地震结构张量特征向量的确定、估算地层倾角方法及装置
By pre-training and unsupervised learning of deep neural network models, the problem of noise impact in the calculation of seismic structure tensors in 3D seismic data is solved, and efficient and accurate estimation of stratigraphic dip angle is achieved, which is applicable to the interpretation of seismic data in different work areas.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA NAT PETROLEUM CORP
- Filing Date
- 2023-06-30
- Publication Date
- 2026-07-17
AI Technical Summary
Existing techniques for calculating seismic structure tensors from 3D seismic data are significantly affected by noise, leading to unstable calculation results. Furthermore, the discrepancy between the attribute labels used in supervised training and the actual data affects the generalization ability, reducing the accuracy and efficiency of stratigraphic dip estimation.
A deep neural network model is pre-trained, and parameters are adjusted through unsupervised learning. Combined with convolutional neural networks and loss function optimization, stable calculation of feature vectors is achieved. Transfer learning and parameter fine-tuning are performed using actual seismic data to avoid dependence on label data.
It improves the efficiency and applicability of seismic data interpretation, enabling accurate estimation of stratigraphic dip angles under different work areas and data characteristics, simplifying the calculation process, and enhancing the stability and accuracy of seismic interpretation.
Smart Images
Figure CN119224837B_ABST