Comparison among simplified reliably-base design methods via case analyses

可靠度分析設計方法中以蒙地卡羅模擬(Monte Carlo simulation, MCS)最具代表性，此法需要大量樣本以降低偏差，須時甚久，因此發展出一階可靠度法(First-Order Reliability Method, FORM)，但因其某些基本假設條件使設計的精確度稍有不足。有鑑於截至目前為止所發展之可靠度設計方法的各種難處，而想驗證分位數值理論法(Quantile Value Method, QVM) [1]是否能夠優越的完成可靠度設計，以期能於大地工程規範中為納用，而能使大地工程設計規範更為進步。 本文主要討論分位數理論法(QVM)，並用簡單的大地工程設計案例來驗證此法的穩定性，及此法與一階可靠度法(FORM)之比較。比較方法為：用此兩種方法設計一個只包含兩個隨機變數的簡單樁基礎設計案例。在簡單的樁基礎設計例子中，此兩種方法都可以得到精確解，並可同時於標準常態空間中比較其幾何狀態、設計的準確度、維持統一可靠度水準的能力。之後以三個接近真實的大地工程案例討論QVM在實際工程設計上應用之表現，及討論可能遇到的問題。 第二部分為討論現地調查資訊是否能夠反映在最後設計尺寸中。此部分將用四種簡易的可靠度設計法校準部份係數/分位數值，並將這些部份係數套用至一淺基礎設計案例中。而現地調查的資訊將用“虛擬現場模擬”來模擬。最終能夠取得不同試驗精度的OCR、N60、qT”值之資訊，這些資訊搭配不同的精度共有20種組合。在這20種組合之下，移植四種校準方法之部份係數淺基礎案例中，檢驗結果，而能歸納出能夠良好反應現地調查資訊之設計方法。Quantile Value Method based on quantile is to be discussed in this paper. First, verifying the QVM. A simple case is just including two randam variables which used to verify the QVM. The QVM will transplant design location to the validation case while design point used in the design value method (DVM). The QVM validation results will be compared with the DVM results in the standard Gaussian space. They showed that assume constant partial factors cannot achieve uniform reliability levels for scenarios with non-constant COVs. Second, there are three more complicated and realistic geotechnical design examples will be studied. The main conclusions in the simple case still hold in these realistic cases. In addition, we observed some interesting results in these realistic cases. Linking site investigation efforts to final design savings with simplified reliability-based is investigating in the last. There are four simple reliability-based design code used to calibrate partial factors under virtual site investigation information. The virtual site survey information is divided into 20 groups based on accuracy and the number of test. According to different calibration methods there are one or twenty partial factors/quantiles will be transplant into validaton pad foundation case. The results showed the the QVM can effectively link site investigation efforts to final design davings