# US 2021/0239879 A1 · Method For Partitioning A Search Direction When Using Least Squares Reverse Time Migration

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0239879 A1
- Document ID: US20210239879A1
- Status: Published application
- Kind: A1
- Application number: 17/247,528
- Filed: 2020-12-15
- Published: 2021-08-05
- Earliest claimed: 2020-01-31
- Assignees: ExxonMobil Upstream Research Company
- Applicants: ExxonMobil Upstream Research Company
- Inventors: Valeriy V. Brytik; Junzhe Sun; Carey M. Marcinkovich; Hong Zhao; Erik R. Neumann
- Web page: https://patentgrep.com/patent/US20210239879A1

## CPC

- G01V 1/282: {Application of seismic models, synthetic seismograms}
- G01V 20/00: Geomodelling in general
- G01V 1/302: {in 3D data cubes}
- G06F 17/18: for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)}
- G06F 30/20: Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30)
- G06F 30/27: using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
- G06F 2111/10: Numerical modelling
- G01V: GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS (means for indicating the location of accidentally buried, e.g. snow-buried, persons A63B29/02)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method for partitioning a search direction when using least squares reverse time migration (LSRTM) is provided. LSRTM may be used iteratively in order to improve imaging accuracy. As part of LSRTM, multiple local line searches may be performed. In particular, image space may be partitioned, such as by using a set of masks. The search direction, such as the gradient, may be partitioned using the set of masks. Local line searches may be performed for each partition of the search direction, resulting in finding respective line search constants. The respective line search constants may then be used for iterating the model in order to improve imaging accuracy.

## Claims

### Claim 1

A computer-implemented method comprising:
computing, using at least one computer and a subsurface property model, an objective function measuring misfit between model-simulated data and measured geophysical data;
computing, using the at least one computer, a search direction in order to update the model;
partitioning the search direction;
performing multiple line searches for the partitioned search direction in order to generate respective line search constants; and
updating the model using the partitioned search direction and the respective line search constants.

### Claim 2 (depends on 1)

The method of claim 1, wherein the search direction comprises a gradient.

### Claim 3 (depends on 2)

The method of claim 2, wherein the model is updated using least squares reverse time migration.

### Claim 4 (depends on 3)

The method of claim 3, wherein partitioning the search direction comprises:
introducing a 3D partition of unity in image space;
generating masks corresponding to the 3D partition of unity; and
partitioning the gradient using the masks.

### Claim 5 (depends on 4)

The method of claim 4, wherein generating masks corresponding to the 3D partition of unity comprises M1 to Mn, where n is greater than 1; and
wherein partitioning the gradient (Grad) using the masks comprises splitting the gradient via the masks to generate M1Grad to MnGrad.

### Claim 6 (depends on 5)

The method of claim 5, wherein the respective line search constants generated by the multiple line searches comprise α1 to αn; and
wherein the model update comprises Δm=α1M1Grad to αnMnGrad.

### Claim 7 (depends on 6)

The method of claim 6, wherein the respective line search constants are selected independently of one another.

### Claim 8 (depends on 6)

The method of claim 6, wherein the respective line search constants are selected at least partly dependently on one another.

### Claim 9 (depends on 8)

The method of claim 8, wherein at least some of the respective line search constants are smoothed based on other line search constants in proximity thereto.

### Claim 10 (depends on 4)

The method of claim 4, wherein a first set of masks are used to partition the gradient;
wherein a second set of masks are used to perform the multiple line searches; and
wherein the first set of masks is different from the second set of masks.

### Claim 11 (depends on 10)

The method of claim 10, wherein the second set of masks are wider than the first set of masks.

### Claim 12 (depends on 11)

The method of claim 11, wherein the second set of masks includes an entirety of the first set of masks and additional area.

### Claim 13 (depends on 4)

The method of claim 4, wherein a same set of masks are used to partition the gradient and to perform the multiple line searches.

### Claim 14 (depends on 1)

The method of claim 1, wherein partitioning the search direction is based on at least one of spatial frequency or masks.

### Claim 15 (depends on 1)

The method of claim 1, further comprising dynamically determining a number of partitions for the search direction; and
wherein partitioning the search direction is in the number of partitions.

### Claim 16 (depends on 15)

The method of claim 15, wherein dynamically determining the number of partitions for the search direction is based on analysis of a number of events in an image generating by the model.

### Claim 17 (depends on 16)

The method of claim 16, wherein the model is updated iteratively; and
wherein the number of partitions for the search direction remains constant for the iterations of the model.

### Claim 18 (depends on 16)

The method of claim 16, wherein the model is updated iteratively; and
wherein the number of partitions for the search direction changes for the iterations of the model.

### Claim 19 (depends on 1)

The method of claim 1, wherein updating the model comprises a Born-approximation based implementation of tomographically enhanced full wavefield inversion.

### Claim 20 (depends on 1)

The method of claim 1, wherein a full wavefield inversion (FWI) operation comprises a first inversion and a second inversion;
wherein the first inversion includes a linear objective function;
wherein the second inversion includes a non-linear objective function;
wherein partitioning the search direction and performing multiple line searches for the partitioned search direction is performed for the linear objective function; and
wherein a single line search for an unpartitioned search direction is performed for the non-linear objective function.
