Research On Deblending And Its Application To Marine Simultaneous Source Data

The simultaneous source technology(also known as multi-source blended acquisition) has become a popular topic for both industry and academic due to its outstanding features of improving the acquisition efficiency and imaging quality over recent years. However, the seismic record would include one or more adjacent shot gathers(also known as blending noise), some blank traces containing more subsurface information than traditional survey and some other random noise which will contaminate the seismic record; the inhomogeneous blended acquisition developed from conventional blended acquisition could conduct broadband acquisition in a simpler and more fidelity way, but the separation SNR would decreased separated by common methods; Owing to the source characteristics and exploration cost the commercialization speed of marine survey was slower than that of the land’s. Given above-mentioned limitations during the developing stage of simultaneous source acquisition, this thesis has made a systematic study of simultaneous source technology. The thesis primarily studied the following contents: the fundamental theory of simultaneous source acquisition, the separation methodology of simultaneous source data, the theory and developmental tendency of inhomogeneous blended acquisition, the first study of the acquisition and processing of the marine simultaneous sources data in China which will provide theory support and technology accumulation for industrialization of marine simultaneous source technology in the future.The thesis addressed the fundamental theory of simultaneous source acquisition from the aspects of the geometry as well as math physics. Based on the synthesis operator and areal source, the concept of blended operator and blended source were introduced. Then the thesis illustrated source density ratio and survey time ratio, which quantify the advantage of improving spatial sampling rate and acquisition efficiency. The blended source was classified as fully determined and underdetermined simultaneous source acquisition depending on the type of the blended operator. Both the blending mode and the impact on separation method corresponding to each blended operator were discussed from the view of math physics. Based on the theory of simultaneous source acquisition, the marine multi-vessel and single vessel simultaneous source were addressed respectively, and the thesis also carried out the forward modeling of aforementioned two marine geometries. The concept of pseudodeblending was presented from the perspective of theory and forward modeling, which would perform data regularization and QC(quality control) of blending operator.The thesis developed a new method which would solve three problems simultaneously: removal of blending noise, interpolation and some other random noise attenuation, and it will increase the efficiency of the seismic data processing and improve the SNR. Based on the same architecture of inversion, the thesis brought these three problems into L1 norm regularization model, and a fast iterative shrinkage-thresholding algorithm was employed to solve L1-norm minimization problem, which speeded up the performance of shrinkage-thresholding algorithm(ISTA) and preserved the computational simplicity of ISTA at the same time. The method was verified by synthetic data dataset and real dataset respectively.The thesis proposed a new separation method of inhomogeneous blended acquisition, and based on synchrosqueezed wavelet transform this new method would offer better separation SNR and faster computational time than any other methods for high overlapping of sources units’ band. Subjected to the acquisition environment and equipment constraints, the broadband acquisition technology at this stage would make the source or the whole geometry complex. Based on the conventional simultaneous source technology, instead of employing broadband source alone, the inhomogeneous blended acquisition equipped with multi-narrowband sources with different domain frequency was introduced, and this new geometry would fulfil broadband acquisition with less complex devices. The thesis showed the application and development trend of inhomogeneous blended acquisition onshore and offshore. The inhomogeneous blended acquisition could be develop into decentralized blended acquisition which is more simple and efficiency to operation while keep the former advantages. However, when the inhomogeneous and conventional blended acquisitions adopt the same separation methods, the former has low signal-to-blending(SNR) noise. On the basis of the time-frequency differences of source units in inhomogeneous blended array, the thesis developed a new separation method which would offer better separation SNR and faster computational time and it was also proved by synthetic models.The source separation quality has great relationship with the blended operator, hence the thesis mainly studied the two key factor of blended operator: the range where the random source emission time picks randomly and the minimum emission time interval of source units in blended array. Combined with the geometry and seismic project of 2D and 3D marine simultaneous source acquisition implemented first in China, the thesis has developed a design method of optimal blended operator.Aiming at marine simultaneous source data acquired first in China, the thesis proposed a processing strategy of simultaneous source data by combining separation process with conventional processing flows. Then the processing strategy was employed to conduct preprocessing, separation and migration of simultaneous source data. By comparing the pre-stack migration of simultaneous source and traditional dataset from the same area under the same condition of computer hardware, software and processing parameter,the separation method and processing strategy of simultaneous source data were further verified.A new separation method was proposed in this thesis. Before the separation process of the simultaneous source data, the model of direct arrival was built and then subtracted adaptively from the simultaneous source dataset. The fidelity of the signal separated was improved and the S/N of deblended shot gather could be doubled through this new method.