摘要 |
Earth models 50 (Figure 5) from survey areas are input to the artificial neural network (ANN). The result is a trained ANN 60. An initial model of the Earth's subsurface 62 in the survey area is input to the trained ANN 60. The initial model 62 may include spatial distribution of resistivity in three dimensions. The initial model 62 may be constrained by any available information such as petrophysical data. The trained ANN 60 generates a set of expected electromagnetic instrument responses 66. The expected responses 66 are compared, at 68, to the measured responses of the electromagnetic instrument. If the differences between the measured response and the expected response are not below a selected threshold, then at 70 the model is adjusted and the expected response is recalculated in the trained ANN 60. The adjustment and recalculation procedure is repeated until the differences fall below the selected threshold.
|