A Genetic Algorithm for Residual Static Correction

Residual static correction is a necessary step to improve the resolution in the seismic exploration process. It is a challenging task because a large number of parameters need to be adjusted. Some machine learning methods have been proposed to deal with this problem, but the results should be further strengthened. In this paper, we propose the genetic-based residual static correction (GBRS) algorithm with three techniques. First, the original encodings is generated by per-forming floating encoding on the offset of each point. Second, a new encodings is constructed through paired crossover on the original ones. Third, the fitness function is used to select new original encodings to promote the evolution of the population. Experiment data with 50 shots and 50 receivers are generated using a simulation model. Results show that our algorithm usually converges in less 100 iterations to the optimal solution.