遺傳演算法於地層下陷模式參數檢定之應用

Abstract

本研究的主要目的在於發展地層下陷數值模擬之參數優選模式，並建立所需之地層下陷 數值模式。在參數優選模式方面，我們採用非傳統之參數檢定方式－遺傳演算法(Genetic Algorithms;GAs)。此一演算法係基於自然界物競天澤之理念，可有效地選擇計算表現良 好的點，在處理多峰函數的最佳化問題上，遺傳演算法大幅提高了收歛於整體最佳值(Glo bal Optimum)的機會，係一新而有效的優選方法。最後將參數優選模式與地層下陷模式整 合，優選所得之模式參數其模擬結果極為良好，故可驗證遺傳演算法確實適用於地層下陷 模擬模式之參數檢定。本研究並同時建立優選模式所須的地層下陷數值模式，此模式係依 據Helm博士(1976)所提出的地層下陷理論為基礎，以德拉基(Terzaghi)一維壓密方程式作 為控制方程式。經過數個案例的測試後，發現符合德拉基解析解及疊加定理，證明此模擬 模式之正確性。綜合以上之結論，可說明本研究之地層下陷模擬模式結合遺傳演算程序優 選參數之效果良好，未來將進一步應用於架構本省地層下陷區之模式。 The main purpose of this study is to develop a parameter optimization model for land subsidence modeling using the Genetic Algorithms(GAs).The procedure o f GAs is based on the law of nature selection and can efficiently identify the points with good performance. The GAs can compute the global optima with highp ossibility even for a problem with multiple local optima. Therefore, it is aef fective algorithm for parameter optimization. The GAs was then combine with a one-dimensional numerical consolidation model. A land subsidence simulation mo del using the parameters obtained from the resulting parameter identification el have a simulated output close to the observed data. This demonstrate that i t is encouraging to apply the GAs in the parameter identification for land-sub sidence modeling. The developed land-subsidence numerical model was based on the consolidation theory proposed by Dr. Helm, which is a modification from t he formula proposed by Terzaghi. Several testing cases are computed to verify the model accuracy. The results are in accordance with a analytical solution b asing on simplified Terzaghi formula and the superposition theory. This resear ch demonstrate that combination of GAs with numerical land-subsidence model h ave a good performance. The developed parameter identification model can be fu rther applied to a field land-subsidence simulation.