Multi-objective particle swarm inversion method for data of magnetic method

Abstract

The invention provides a multi-objective particle swarm inversion method for the data of the magnetic method, wherein the regularized inversion problem is converted into the multi-objective optimization problem. According to the method, firstly, a multi-objective inversion solution set for the data-fitting process and the model-restricted process is solved out simultaneously based on the global optimization method. After that, the relative importance degrees of both the data-fitting process and the model-restricted process are weighed, and then an inversion result is optimally selected out of the inversion solution set. In this way, the factor regularization effect is realized. According to the technical scheme of the invention, the difficult selection problem of regularization factors and the dependence problem of initial models are solved at the same time. Compared with the conventional regularization inversion method wherein only one inversion result is obtained, feasible solutions are obtained as much as possible and then an inversion solution set can be obtained based on the above method. Therefore, the geophysical research personnel can better understand the inversion process through analyzing the solution set, and can also flexibly evaluate and select solutions in the two aspects of the data-fitting process and the model-restricted process. As a result, a result superior to that of the regularized inversion method is obtained.

Description

technical field [0001] The invention particularly relates to a multi-target particle swarm inversion method for magnetic method data. Background technique [0002] Instability and Multiple Solutions in Geophysical Inversion, Tikhonov Regularization Method [1,2] has been widely accepted and applied to solve such pathological inverse problems. The objective function of regularized inversion consists of data fitting, model constraints and regularization factors. The introduction of model constraints and regularization factors can stabilize the inversion process and reduce the multi-solution of the inversion. Among them, the regularization factor is a compromise coefficient between data fitting and model constraints, and its value directly affects the quality of the inversion results. Many scholars have conducted in-depth research on the selection of regularization parameters. How to choose the most appropriate regularization factor is still one of the hotspots and difficultie...

AI Technical Summary

Problems solved by technology

However, currently this type of method considers data fitting more, and combines less with regularized inversion

Since the global search method is more time-consuming than the local search method, the selection of its regularization factor is also more difficult

Images