EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Intelligent Classification of Coal Body Structure in Logging Based on Wavelet Transform and BO-CNN-TransKAN Model

Huiyuan Bian, Jun Zhou, Yongfa Xu, Yue Chen, Wei Wang, Yinting Wu, Jiajun Ji

Lithosphere · 17.11.2025 · 2025 (4)

DOI: 10.2113/2025/lithosphere_2025_108 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Lithosphere, online erschienen am 17.11.2025. Verfasst von Huiyuan Bian und 6 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 54 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate determination through logging of the coal body structure (CBS) contributes to deep exploration and development and stratigraphic optimization of coal bed methane. Due to the “boundary effect,” the vertical definition of conventional logging data is low, making it harder to accurately identify the CBS. In this study, we propose a method to enhance the longitudinal determination of logging curves using the wavelet transform (WT). First, the mean absolute error (MAE) and signal-to-noise ratio are utilized to select the optimal wavelet basis functions for different logging curves. Subsequently, wavelet decomposition is applied to the original logging curves, with the high-frequency wavelet components being weighted and then reconstructed into new logging curves. This process enhances the vertical resolution of thin coal seams. Then, convolutional neural network (CNN) and Kolmogorov-Arnold network (KAN) were integrated with the Transformer model to predict the CBS using the reconstructed logging data. In this paper, we propose the BO-CNN-TransKAN model for predicting the structure of logging CBS based on WT. The model integrates the capacity of local feature selection of CNN, dynamic message modeling of Transformer, and nonlinearity of KAN network. The experimental results demonstrate that the WT-fused BO-CNN-TransKAN model achieves superior classification performance, attaining an accuracy of 93.67%, surpassing both traditional machine learning models and models without WT-based reconstruction. This model is applied to predict the CBS of the coal seam in Well A within the Ordos Basin. The forecast results show a close agreement with the actual CBS, confirming the model’s effectiveness. The WT-fused BO-CNN-TransKAN model offers a novel approach to identifying CBS using logging data.

Zitieren

APA

Bian, H., Zhou, J., Xu, Y., Chen, Y., Wang, W., Wu, Y., & Ji, J. (2025). Intelligent Classification of Coal Body Structure in Logging Based on Wavelet Transform and BO-CNN-TransKAN Model. Lithosphere, 2025(4). https://doi.org/10.2113/2025/lithosphere_2025_108

Vancouver

Bian H, Zhou J, Xu Y, Chen Y, Wang W, Wu Y, et al. Intelligent Classification of Coal Body Structure in Logging Based on Wavelet Transform and BO-CNN-TransKAN Model. Lithosphere. 2025;2025(4). doi:10.2113/2025/lithosphere_2025_108

BibTeX
@article{Bian2025,
  author = {Bian, Huiyuan and Zhou, Jun and Xu, Yongfa and Chen, Yue and Wang, Wei and Wu, Yinting and Ji, Jiajun},
  title = {Intelligent Classification of Coal Body Structure in Logging Based on Wavelet Transform and BO-CNN-TransKAN Model},
  journal = {Lithosphere},
  year = {2025},
  volume = {2025},
  number = {4},
  doi = {10.2113/2025/lithosphere_2025_108}
}
Kostenlos starten →