1PARALLEL REGULARIZATION INVERSION OF 2-D MT DATA BASED ON MPI

The use of parallel technology can improve the efficiency in the magnetotelluric data processing.Partial derivative matrix calculation occupy the large amount of computing time and make the inversion slow.This article propose for the objective function to use the regularization inversion method based on the damping Gauss-Newton method,and to use parallel calculation to partial derivatives and the remaining part of the reservation serial computing model.The equivalence parallel strategy are adopted among the different frequency.Simulating the multi-process parallel computing speed through two theoretical models,the results show that the correctness of the algorithm and efficiency and stability.