Predicting water saturation using artificial neural networks (ANNS)
AIAP'07 Proceedings of the 25th conference on Proceedings of the 25th IASTED International Multi-Conference: artificial intelligence and applications
| Hi-index | 35.68 |
Multiple reflections in seismic data are generally considered to be unwanted noise that often seriously impedes correct mapping of the subsurface geology in search of oil and gas reservoirs. We train a backpropagation neural network in order to recognize and remove these multiple reflections and thereby bring out the primary reflections underneath. The training data consist of model data containing all multiples and the corresponding seismic sections containing only the primary arrivals. The basis for the modeling is data from a real well log that is typical for the area in which the data were gathered. In contrast to existing conventional deconvolution methods, the neural network does not depend on such restricting assumptions concerning the underlying model as, for example, the Wiener filter, and it has the potential to be successful in cases where other methods fail. A further advantage of the neural net approach is that it is possible to make extensive use of a priori knowledge about the geology, which is present in the form of well log data. Tests with realistic data show the ability of the neural network to extract the desired information