| Seismic exploration data can be divided into active source seismic data and passive source seismic data.Conventional active source data have a expensive acquisition costs,exploration and construction are difficult,and often require a lot of material and financial support.However,passive source data is easy to obtain,low cost,easy to construct,and does not require excitation of large seismic sources.It has important research value in the fields of monitoring while drilling,underground space exploration,urban geophysical prospecting,oilfield fracturing monitoring,etc.It has also gained more and more attention in recent years.Passive source seismic data can be divided into pulse-type passive source and ambient noise passive source style.The former has low seismic energy and is usually used for processing research such as microseismic monitoring and microseismic positioning.Ambient noise passive source data is the most widely existing data type in nature.Natural background noise and noise generated by human activities can be regarded as ambient noise passive source data.Longer seismic record acquisition time can compensate for the weak seismic energy to a certain extent.This long record style makes it possible,that ambient nosie data can be used for seismic inversion and imaging.Usually,passive seicsimc data has a low signal-to-noise ratio,which makes passive seismic data often play a rola as a supplement and aid to the active source data,such as using the result of passive source data processing as a prior condition for the active source data,or using it when the active source data volume is missing.Passive source data is supplemented,and there are very few inversion and imaging studies using passive source data.With the continuous development of exploration technology,high-resolution inversion technology represented by full-waveform inversion and high-resolution imaging technology represented by reverse-time migration imaging have enabled seismic exploration to achieve considerable development and progress.It is directly used passive source seismic data can be used for inversion of underground physical parameters and imaging.At present,the processing methods of passive source seismic data are mostly based on seismic interferometry,such as cross-correlation method,deconvolution method,multi-dimensional deconvolution method,etc.Seismic interferometry is a powerful tool.The design idea of this type of technology is the data reconstruction method,which converts the passive source seismic data type to the active source seismic data type,and then the active source seismic data processing tool can be used for further data processing.Based on the reconstruction of passive source data,this paper also studies the ideas of passive source data direct migration imaging.The combination of the two makes it possible to rely solely on passive source seismic data for seismic imaging and inversion.First,the full waveform inversion method is researched.The adaptive momentum algorithm is combined with the multi-scale source encoding full waveform inversion,this kind of combination improves the efficiency of full waveform inversion and also makes full waveform inversion achive a better result.Subsequently,this paper proposes a passive source reverse time migration technology based on direct migration imaging,and according to the characteristics of passive seismic data,proposes imaging conditions for full waveform separation and normalization.Finally,this paper proposes an ambient noise passive seismic data processing method based on full waveform inversion and direct migration imaging.This method combines full waveform inversion,seismic interferometry technology and direct migration imaging technology.Firstly,the virtual shot gathers reconstructed by seismic interference technology,based on the characteristics of the virtual shot gathers,combined with the efficient full waveform inversion method previously studied,I developed a multi-scale source independent source encoding full waveform inversion method.The velocity model result of full waveform inversion is processed using reverse time migration technology based on direct migration imaging.Without relying on active seismic data,starting with a simple initial velocity model,a good imaging results are finally obtained.The methods and results presented in this article can be summarized as follows:(1)Combining the source encoding strategy in the full waveform inversion with the adaptive momentum optimization algorithm in the field of machine learning,a multi-scale source ecoding full waveform inversion based on the adaptive momentum algorithm is proposed.In order to improve the efficiency of the full waveform inversion and process data close to the dominant frequency in the actual situation,this paper adopts the multi-scale strategy of overlapping frequency bands and the source encoding technology in the full waveform inversion.In order to deal with the stochastic optimization problem brought by the source encoding technology,the adaptive momentum algorithm(Adam)is introduced from the field of machine learning,and the adaptive momentum source encoding full waveform inversion is proposed.Under the simple initial model,we compared the mini-batch full waveform inversion under the machine learning architecture and the conventional full waveform inversion.Numerical examples show that the proposed method has the highest computational efficiency and the best inversion accuracy results.(2)A passive source reverse time migration imaging method based on direct migration imaging is proposed.This paper derives the formula of direct migration imaging of ambient noise passive data in reverse time migration imaging,so that reverse time migration can be applied to ambient noise passive source seismic data.A passive seismic reverse-time migration imaging strategy with parallel subsection stacking is proposed,which improves imaging efficiency while reducing computational memory pressure.Numerical examples show that under the circumstance of the source location is unknown and the source excitation time is unknown,the passive noise source data can be directly imaged without using seismic interferometry to reconstruct the virtual shot collection.(3)A method that using passive seismic direct imaging boundry condition on reverse time migration is proposed,besides,a imaging condition designed for passive seismic data is proposed,which combined full-wavefield decomposition and normalization.This kind of imaing condition improves the imaging quality,enhancing the illumination deficiency caused by unknown passive sources.To remove the low-frequency artifacts of the cross-correlation,this paper using f-k domain wavefield decomposition method.To achive a better results under a low signal noise ratio,a normalized method is applied on full wavefield decomposition.Through this way,different parts in results are indenpently normalized,the final results are benefited by this normalization.Further more,this normalization can also overcomes the illumination problems caused by random fashion source posion.(4)A method for full waveform inversion and direct imaging processing of passive ambient noise data is proposed.First,I use seismic interferometry to reconstruct the virtual shot gathers.Combined with the previous research,a passive seismic multi-scale source encoding and source independent full waveform inversion is proposed.There are several improvements designed for passive data,such as it the selection of reference traces,the construction of the source encoding super shots,and the integral stochastic gradient desent optimization algorithms have been improved.After the inversion results are processed,the direct reverse time migration imaging method proposed above is used for imaging.This method integrates and improves seismic interferometry,full waveform inversion and direct reverse time migration imaging.On the basis of not relying on active source data,only passive source data is used,and successfully got the shallow geological structure,and it also reflects part of the deep structure. |
PDF Full Text Request