Popis: |
In this study we investigated several approaches and proposed a workflow for smooth 2D modeling of MT data using particle swarm optimization. Due to the known problems of global optimization methods, these methods are geberally used in 1D problems or constrained heavily. In this study we suggested a workflow for 2d modeling purposes which relies solely on the ability of the swarm optimization algorithm to recover a smooth 2D model. In global optimization methods partial derivatives are generally not calculated, preventing implementation of traditional methods to obtain smooth models. Due to this, we forced the model smoothness by an operator. Also due to the nature of the particle swarm optimization method, many artefacts are found to be emerging in the recovered models. Hence, we also employed minimum gradient support to prevent artefacts. The proposed approach is implemented on two differet synthetic datasets representing MT and RMT cases. The same datasets are also modeled with a smooth inversion algorithm for comparison. |