Efficient line search methods for multi-parameter full wavefield inversion

Patent metadata

Field	Value
Publication number	US-9977142-B2
Application number	US-201514670039-A
Country	US
Kind code	B2
Filing date	Mar 26, 2015
Priority date	May 9, 2014
Publication date	May 22, 2018
Grant date	May 22, 2018

Abstract

Method for simultaneously inverting full-wavefield seismic data ( 51 ) for multiple classes of physical property parameters (e.g., velocity and anisotropy) by computing the gradient ( 53 ), i.e. search direction ( 54 ), of an objective function for each class of parameters, then applying (preferably exhaustive) first-pass independent line searches to each parameter class to obtain the corresponding step size ( 55 ) along the search direction for each parameter class; then without yet updating the model, using the step sizes to define a relative scaling between gradients of all parameter classes. Next, each scaled search direction is recombined to form a new search direction ( 56 ), and a new second-pass line search is performed along the new search direction ( 57 ), and all parameters are simultaneously updated with the resulting step size ( 58 ). Alternatively to the preceding alternating two-pass embodiment, the model may be updated after each first-pass line search, and no second-pass line search is performed.

First claim

The invention claimed is:

1. A computer-implemented method for iteratively inverting seismic data to simultaneously infer a model for at least two physical properties of the subsurface, said method comprising:
(a) for each physical property, computing, with a processor, a gradient of an objective function with respect to parameters of the physical property, said objective function measuring misfit between all or part of the seismic data and corresponding model-simulated seismic data;
(b) for each physical property, computing, with the processor, a search direction in model space from the gradient;
(c) alternating line searches between or among the search directions for the at least two physical properties to determine, with the processor, optimal step sizes along each of the search directions;
(d) using the optimal step sizes to update the model with the processor; and
(e) generating, with the processor, an image of the subsurface displaying the at least two physical properties according to the updated model;
wherein (c)-(d) comprise:
(i) performing a first line search for a first physical property, using an initial model, and saving the optimal step size from the first line search but not updating the model yet;
(ii) performing a second line search for a second physical property, using the initial model, and saving the optimal step size from the second line search but not updating the model yet;
(iii) performing a line search using the initial model and saving the optimal step size as in (ii) for any additional physical properties being modeled; and
(iv) using the saved step sizes from (i)-(iii) to compute a new search direction, then performing a second-pass line search along the new search direction, determining a second-pass optimal step size and using it to generate an updated model for a current iteration of the inversion.

2. The computer-implemented method of claim 1 , wherein in (a) one or more parts of the seismic data are selected and used for computing the gradient for each physical property, said one or more parts being determined according to a selected criterion based on sensitivity of the physical property to the data.

3. The computer-implemented method of claim 2 , further comprising, in a next iteration of the iterative inversion, expanding the selected part or parts of the seismic data for at least one of the at least two physical properties to include more parts of the seismic data.

4. The computer-implemented method of claim 1 , wherein the first and second and any additional line searches in (iii) are exhaustive line searches, performed to find inflection points in the objective function as a function of the step size.

5. The computer-implemented method of claim 1 , wherein some source shots in the seismic data are discarded and not used in the method in order to reduce computing cost and time.

6. The computer-implemented method of claim 1 , wherein the data inversion is full wavefield inversion.

7. The computer-implemented method of claim 1 , wherein the at least two physical properties comprise at least two of P-wave velocity, S-wave velocity, normal move-out velocity, one or more anisotropy parameters, and attenuation.

8. The computer-implemented method of claim 1 , wherein the search direction is computed without preconditioning the gradient with a Hessian of the objective function.

Assignees

• Tang Yaxun
• Ayeni Gboyega
• Exxonmobil Upstream Res Co

Inventors

• Tang Yaxun
• Ayeni Gboyega

Classifications

• G01V2210/63

  Seismic attributes, e.g. amplitude, polarity, instant phase · CPC title

• G01V1/32 Primary

  Transforming one recording into another {or one representation into another} · CPC title

• G01V1/282 Primary

  Application of seismic models, synthetic seismograms · CPC title

• G01V2210/42

  Waveform, i.e. using raw or pre-filtered trace data · CPC title

• G01V1/303

  for determining velocity profiles or travel times · CPC title