在大地电磁二维Occam反演中求取拉格朗日乘子方法改进

Occam反演具有可靠性和加上各种先验信息可行性的优点，但是因为求取拉格朗日乘子需要很长时间，所以前人提出了几个直接选取方法。笔者改进了二维大地电磁测深Occam反演中求取拉格朗日乘子的方法。在搜索过程中加上观测数据拟合差与其期望值的比较。这样，在Occam反演的拟合差下降阶段的最后迭代和模型光滑阶段，能够排除多余的正演计算。这种改进对反演结果没有影响，但却使得反演速度加快并且更符合于Occam思想。根据模型实验和野外数据处理，可以减少正演次数20%~50%。

Improvement of Choosing Lagrange Multiplier on MT 2D Occam Inversion

Occam’s inversion is reliable and  able to incorporate varying levels of the priori information. The Choice of Lagrange multiplier is computationally expensive and some direct selection methods are then suggested in this paper.  We modified the selected method of Lagrange multiplier in 2D magnetotelluric Occam’s inversion. In the search progress, the misfit between  the forward model response and  observed data are added and compared with  the  target level. This can eliminate superfluous forward calculation in the last iteration of the phase of bringinge misfit down and the phase of smoothing out the unnecessary structure. This modification  can accelerate inversion and more accordant with   the Occam’s idea.Through inversions of several synthetic and field data,it is improved that forward caculation number can be reduced to 20%-50%.