Analysis of onedimensional seismic waveform inversion by Regularized Global Approximation.pdfVIP

Analys i s of onedimens ional s e i smic waveform invers ion by Regular ize d Global Approximation Gennady Ryzhikov Mar ina Biryulina and Shmar iahu Keydar key words global optimization forward mo deling invers ion nonuniquene ss ob jective function regular ization soft constraint Ab stract Direct analys i s of normal incidence s e i smogram invers ion with re sp ect to a velo city prole i s available now due to applying of a new global optimization algor ithm The latter i s bas e d up on regular ize d global approximation of an ob jective function which i s not supp os e d to b e dierentiable The new technique allows to s ee clearly a nonuni quene ss of the invers ion problem no matter how high i s a quality of the input data It i s induce d by a few f actors a source wavelet i s a function of a nite f requency band an eective wave length of the sounding s ignal i s increas ing jointly with the velo city and the p ower of a me dia re sp ons e i s decreas ing with re sp ect to the depth The nonunique ne s s means that there i s no invers ionpro ce s s ing enable to solve the problem if it do e s not take into account a pr ior i information ab out the recovere d velo city prole It i s shown how an a pr ior i assumption ab out a trend of the prole can e ss entially re duce the nonuniquene ss of the problem The corre sp onding regular ization has a form of a soft constrain on the mi st function and leads to an unbias e d e stimation of the velo city prole when the latter i s monotonous function with re sp ect to the depth On the other hand the regular ization sugge ste d allows to reconstruct nonmonotonous functions as well which leads to a bias e d e stimation of the velo city prole as like as any conventional regular ize d invers