Non-linear multiparameter inversion using a hybrid global search algorithm: applications in reflection seismology

摘要: SUMMARY Many interesting inverse problems in geophysics are non-linear and multimodal. Parametrization of these leads to an objective function, or measure agreement between data model predictions, that has a complex topography with many local minima. Optimization algorithms rely on gradients the function search space locally may become trapped By combining simulated annealing downhill simplex method, hybrid global algorithm is presented this paper for non-linear, multimodal, problems. The shares advantages both methods perform well if suitable, able explore efficiently full space. also utilizes larger more memory store information than algorithms. effectiveness new scheme evaluated three problems: minimization multidimensional Rosenbrock 1 -D, acoustic waveform inversion, residual statics. performance compared genetic shown converge rapidly have higher success rate locating minimum cases investigated.