PROPOSAL OF CONSTRUCTIVE ALGORITHM AND DISCRETE SHAPE DESIGN OF THE STRONGEST COLUMN

This paper deals with the proposal of a new genetic algorithm based optimization algorithm referred to as the constructive algorithm and its application to the discrete shape design of the strongest column with maximum buckling load of the first mode under constraint of constant weight. The buckling load analysis is performed by the finite element method. Introducing a set of system parameters concerned with the bending rigidity of each finite element, the problem is reduced to that of finding the best set of the system parameters from the great number of sets satisfying the constraint. The basic idea of the constructive algorithm is as follows. One individual is generated in which a set of system parameters is encoded. Then a set of strings with an updating rule of the system parameter set is generated. The strings are then stored to computer memory with fitness values based on a certain rule and the evolutionary operation of standard genetic algorithms is applied to the strings. To demonstrate the efficiency of the constructive algorithm numerical calculations were performed. It was shown that the constructive algorithm exhibits high performance in finding the best shape from the great number of shapes satisfying the constraint.