Neural Network Inversion of EM39 Induction Log Data

  • Lin Zhang
  • Mary Poulton
Part of the Modern Approaches in Geophysics book series (MAGE, volume 21)


A modular neural network (MNN) is applied to invert well-logging curves from a Geonics EM39 induction-logging tool. In the interpretation scheme, there are several subsets of networks that depend on the relative resistivities of adjacent layers, e.g. R1>R2>R3, R1<R2>R3, etc. The well-logging curves are subdivided into fixed-length windows by each sub-network. The results are estimations of resistivity and thickness of each layer. The networks are examined for their ability to compute the correct output patterns for the corresponding input patterns of the training set, and the ability to interpret the new patterns that are not present in the training set. The networks are tested using synthetic and field data. The results show that the neural networks can facilitate interpretation of well-log data. When tested on a multi-layer case, the large shoulder effects can make the prediction difficult, especially for the thin resistive and conductive layers. However, the estimates of resistivity and thickness for each layer are sufficiently accurate that an interpreter will not be mislead as to the geological material properties. This approach could also be used for other logging tools or suites of tools.


Apparent Resistivity Ridge Regression Training Pattern Local Expert Synthetic Model 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. Haykin, S., 1994, Neural Networks: A Comprehensive Foundation: Prentice Hall, New York.Google Scholar
  2. Hoerl, A., and Kennard, R., 1970a, Ridge regression: Biased estimation for nonorthogonal problems: Technometrics, 12, 55–67.Google Scholar
  3. Hoerl, A., and Kennard, R., I 970b, Ridge regression: Applications to nonorthogonal problems: Technometrics, 12, 69–82.Google Scholar
  4. Hoversten, G., Dey, A., and Morrison, H., 1982, Comparsion of five least-squares inversion techniques in resistivity sounding: Geophys. Prosp., 30, 688–715.CrossRefGoogle Scholar
  5. Inman, J., Ryu, J., and Ward, S., 1973, Resistivity inversion: Geophysics, 83, 1088–1108.CrossRefGoogle Scholar
  6. Inman, J., 1975, Resistivity inversion with ridge regression: Geophysics, 40, 798–817.Google Scholar
  7. Marquardt, D., 1970, Generalized inverses, ridge regression, biased linear estimation and nonlinear estimation: Technometrics, 12, 591–612.Google Scholar
  8. Poulton, M. M., and Birken, R. A., 1998, Estimating one-dimensional models from frequency-domain electromagnetic data using modular neural networks: IEEE Trans. Geosci. Remote Sensing, 36, 547–555.CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2003

Authors and Affiliations

  • Lin Zhang
    • 1
  • Mary Poulton
    • 2
  1. 1.Chevron Petroleum Technology CompanyHoustonUSA
  2. 2.Department of Mining and Geological EngineeringUniversity of ArizonaTucsonUSA

Personalised recommendations