Abstract
The performance of gradient methods such as Gauss-Newton method. Davidon-Fletcher-Powell method and modified Successive Linear Programming is compared for the solution of identification of dynamic soil properties. The identification problem is formulated on the basis of multiple reflection and refraction theory. Through numerical analysis, we showed that modified Successive Linear Programming is as good as Gauss-Newton method on the speed of convergence of the solutions, and is superior to others on the property of gloval convergence.
