Autor: |
Sadeghi, Hossein, Suzuki, Sadaomi |
Jazyk: |
angličtina |
Rok vydání: |
1995 |
Zdroj: |
Memoirs of the Faculty of Science, Kyushu University. Series D, Earth and planetary sciences. 29(1):39-48 |
ISSN: |
0916-7390 |
Popis: |
We study neural networks for interpretation of travel time data of refracted waves in a field experiment. The designed multilayer network consists of an input layer, two hidden layers and an output layer. Fifty examples of simple velocity structure consisting of two layers separated by a plane interface dipping less than 10 degrees are used to train the neural network. Surface layer velocity, travel times and distances between two sources and five receivers are used as the network inputs. Target output is the expected velocity structure model. The trained neural network produces results that are close to the actual output of the new examples not included in the training set. The trained neural network also gives acceptable results for travel time data of field experiment. We demonstrate that neural networks are able to perform transformation giving the correct connections between data and model parameters in seismic refraction surveys. |
Databáze: |
OpenAIRE |
Externí odkaz: |
|