Fault detection method and device

Abstract

The invention provides a fault detection method and device. The method comprises the following steps: acquiring a fault sample data set; constructing an initial convolutional neural network model for fault detection based on a deep learning algorithm according to the fault sample data set; based on a self-attention gate mechanism, eliminating specified features and noise influences in the initial convolutional neural network model, and establishing a convolutional neural network model used for fault detection; acquiring real-time seismic data; and inputting real-time seismic data into the convolutional neural network model to obtain a fault detection result. According to the method, fault detection of seismic data is realized by establishing the convolutional neural network model, significant features of the fault in the model are emphasized by introducing a self-attention gate mechanism, the accuracy of fault feature selection is improved, and the prediction accuracy of the established convolutional neural network model is improved while the calculation efficiency is ensured. Therefore, the fault detection precision is improved.

Description

technical field [0001] The invention relates to the technical field of seismic signal processing and interpretation, in particular to a fault detection method and device. Background technique [0002] A fault is a structure in which the crust is fractured by force, and the rock blocks on both sides of the fracture surface undergo significant relative displacement. The scale of the faults varies. The larger ones can extend up to thousands of kilometers along the strike and can cut through the crust downwards. They are usually composed of many faults and are called fault zones; the smaller ones are measured in centimeters and can be found in rock specimens. . [0003] Oil and gas are produced under the control of faults, and exist according to fault sealing; faults can allow oil and gas to migrate, escape, or cut the reservoir into several parts. Therefore, the results of fault interpretation are an important basis for geological modeling and subsequent reservoir evaluation ...

AI Technical Summary

Problems solved by technology

There are two fault detection methods in the prior art. One is to manually pick faults from the seismic section by using the experience of interpreters and the assistance of some seismic attributes, which often takes a lot of time and the accuracy is difficult to guarantee; the other is automatic fault tracking technology. , generally edge detection is performed on slices along the layer, and at the same time combined with attribute information, the efficiency is greatly improved compared with manual interpretation, but due to inaccurate selection of fault features, the accuracy often cannot meet production requirements

Therefore, there is a problem of low detection accuracy in the existing tomographic detection technology

Images