基于非结构化网格的2.5维时间域激发极化反演

针对我国西部山区起伏地形,提出通过Delaunay三角剖分生成非结构化三角网格,可以根据实际地形情况,较好地控制节点分布,从而降低初始模型的几何离散误差。在反演中对模型参数施加最小光滑和背景约束信息,从而提高反演的分辨率。通过在正演求解稀疏矩阵时采用稳定型双共轭梯度法和计算系数矩阵时利用电位互换等手段,可有效地提高反演计算速度。最后,为验证反演方法的有效性,编制了跨平台反演软件;并对2例地电模型和实测数据进行反演试算。计算结果表明:反演耗费时间少,仅迭代3次拟合差便趋于稳定,反演结果能较好地刻画异常体形态。

2. 5-D Time Domain Induced Polarization InversionBased On The Unstructed Mesh

According to Delaunay subdivision,Unstructed triangular grid was developed for the resistivity model with fluctuated geometry. The resoluion of inversion which 2.5-dimensional (2.5D) induced polarization(IP) data with waved geometry was improved by including model smoothness and background prior information. We calculate sensitivity matrix with a combination of reciprocal theory and sparse matrix with Bicgstab. So,the computation efficiency was significantly improved,and we make the cross-platform program for testing the effectiveness of the inversion method by two examples. The computation results show that the inversion converg within 3 iterations and takes very little consumtation time. The inversion result show a good interpretation of the electric models.