HESSIAN MULTIPLICATION ON VECTOR APPROXIMATION METHODS IN WAVE FIELD FULL INVERSION Russian patent published in 2017 - IPC G06F17/16 G01V1/50 

FIELD: data processing.

SUBSTANCE: invention relates to geophysical exploration, and, particularly, to seismic data processing. Method of depths area physical properties discrete model determining, which is model or depths model, by measured geophysical data iterative inversion, obtained from depths area, and containing: target function Hessian matrix approximation, which then is multiplied by vector describing Hessian multiplication on vector using computer, with help of single simulated direct wave propagation and single target function gradient calculation, in depths modified model, besides, it is necessary to use only three direct wave or backward wave spreads, wherein this approximation is based on approximation equation for Born scattered pressure field , where Hessian multiplication on vector is approximated by gradient calculating using as artificial excess, then calculating model parameters direction in space for updating to current model by Hessian matrix inversion multiplying by target function gradient, where Hessian matrix inversion is iteratively calculated using conjugated gradients method, wherein approximation of Hessian multiplying on vector is used to estimate Hessian matrix multiplication on environment disturbance vector, performing linear search for model updating amplitude determination using calculated direction, addition of model updating to current model to generate updated model and use of updated model for geophysical survey.

EFFECT: technical result is increased rate of value assessing, known as Hessian multiplication on vector, which occurs in certain methods for numerical solution of partial differential equations.

14 cl, 4 dwg