Use this URL to cite or link to this record in EThOS:
Title: Seismic reflectivity and impedance inversion in multichannel fashion
Author: Wang, Ruo
ISNI:       0000 0004 6348 1008
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2015
Availability of Full Text:
Access from EThOS:
Access from Institution:
Seismic reflectivity inversion is an important step in both signal processing and quantitative interpretation since reflectivity contains the information of impedance and other elastic parameters. Conventional methods assume stratified media and perform deconvolution on seismic data trace by trace. However, when using these single-channel methods, the lateral coherency of the result may be affected when the input seismic traces have low signal-to-noise ratios (SNRs) or complex structures. In this thesis, the development of multichannel inversion algorithms will be investigated to improve the continuity of the reflectivity profiles and suppress the noise. An example of a widely-used seismic reflectivity inversion method that is applied to single-channel seismic trace is the basis pursuit method. In this thesis, an FX prediction filter is incorporated with the conventional BP method, to investigate the potential benefit of multichannel implementation. Since the dictionary used in BP is huge in size, the matrix operations are very time consuming, yielding a long overall computational time. To improve the efficiency, a GPU-accelerated basis pursuit method is further implemented under CUDA architecture. Numerical results with the same accuracy are obtained and a speedup factor up to 145 for the whole process has been achieved. To implement a multichannel reflectivity inversion, the curvelet transform is employed. A comparative study on the performance of the curvelet deconvolution with two other widely used methods, the least-squares method and Lp-norm deconvolution, is further conducted. Since the deconvolution based on the curvelet transform offers a good trade-off between the lateral continuity and sparseness, the curvelet deconvolution result is used as the initial model to enhance the Lp-norm deconvolution. Numerical results show that the lateral continuity of the spiky reflectivity profile can be further improved. Moreover, I develop a proper multichannel deconvolution method based on the Cauchy constraint. In this algorithm, a multichannel prediction operator is integrated into the iteration process. In this way, the information of the adjacent traces is exploited during the inversion procedure. This method can provide results with improved lateral coherency, better structure characterisation ability and lower residual energy ratio. I also develop two modified processes based on the original Cauchy constrained multichannel method. These two modifications can give better quality for the reflectivity inversion results, when compared with the original algorithm. Finally, using a similar concept as the multichannel deconvolution method with the Cauchy constraint, I apply the multichannel inversion algorithm to seismic impedance inversion, with the input reflectivity series obtained from the previous inversion steps. Numerical results show impedance profiles with better structure identification and lateral continuity.
Supervisor: Wang, Yanghua Sponsor: China Scholarship Council
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral