
Full text loading...
New technology for seismic inversion using multilayer neural network (NN) is proposed. Special set of seismic attributes based on offsets serves as an input to the neural network. Logs measured at the well serve as training data. Genetic algorithm and back-propagation are used in NN to calculate weights. This allows implicit derivation of the non-linear operator. The method naturally extends to other logs which have meaningful physical relationship to the seismic data, such as porosity or density.