Study And Application Of Particle Swarm Optimization Algorithm In Rayleigh Wave Inversion And Converted Wave Static Correction

Seismic tomography, residual static correction, deconvolution, AVO and other issuesare the inverse problem of seismic exploration. Inversion can be divided into linear(quasi-linear) and nonlinear inversion according to different algorithms. Theadvantages of linear inversion are comparatively ripe theory, fast operation and so on.However, the disadvantage of it is falling into the local optimal solution since thealgorithm can be easily influenced by initial model. Optimization of nonlinearinversion doesn’t depend on initial model and has the ability of jumping out of localoptimal solution and finding global optimization in theory, but the weaknesses areslow convergence speed and the algorithm performance are largely influenced byparameter selection.Firstly, we introduce particle swarm optimization in this paper. Particle swarmoptimization is one of the nonlinear optimizations and its main thought is simulatingthe behavior of a flock of birds or a school of fish. Standard particle swarmoptimization own advantages such as, fast convergence speed, easy to implement,being prone to parallel operation and less call parameters. However, the disadvantageof it easily occurs “premature”. In this paper, we introduced group particles to formgroup particle swarm optimization based on standard particle swarm optimization.Group particles are one special class and all the group particles keep a certain distanceduring the algorithm execution process, thus it decreases “premature” of particleswarm optimization phenomenon caused by aggregation of particles in some degreeand the new algorithm has powerful ability of global searching.Secondly, our study focuses on application of group particle swarm optimization applied on the inversion of Rayleigh wave dispersion curves. Rayleigh waves are atype of seismic wave caused by the interference of P and SV wave at free surface.Rayleigh waves have dispersion phenomenon in multilayer-medium and their phasevelocity changes along with frequency is defined as dispersion curves. Dispersioncurves relative to compressional and shear wave velocities, density and layerthickness, what’s more, shear wave velocity and layer thickness among them areplaying more important roles. Thus we can use dispersion cures to inversenear-surface shear velocity structure. Levenberg–Marquardt algorithm is one ofcommon algorithms for Rayleigh waves inversion and a linear algorithm which hasthe merits, such as fast convergence speed. However, its results depend on the initialvalue. Group particle swarm optimization as a nonlinear algorithm can be independentof initial value and get global optimal solution, but it has lower convergence speed.Rayleigh waves inversion can only inverse fundamental mode of dispersion curve andalso can joint inverse multi-modes. Research results show that adding higher modes toinversion can increase accuracy of results and stability of inversion. The experimentsresults also prove the above conclusion.Last but not least, we study the problem that large short wavelength shear wave staticscorrection. With the development of society, the demand for resources is increasingand which require geologists and geophysicists to get a more meticulous and accurateunderstanding of underground. On one hand scholars continue to developcompressional wave exploration technology. On the other hand they research anddevelop new theories and technologies of seismic exploration, such asFour-dimensional seismic exploration, multi-wave and multi-component seismicexploration. Shear wave has some properties which compressional wave does nothave, such as its propagation which is seldom sensitive to the porous fluid and shearwave splitting through fractures. Shear wave material plays an important role in determination of lithology, fracture and oil-gas possibility. Shear wave explorationwhich needs excitation of shear wave source is hard to be applied to practicalproduction, since shear wave source is expensive and the penetration depth of shearwave is small. Shear wave exploration using converted wave information hasadvantages of larger exploration depth and lower cost comparing with the one usingshear wave source. Static correction is an important step of seismic exploration and itwill largely influence the following steps. For converted wave, its excitation iscompressional wave and we can get shear wave information, so its static correctioncan be divided into static correction at source and static correction at receiver.Normally, compressional wave seismic data processing has been finished beforeconverted wave seismic data processing, so the static correction at source can beconsidered already being known which means converted wave static correction is thestatic correction at receiver. Since shear wave propagation is not influenced by theporous fluid, its low velocity zone is always below water table and shear wavevelocity is lower, its static correction is often2~10times as much as compressionalwave static correction at the same position. Converted wave static correction atreceiver is often large and changes dramatically and short wavelength converted wavestatics correction is typical nonlinear and multi-minimum problem, so it is hard to geta good result when using linear method. In order to estimate static correction of shortwavelength converted wave at receiver is large. In this paper the stack section energyis considered as objective function and group particle swarm optimization and themaximum energy method are serially fused, then we can estimate static correction ofshort wavelength converted wave at receiver by using serial algorithm. Model testproves that this method can get better results.