Identification of unknown parameters in the dynamic analysis of a subway track by genetic algorithms

An identification method is developed for unknown material parameters in a dynamic subway-track model. The proposed method is formulated in the framework of genetic algorithms (GAs). An individual in the GA is defined by a set of real unknown parameters not by a coded string. The evaluation of fitness is carried out based on a three-dimensional dynamic analysis, in which the subway track, tunnel and ground are modeled by means of the coupling of boundary element (BE) and finite element (FE) methods. To save the computational work, a neural network (NN) in which the material parameters are input data and the spectrum ratios are output data is introduced. The developed method is validated through application to numerical models and in situ measurements.