New few parameters differential evolution algorithm with application to structural identification

Differential evolution algorithm (DEA) is a stochastic, population-based global optimization method. In this paper, we propose new schemes for both mutation and crossover operators in order to enhance the performances of the standard DEA. The advantage of these proposed operators is that they are “parameters-less”, without a tuning phase of algorithm parameters that is often a disadvantage of DEA. Once the modified differential evolutions are presented, a large comparative analysis is performed with the aim to assess both correctness and efficiency of the proposed operators. Advantages of proposed DEA are used in an important task of modern structural engineering that is mechanical identification under external dynamic loads. This is because of the importance of using a “parameters-less” algorithm in identification problems whose characteristics typically vary strongly case by case, needing of a continuous set up of the algorithm proposed. This important advantage of proposed optimizers, in front of other identification algorithms, is used to develop a computer code suitable for the automatic identification of a simple supported beam subject to an impact load, that has been tested both using numerical simulations and real standard tests dynamic. The results point out that this algorithm is an interesting candidate for standard applications in structural identification problems. Keywords: Differential evolution, Parametric identification, Structural identification, Optimizatio