- 學位論文
液化可信度分析 – 新竹市為例
Reliability Analysis of Liquefaction – Hsin-Chu City for Example
摘要
台灣位於西太平洋地震環帶，地震發生十分頻繁。1999 年「921 集集大震」期間，液化發生於台灣中西部多處砂質地盤而興起國內學者廣泛的探討。而評估廣域地區之液化潛能通常採用基於 SPT-N 值簡易法所計算得之安全係數。安全係數方法屬於定值性分析且為工程界普遍所接受者。然而，其最大的缺點為無法合理地考量相關參數之變異性，尤其當此變異性資料已經可得時。 本論文主旨在於建立一種簡單評估一場址液化機率 (PL) 之可信度分析架構。選用之極限函數由修正Seed簡易經驗法所定義 (NCEER, 1997)，其共需要七個輸入參數，即最大地表加速度Amax、 地震規模M、深度Z、土壤單位重 g、SPT-N值、土壤細料百分比FC及地下水位深度Dw；而對應之 PL 值及其他者則由通用可信度程式 CALREL 來計算。具高科技產業之新竹市被選為分析示範區，現地審慎篩選了共 432筆鑽孔資料，並先找出每孔位最可能之液化土層深度 (即為 Z)；而後所有鑽孔分成幾個地質相似之群落，每個群落再計算出其共有Z、SPT-N、γ、F及Dw 之變異係數。對於每個孔位，分別以Seed簡易經驗法及 CALREL 求其小安全係數 (FSm) 及 PL 值，接著製作出新竹市全區 FSm 及 PL 之等值圖以供比較。文中將輸入兩種震度值，一為現地 921 地震平均實測值及新公佈國家建築規範者。 於輸入921地震平均值時，所得 FSm 及 PL 等值圖皆顯示新竹市全區幾乎屬於非液化區，此與現地並無任何液化災害報告相吻合；於輸入新國家規範震度時，所得 FSm 及 PL 等值圖皆顯示新竹市大部份地區有極高之液化潛能，此點可提供該區政府未來施政之參考。而FSm 及 PL 等值圖所顯示危險區域分佈之主要差異乃是每孔位輸入參數之變異性並非相同所致。CALREL之敏感性分析也同時求出新竹市各輸入參數之重要性排序，由大至小分別為Amax、M、Z、SPT-N、γ、Dw、FC。 本論文所建立之一種評估液化機率之可信度分析架構於成功應用於台灣其他地區前仍需要許多修正。這些修正工作包括了與更多液化場址之比對、考量現地常態地下水位之變動及採用其他SPT-N 值簡易評估法等。
並列摘要
Taiwan is located in the western Pacific seismic rim, where earthquake occurs quite frequently. During the 921 Gi-Gi Earthquake, liquefaction took place in many sand-deposited sites of the mid-western Taiwan, this drawing great attention of local geotechnical engineers. The factor of safety (FS) determined by several SPT-N-based methods is commonly used to assess the liquefaction potential of a large region. The FS approach is deterministic and widely adopted in most engineering applications. However, it is extremely difficult for it to take the variability of relevant parameters in account, especially when such variability information is available. The objective of this thesis is to establish a simple framework of reliability analysis for evaluating the liquefaction probability (PL) of a site. The selected limiting function was derived from the modified Seed method (NCEER, 1997), in which seven input parameters are required, i.e., maximum ground acceleration Amax, earthquake magnitude M, depth Z, soil unit weight g, standard penetration value SPT-N, soil fine percentage FC, groundwater table depth Dw. And the corresponding value of PL and the like were computed by a universal reliability program, CALREL. The Hsin-Chu city with a hi-tech industry distributed was taken for an illustrative site. By careful screening-out, data of 432 boreholes were collected, followed by locating the most likely liquefied depth of each borehole. Subsequently, these boreholes were divided in several similar geological groups, for each of which the coefficients of variance for Z, SPT-N, γ, FC, and Dw were determined. For each borehole, its minimum FS (FSm) was computed by the modified Seed method and its PL by CALREL. Then, the contour maps of FSm and PL over the whole city were generated for comparison. Two kinds of input earthquake intensity were used: one was the site average Amax measured during the 921 Gi-Gi Earthquake and another specified by the newly-issued national building code. For the input case of 921 Gi-Gi Earthquake, both FSm and PL contour maps fairly demonstrated the Hsin-Chu city as a non-liquefied zone, which is compatible to the fact of no liquefaction disaster reported after the quake. For the input case of new building code, both FSm and PL contour maps exhibited that most of the Hsin-Chu city have a high liquefaction potential, this providing pre-warning to the local government. The major difference of indicated critical zones between FSm and PL contour maps is attributed from that the value of PL for each borehole is affected by its own variability in input data. The significance order of input parameters for the site was also determined by the sensitivity analysis of CALREL, and from greatest to lest it was Amax > M > Z > SPT-N > γ > Dw > FC. The framework of reliability analysis for evaluating the liquefaction probability established in this thesis definitely requires additional revisions, before successfully applying it to the other areas in Taiwan. These revisions may be composed of calibrating with more liquefied site measurements, considering the fluctuation of normal Dw, adopting another suitable SPT-N-based method, etc.