进化策略算法在地面沉降量计算参数反演中的应用

多数情况下,评估的研究区缺乏多年积累的地下水位及开采量资料。因此,通常采用土力学模型,根据太沙基固结理论,进行地面沉降量预测。在地面沉降量计算中,土层变形参数值的选取直接影响计算结果的精确性。深层含水层埋深通常大于100m。通过现场勘探,采取土样进行岩土物理力学性质试验获取参数值,成本过于昂贵。通过参考区域土层物理力学性质资料,结合工程经验获取参数值,预测结果的精度得不到保证。针对上述情况,作者提出用非线性优化算法-进化策略算法,根据实测地面沉降量反演土层变形参数值。通过这种方法确定参数值,既节约成本,又保证了计算结果的可靠性。进化策略算法通过模拟生物遗传及进化过程,利用转移概率来帮助指导搜索。搜索结果不依赖于初始点的选择,对于求解全局最优解有很强的鲁棒性。作者将进化策略算法用于某一工程实例土层变形参数的反演,结果表明了该算法的可行性及稳健性,值得在工程实践中推广应用。

Appliance of evolution strategy algorithm in obtaining back analyzed parameters for calculation of land subsidence

In most cases,the soil mechanic model based on the theory of 1 dimensional Terzaghi consolidation is adopted to predict the amounts of land subsidence,since the lack of systematical accumulated information about the level of groundwater and exploitation quantity of groundwater.The values of deformed parameters are critical to the accuracy of results in land subsidence calculating.It is expensive to obtain the parameter's value by physical mechanical test on soil samples got by means of geological exploration,for the depths of aquifer are usually more than 100 meters.Meanwhile the accuracy of results can not be guaranteed,if the value of parameters is determined through referencing regional soil's physical mechanical information.According to the above reasons,the author puts forward a nonlinear optimal method that is evolution strategy method which uses the measured surface subsidence to inverse value of parameters.It proves that the cost of this method is decreased and the accuracy of results is reliable as well.The evolution strategy algorithm based on the principle of simulating biological heredity and evolution searches the overall optimal solution by the rule of transverse probability,and the searching results are independent from the initial values.In this paper,the algorithm is adopted to obtain back analyzed parameters of land subsidence computing model in a practical project,and the numerical results confirm that evolution strategy algorithm is feasible and stable.It can be widely used in future practice.