基于粒子群支持向量机的三维含水层渗流参数反馈识别

实际岩土体含水层渗流一般是三维、空间各向异性的。针对三维渗流参数识别的数值正算时间过长，易限于局部最优解的问题，提出了一种基于支持向量机和粒子群优化算法的含水层渗流参数反馈识别方法。采用正交试验设计和有限元程序生成学习样本，利用支持向量机高度非线性映射能力，建立水头与渗流参数之间的映射关系，进而以识别误差目标函数为适应值，通过粒子群优化算法反馈搜索得到渗流参数。该方法可直接利用现有大规模渗流有限元程序进行三维含水层渗流参数识别。算例表明，该方法具有良好的效率和精度。

Feedback identifying seepage parameters of 3D aquifer based on particle swarm optimization and support vector machine

For actual engineering, the aquifers are generally three dimensional anisotropic problems, aiming at the computing time too long and local optimization limitation of conventional method, a new method of three dimensional seepage parameters identification based on support vector machine and particle swarm optimization is proposed. Adopting orthogonal experimental design and finite element program producing training samples, using SVM mapping, the nonlinear relation between water heads and seepage parameters is established. Then taking error objective function as the fitness value of particle swarm optimization, the seepage parameters should be identified by PSO. The method can directly utilize the large scale seepage finite element program. The computing example is given to prove that the method has favorable efficiency and precision.