Comparison of local and global search algorithms for inverse electromagnetic problems

Inverse problems of electromagnetic sounding often aim at extracting a one-dimensional piecewise constant conductivity profile of the marine environment from electromagnetic field measurements. Generally, solving inverse problems like this, local Newton like search methods have been shown to be sufficient, as long as the parameter search space is reasonably simple. In this work a performance comparison between a local algorithm and two global methods, a genetic algorithm and a simulated annealing, is made for inverse electromagnetic problems. The comparison is based on both simulated data and experimental results from field trials in the Stockholm archipelago. All results indicate that the one dimensional inverse problems are simple enough to be solved by a local algorithm, which performs better than the global methods. For more complicated searches, involving e.g. multidimensional environment models, the global algorithms will be more competitive.