Integrating Bi-LSTM and Extended Geological Constraints Method for Automatic Logging Curve Stratification

Logging curve stratification plays a crucial role in the interpretation of geological data. As the volume of logging data increases, logging curve stratification becomes increasingly challenging and time-consuming. Current automated log stratification methods primarily rely on pattern recognition techniques, which are limited to processing single curves and overlook the complex nonlinear relationships between different logging curves, resulting in suboptimal performance in thin layer stratification. In this study, we propose a deep learning-based approach that simultaneously handles multiple logging curves while incorporating geological constraints to enhance the impact of spatial features on logging curve stratification. Furthermore, we optimize the loss function and introduce a new evaluation metric to better assess the model's performance. Experimental results demonstrate that our proposed method performs better in terms of stratification error and thin layer detection.