Particle Model Method for Stress Analysis of Soil (II)

A Numerical Method for Unconstrained Nonlinear Optimization Problems

Abstract

This report develops a numerical method for unconstrained nonlinear optimization problems. This method is based on a kinetic point of view and is named Dynamic Damping Method (DDM).
The DDM has following merits.
(1) The DDM optimizes an object function by only its gradient vector.
(2) The amount of calculation for optimization is small in subroutine DDM.
(3) Its programme is simple.
The experiments of DDM indicate good convergence. The computing speed by DDM is as fast than that by Davidon's Method or by Fretcher and Reeves' Method.
The DDM seems to be effective to the problems which have many independent variables and large amount of object function calculation.
The DDM is, also, able to find an attractor or a repeller of the vector field without potential.