Evolutionary Programming for Inverse Problems in Civil Engineering

Abstract

Two main contributions to evolutionary computation and their applications are reported in this paper: (1) a new evolutionary programming (EP) algorithm is proposed with several improved techniques added to the conventional EP algorithm; and (2) the improved EP algorithm is incorporated with the finite-element method (FEM) to form an integrated algorithm, called FEMEP, for solving the inverse problems in civil and structural engineering. In the improved EP algorithm, two types of mutation—namely, random and uniform—are proposed to increase the efficiency and to retain the versatility of the conventional EP algorithm. Simulated annealing is adopted to prevent the chromosomes in the early generations from stagnating in a local optima. The algorithm is examined with two typical numerical examples. Results show that the improved EP algorithm is efficient and versatile. It is able to find better solutions than genetic algorithms and other evolution programs. The FEMEP algorithm is then applied to the inverse problems, with two parameter identification problems as illustrations. Although simple, they demonstrated the potency of the proposed integrated algorithm to solve inverse problems in civil and structural engineering.