CAJ | 학술논문

利用地温探测堤坝工程中的多重集中渗漏通道位置是一种反问题，为取得全局最优解及突破现有方法的局限性，采用不同优化算法相互配合的温度全局最优化方法，以达到有效准确探测显著集中渗漏通道的位置。结合各算法优势及目标函数不连续性等特点，选用全局搜索（GS ）、多初始点（MS ）及遗传算法（GA ）3种方法联合补充优化。首先利用GS搜索局部最优点；然后结合目标函数不连续点及上述局部最优点，人工增加MS优化起始点进行多初始点优化；最后依据上述局部解，利用GA混合算法优化，从而有效提高全局最优化的概率。根据堤坝工程集中渗漏通道低温补给的特点，对优化目标函数进行了具有特定物理意义的修正；为减少每次迭代的优化参数，基于温度测试精度提出分步优化方法，根据温度或修正目标函数残差吸引盆特征确定某步优化的集中渗漏数量；由修正残差与温度测量精度相对大小确定优化是否终止，对应修正残差最小的优化结果为集中渗漏通道位置的真实解。结合工程实例，计算了某水库左坝肩的多重集中渗漏通道位置，分步优化了两个主要的集中渗漏位置，直接法优化3个集中渗漏位置。文献及工程实践证明了该法的正确性及有效性，为了堤坝等水利工程岩土体的集中渗漏整治提供了科学依据。
이용지온탐측제패공정중적다중집중삼루통도위치시일충반문제，위취득전국최우해급돌파현유방법적국한성，채용불동우화산법상호배합적온도전국최우화방법，이체도유효준학탐측현저집중삼루통도적위치。결합각산법우세급목표함수불련속성등특점，선용전국수색（GS ）、다초시점（MS ）급유전산법（GA ）3충방법연합보충우화。수선이용GS수색국부최우점；연후결합목표함수불련속점급상술국부최우점，인공증가MS우화기시점진행다초시점우화；최후의거상술국부해，이용GA혼합산법우화，종이유효제고전국최우화적개솔。근거제패공정집중삼루통도저온보급적특점，대우화목표함수진행료구유특정물리의의적수정；위감소매차질대적우화삼수，기우온도측시정도제출분보우화방법，근거온도혹수정목표함수잔차흡인분특정학정모보우화적집중삼루수량；유수정잔차여온도측량정도상대대소학정우화시부종지，대응수정잔차최소적우화결과위집중삼루통도위치적진실해。결합공정실례，계산료모수고좌패견적다중집중삼루통도위치，분보우화료량개주요적집중삼루위치，직접법우화3개집중삼루위치。문헌급공정실천증명료해법적정학성급유효성，위료제패등수리공정암토체적집중삼루정치제공료과학의거。
This paper is to decrease the tentative calculations and get the global optimum in higher possibility.The temperature global optimization is employed to locate the concentrated seepage pipes in rock-soil body of the dam or abutment.Global search(GS),Multistart(MS)and Genetic algorithm(GA)are selected to estimate the positions of the concentrated seepage pipes with the discontinuous objective function associated with cooperation and complementarity of these algorithms.Firstly,the local optima are searched by GS.Then in the consequent MS optimization,the local optima are inputted as the initial starts to increase the probability of searching the global optima.Finally,the result is verified or improved by genetic algorithm and its hybrid algorithm.According to the characteristics of concentrated leakage passage recharged by water in lower temperature,the objective function in the optimization is modified with this specific physical meaning.The step-by-step optimization is set up to reduce the parameter amount in each iteration.The number of the concentrated seepage pipes is ascertained by the attraction basins composed by temperature datum or their residuals.The optimization is terminated when the modified residual is less than the data detection precision.The real solution for the locations of the concentrated passages is the iteration result with the least residual.In the case study,two intense concentrated seepage pipes and three minor ones are estimated by step-by-step and direct optimizations in the left leaky dam abutment of Douhe reservoir,where the locations of the multi-seepages are firmed by engineering practice partly.