A lithology identification method and system based on measured analog drilling feature unification
By constructing and processing a simulated drilling dataset, combining a one-dimensional convolutional neural network and a bidirectional long short-term memory network, and using an adversarial training mechanism to eliminate data distribution differences, high accuracy and robustness of lithology identification were achieved, solving the time-consuming and labor-intensive problem of lithology identification in tunnel construction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG UNIV
- Filing Date
- 2026-05-12
- Publication Date
- 2026-07-24
AI Technical Summary
Existing lithology identification methods are time-consuming, labor-intensive, and costly in tunnel construction. Machine learning models rely on a large number of field samples for training, but their accuracy is low in small sample scenarios. Simulated data and measured data are difficult to mix directly to improve model performance.
By constructing field datasets and simulated datasets, physical features are derived and preprocessed. One-dimensional convolutional neural networks and bidirectional long short-term memory networks are combined to extract drilling data features. Adversarial training and gradient inversion mechanisms are used to eliminate the distribution differences between simulated data and measured data, thereby achieving feature unification.
High accuracy and robustness of lithology identification were achieved with limited field data, overcoming the bottleneck of traditional methods that rely on large amounts of field-labeled data, and improving the model's adaptability and generalization ability in real engineering scenarios.
Smart Images

Figure CN122174028B_ABST