Fault Detection And Poststack Data Denoising Based On Deep Learning

Fault identification is an important task of seismic interpretation.With the advance of deep learning,deep learning methods have attracted huge attention in the industry as they have substantially improved automation compared to traditional manual or semi-manual based fault detection methods.A large number of research has proved that the deep learning method has achieved a good application result in seismic interpretation.Therefore,starting from building neural network,this paper explores the application prospect of fault identification and horizon tracking technology using deep learning.Aiming at 2D fault detection and horizon tracking task,this paper utilizes synthetic seismic data and uses the strong representation ability of the deep learning network to build a neural network to learn geophysical knowledge in synthetic seismic data.The trained network can be directly used to solve the problems of fault identification and horizon tracking.U-net has achieved good results in the application of2 D real seismic data.To further enhance the effectiveness of the deep learning methods,we propose a novel network architecture,named R2SE-Unet,to solve the 3D fault segmentation and horizon tracking problems.In the neural network,we design a recurrent residual-SE convolution unit(RRCU-SE)to store information over time in 3D seismic data.This component promotes the spread of 3D volumetric information and help the network learn spatial correlations in 3D images.In addition,to reduce the impact of insufficient spatial resolution resulting from the base architecture of U-net,we add an attention unit between skip connection operations.These two new units enable our R2SE-Unet to exploit semantic information more accurately in the feature maps.After many experiments on region-based loss functions and distribution-based loss functions,we also propose a novel loss function,which takes the advantage of generalized dice(GDice)loss and balanced binary cross entropy(b-BCE)loss,named Gdice-bce,to effectively train R2SE-Unet.Although only synthetic seismic data samples are used to train the network parameters,our R2SE-Unet could produce more reliable fault feature maps on field seismic data than other conventional fault-detection neural networks.In terms of denoising,seismic coherent noise and random noise usually exist in the post-stack seismic data,affecting the resolution and integrity of seismic images.It is difficult to remove the coherent noise since the pattern of coherent noise is highly related to signals.Recently,deep learning has proven to be uniquely advantageous in image denoise problems.To enhance the quality of the post-stack seismic image,in this paper,we propose a novel deep residual-learning-based neural network named DR-unet to efficiently learn the feature of seismic coherent noise,which includes an encoder branch and a decoder branch.Moreover,to solve the problem of training data,we propose a workflow that combines real seismic noise with synthetic seismic data to construct the training data.Experiments show that our method can achieve good denoising results in both field and synthetic seismic data.