Neural NetworkとRosenbrock法による直流電位差非破壊診断

概要

論文の詳細を見る
This paper proposes an effective algorithm for non-destructive testing in DC potential difference method The purpose of this paper is to identify the position and the size of cracks on backside of a conductor quantitatively. The testing using DC potential difference method is effectively performed by combination of numerical analysis method and inverse analysis method. The boundary element method is adopted as the numerical analysis method for calculating the potential distribution on the conductor. In performing non-destructive testing, accurate information of crack is needed instantly. Therefore, neural network optimized by simulated annealing method is used as the first inverse analysis method because sufficient solution is obtained instantly. Furthermore, Rosenbrock's method is used as the second inverse analysis method to identify more accurately from the result by neural network As a result, the proposed non-destructive testing procedure enables to identify cracks instantly and accurately than in the case obtained only by neural network
2003-09-15