The transient electromagnetic inversion of detection signals mainly depends on fast inversion in the half-space state. However, the interpretation results have a certain degree of uncertainty and blindness, so the accuracy and applicability of the three-dimensional full-space inversion need to be investigated. Two different three-dimensional full-space inversions were carried out. First, the numerical characteristic parameters of the response signals were extracted. Then, the correlations between the numerical characteristic parameters and physical parameters of the water-bearing abnormal bodies were judged, and the judgment criterion of the iterative direction was proposed. Finally, the inversion methods of the iterative algorithm and the BP neural network were utilized based on the virtual example samples. The results illustrate that the proposed numerical characteristic parameters can accurately reflect the response curve of the full-space surrounding rock. The difference in the numerical characteristic parameters was used to determine the update direction and correction value. Both inversion methods have their advantages and disadvantages. A single inversion method cannot realize the three-dimensional inversion of the physical parameters of water-bearing abnormal bodies quickly, effectively and intelligently. Therefore, the benefits of different inversion methods need to be considered to comprehensively select a reasonable inversion method. The results can provide essential ideas for the subsequent interpretation of the three-dimensional spatial response signals of water-bearing abnormal bodies.
Keywords: iterative algorithm; neural network; response signals; three-dimensional inversion; transient electromagnetic method; water-bearing abnormal bodies.