored in a newly established database. Through thorough analysis of this data, a multivariate probabilistic model has been developed to predict the ultimate bearing capacity of driven piles under axial loading. During the literature review phase, various parameters related to the bearing capacity of driven piles under axial loading were identified and collected. After the database construction, preliminary checks were conducted based on empirical formulas proposed in previous literature to observe data trends and confirm the validity of the collected data. The database contains various parameters related to ultimate bearing capacity and field test data, including results from cone penetration tests (CPT) and standard penetration tests (SPT). The main parameters include pile length (L), pile diameter (B), ultimate bearing capacity (𝑞𝑢𝑙𝑡), SPT-N values, and CPT-qc values. The Hierarchical Bayesian Model (HBM) was employed, combined with the Johnson distribution system, Gibbs sampling, and conjugate conditions in Bayesian analysis, to study the correlations between parameters and capture the statistical behavior of the sites within the database. Concurrently, limited known data from target sites were used to estimate the distribution of unknown data. The results indicate that as the amount of known information increases, the HBM effectively reduces the uncertainty in predicting the bearing capacity of driven piles. If further applied in reliability design, this approach can significantly enhance accuracy and reduce costs.