Algorithms for System Identification

Implementations of different algorithms designed for material constant identification are discussed in this contribution. Identification is performed by varying the input variables (i.e., the material constants) and juxtaposing the results obtained by analysis of the model and some benchmark example. In order to reduce the iterations needed to achieve a good agreement with desired results, different numerical strategies can be employed. One of the possibilities is to use a genetic algorithm. The combination of finite element analysis and identification algorithm is a strong tool but it is time consuming and very demanding in computational resources. A surrogate modeling can be employed to reduce computational time. Generally, it consists in replacing the original model with a simplified one. Two approaches are taken into consideration herein: the polynomial chaos expansion and the artificial neural network. The efficiency of the above-mentioned algorithms is to be assessed in terms of computational resource.