Research On Denoising Method Of DAS Seismic Signals

With the development of petroleum exploration technology in China,the geological environment in the exploration area has become more and more complex.These complex factors have aggravated the noise pollution of the seismic data.Distributed fiber acoustic sensor(DAS)technology is a revolution.However,the noise generated in the process of DAS has reduced the signal-to-ratio of the seismic data,which brings great difficulty to the multi-channel processing of the seismic data.Hence,the development of effective denoising technology has currently become an urgent requirement in the petroleum exploration research.In this paper,the denoising in petroleum exploration and acquisition has been firstly introduced to manifest its indispensability.Next,some researches related to noise suppression at home and abroad have been summarized.In addition,the basic theory,such as low rank matrix decomposition and tensor decomposition,related to the denoising method indicated in our paper has been elaborated.Finally,two methods of noise attenuation of seismic data have been proposed which has been an improvement of exisiting methods.Our detailed work are as follows:First,this paper proposes a denoising method based on low rank matrix approximation which can be applied to DAS data since the singular spectrum analysis is inadequate to suppress the non-Gaussian noise.This method constructs a new objective function by raising a non-convex penalty function.The limit of the non-convex penalty function can convert the non-convex problem into a convex optimization problem to further obtain the optimal solution of the objective function.And the Hankel matrix is able to reduce the rank of the data,which facilitates to separate the noise and the signal.Besides,the truncated singular value decomposition can not only remove the noise of the original data,but also greatly improve the computational efficiency.Second,the traditional robust principal component analysis method is usually used to process 2D seismic data,and it is inefficient to deal with non-Gaussian noise since it does not fully utilize the information among multidimensional seismic data.In view of this problem,this paper raises a DAS data denoising method based on tensor robust principal component analysis.This method applies the tensor robust principal component analysis to 3D seismic data and constructs an objective function based on tensor.The function can be iteratively solved by the alternating direction method of multipliers.In the iterative process,Hankel tensor random singular value decomposition is adopted to acquire the tensor nuclear norm.With the low rank signal tensor and the sparse noise tensor deduced,the seismic data denosing is accomplished.Finally,the method proposed in this paper which is based on low rank matrix approximation and tensor robust principal component analysis is applied to the actual seismic work area.The results show that our method is promising to suppress nonGaussian noise.