Document Type : Regular Article
Authors
1
Department of Civil Engineering, LGGC Laboratory, University Blida 1, Algeria
2
Department of Civil Engineering, LGSDS Laboratory, Ecole Nationale Polytechnique, Algiers, Algeria
3
University Gustave Eiffel, UPEC, CNRS, Laboratory Modélisation et Simulation Multi Echelle (MSME 8208 UMR), Marne-la-Vallée, France
4
Department of Geotechnics and Hydraulics, LEEGO, University of Sciences and Technology Houari Boumediene, Algiers, Algeria
10.22115/scce.2023.391609.1627
Abstract
Predicting the propagation of ground vibration wave, induced by highways, railways, electrical power plants and other industrial activities, in their vicinity is of paramount importance to evaluate the degree of nuisance to populations, buildings and sensitive equipment. The present study proposes a generalized artificial neural network (ANN) model able to predict the ground vibrations, at any distance in any type of soil, due to any source of vibration. The ANN network inputs are the distance between the measurement point and the excitation source, the soil characteristics for homogeneous soil or their equivalent characteristics for heterogeneous soil, as well as the amplitudes of the source acceleration time history. Its outputs are the acceleration time history responses of the free field points located at given distances from the excitation source. A large dataset including realistic and unrealistic soil properties is artificially generated in order to train and test the ANN. The database covers the characteristics of homogeneous and heterogeneous soils having homogenized properties. The time delay between the source and the distant point is accounted for using the cross-correlation product technique. The test results show that the correlation between the target ground acceleration signals and those predicted by the trained ANN model reaches 98%. Furthermore, the predictions of ground acceleration peaks have good accuracy, i.e. less than 2% error. Considering the unlimited possibilities to sample the soil properties combined to the homogenization and the synchronization techniques, the ANN model can be generalized to many practical applications.
Keywords
Main Subjects
)