Abstract:In order to cope with the problem of logging data extrapolation that is restrained by seismic data,we revise the energy function of BP neural network and add an inversion objective-function as a new energy restraint ,so that people can do adaptive extrapolation of logging data and then combine BP network with seismic data inversion. We expound the theory and algorithm of the technique,and offer application example in the article. The extrapolation result of real logging velocity shows that this technique will find good application because it brings fast operation,good stability and high reliability.