MODIFIED LIMIT STEP LENGTH ITERATION ALGORITHM IN CORRELATION NORMAL SPACE

When the structural failure function is highly nonlinear, the initial step length is difficult to determine for the application of the limit step length iteration method. To solve this problem, the modified limit step length iteration algorithm is presented in this paper. Illustrative examples demonstrate the reason of zigzagging movements and non-convergence in the application of the limit step length iteration method. To ensure that each iteration step length is optimized, the golden section method is introduced for one dimensional search of the step length. A new merit function is given according to the extreme value condition of extensive Lagrange function, and the calculation procedure of the modified limit step length iteration method is presented in correlation normal space. Numerical examples demonstrate the accuracy of the modified limit step length iteration method. It has better convergence when the structural failure function is highly nonlinear, and the number of iteration step is less than that of modified iterative algorithm.