| Since2003, Holbrook successfully apply the seismic reflection method tophysical oceanography studies at the first time, seismic oceanography has been rapiddevelopment. At present, the seismic oceanography has been able to image somephysical oceanography phenomenons such as marine fronts, water mass boundaries,mesoscale vortex and internal wave. In addition, some geophysicists also use AVOinversion, wave impedance inversion and full waveform inversion to obtain thevelocity, temperature and salinity of water from the seismic data.Currently seismic oceanography is used most in deep oceans, compared with thedeep oceans, coastal waters are more closely related to human activities. Thethermohaline structure of water is important for building large Sea-Crossing Bridge,subsea tunnel and coastal port.The full waveform inversion seeks to minimize the amplitude misfit betweenobserved and synthetic seismic records to inverse the optimal model. It has thepotential to accurately depict the details of the water structure. Seismic oceanographydata are especially well suited to one-dimensional waveform inversion approaches,because thermohaline boundaries in the ocean are very nearly flat and horizontal,lateral variations in sound speed are small, no converted shear waves are present andinterbed multiples are negligible.This article uses the time domain acoustic wave equation to forwardone-dimensional water seismic waveform. Since the shallow sea has less modelparameters and forwarding calculation cost little, nonlinear inversion can beconsidered. Compared with the local linearization,nonlinear inversion has a smallerdependence on the initial model and no need to calculate derivative information.Due to the complex nonlinear relationship and a number of extreme pointsbetween the objective function and the model parameters in full waveform inversion,conventional genetic algorithm prone to premature leaving the inversion results into local minima. To solve this problem, niche technology can be used to enrich thediversity of GA population. In addition, as PSO has the advantages of condensedconcept, less parameter settings, fast convergent rate and low population sizerequirement, it can quickly and greatly increase the fitness of random initialpopulation before NicheGA inversion and accelerate the convergence rate. Smallerinertia weight PSO can also local search the NicheGA inversion results for furtheroptimization. Therefore, this paper uses a algorithm based on PSO combined withNicheGAwhich get good results by testing the actual CTD synthetic records. |
