標題: 變形整合測量反算分析模式Development of an Inverse Analysis Model based on the Global Deformation Data 作者: 張裕民Chang, Yu-Min陳春盛Chun Sung Chen土木工程學系 關鍵字: 反算分析;Invesse Analysis 公開日期: 1996 摘要: 在分析及預測大地變形值時，必須預先取得正確的大地參數值。以往利用土壤試驗方法 求取大地參數值，常因地層土壤的複雜性，以及大地活動之影響，不易求取大地參數之正 確值，因而導致與現場實際變形值不一致的分析結果。因此，利用現場觀測之變形資料， 反算分析符合現場之大地參數值，當較為實際且正確。 變形現場觀測的方式有很多種，為了求取正確詳盡的變形資料，在本研究中，首先建立 將各種變形量測資料之整合分析模式，再將整合分析之整體變形值作為反算分析大地參數 值之輸入數據。在反算分析模式方面，本研究突破傳統僅考慮變形觀測值之最小二乘法模 式，而將變形觀測值、大地參數值及二者間之調整權值合併考慮，發展一套新的反算分析 模式；同時，也建立該模式之評估方式，以對分析結果進行評估。為了預測大地變形，本 研究推導大地變形之卡曼濾波模式來進行分析。經過應用於土堤填土工程之實例分析及評 估，發現該模式收歛快速且分析之精度良好。 In analyzing and predicting the geotechnical deformations，the accurate value s of geotechnical parameters must be obtained. In the recent years，the values were obtained by soil test methods. However，it is difficult to obtain the ac curate values due to the complication of soil materials. The deformation data analyzed by the inaccurate parameter values are also inaccurate compared with the field deformation data. Thus it will be more practical and accurate to get the geotechnical parameters by inverse analysis using the field deformation d ata. There are many types of field deformation data. In order to get the ac curate global deformation data, this study proposes an integrating analysis mo del of various deformation data and adopts the global deformation data as the input data in the inverse analysis process. In addition，this study develops a new inverse analysis model. Such an analysis model has modified the conventio nal least squares model concerning only the field deformation data, and it con siders wholly the field deformation data，the geotechnical parameters data and the weighted value. Meanwhile, this study suggests an evaluation method on th e analysis results. For predicting the geotechnical deformations, this study d evelops a Kalman filtering model to compute the future deformations. By analyz ing the field case of a three-meter high embankment, it proves that the new i nverse analysis model has rapid convergence as well as analytical predictabili ty. URI: http://140.113.39.130/cdrfb3/record/nctu/#NT850015064http://hdl.handle.net/11536/61438 Appears in Collections: Thesis