Details of the Abstract
| Title of paper | Application of CNN-BiLSTM transient electromagnetic inversion in the detection of coal seam gob |
| List of authors | Jifeng Zhang,Yu Shi , Zhipeng Qi |
| Affiliation(s) | Chang’an university |
| Summary | The traditional inversion method of transient electromagnetic is easy to fall into local optimal when dealing with non-uniform geoelectric structure, and it is difficult to meet the practical exploration requirement. In this paper, an inversion method based on convolutional bidirectional long short time memory neural network (CNN-BiLSTM) is introduced, which is applied to the precise inversion of fixed source large loop transient electromagnetic. This network structure has strong ability of extracting spatial features and understanding sequence data, which solves the problem of slow computation efficiency and insufficient accuracy of traditional inversion. Using the apparent resistivity of the three-layer model as the sample input and the real model as the sample target, the network is trained, and batch normalization and dropout techniques are used to accelerate the convergence of the network. Through numerical simulation experiments, the inversion efficiency of this method is much better than that of the traditional method, and it also has excellent inversion accuracy and geoelectric stratification ability. The CNN-BiLSTM inversion is applied to the measured coal seam gob detection, and the inversion effect is good, and it is consistent with the drilling verification results. This work provides a new and efficient inversion method for the transient electromagnetic exploration field, and has potential application prospects in other fields. |
| Session Keyword | 2.0 EM theory, modelling and Inversion |
| File upload |
2.0_application_of_cnn-bilstm_zhang.pdf
|