Vrms inversion by ray tracing

<< Click to Display Table of Contents >>

Navigation:  Velocity >

Vrms inversion by ray tracing

 

Description

This module inverts a VRMS (root-mean-square velocity) model to produce an updated interval velocity (VINT) model by iterative ray tracing. For each location in the model, the module shoots rays through a current VINT estimate, computes the resulting synthetic VRMS, compares it to the measured input VRMS, and adjusts the VINT model to reduce the residual. Tikhonov regularisation is applied laterally and vertically to ensure a smooth, stable output model. Both 2D and 3D data are supported.

Use this module when you need a depth interval velocity model that accurately reproduces a given VRMS field and you want a more physically rigorous result than a simple Dix inversion can provide. An optional true VINT reference model can be supplied to constrain or validate the inversion result.

 

Input data

Input Vrms model

The target VRMS velocity model that the inversion aims to reproduce. This is typically the result of semblance picking or another time-domain velocity analysis. The inversion iteratively adjusts the interval velocity model until the synthetic VRMS derived from it matches this input.

Input Vint model

The starting interval velocity model used to initialise the inversion. This could be the result of a previous Dix conversion, refraction tomography, or another initial depth model. A better starting model typically leads to faster convergence.

True Vint model

An optional reference interval velocity model (for example from well logs) used to constrain or validate the inversion. When connected, the inversion can use this reference to weight the solution towards known velocity values in areas of good well control.

Parameters

V0

The near-surface velocity (m/s) used as the boundary condition for ray tracing at the surface. The default is 1500 m/s. Set this to match the velocity at the datum or water bottom for marine surveys.

Output model datum

The reference datum elevation (m) of the output VINT model. The default is 0 m. Set this to the survey datum elevation if the data are referenced to a non-zero datum.

Output model depth

The total depth (m) of the output interval velocity model. The default is 5000 m. Set this to at least the depth of the deepest target horizon. Values deeper than the maximum ray penetration depth will be extrapolated.

Output model sample ratio

The vertical sample spacing (m) of the output VINT model. The default is 10 m. A finer sample spacing produces a more detailed vertical velocity profile but increases the output model size.

Grid params

Container grouping all ray-tracing geometry parameters. These control how densely rays are shot through the model, the maximum offset and angle they cover, and the spatial resolution of the inversion grid.

Ray cluster step X/Y

The lateral spacing (m) between ray-tracing source clusters in the inline and crossline directions. The default is 500 m. A smaller step produces denser lateral ray coverage and a higher-resolution inversion result, but significantly increases compute time.

Ray cluster step Z

The time step (s) between ray-tracing source levels. The default is 0.1 s. This controls the density of reflectors sampled by the ray tracing in the time direction.

Variable step X/Y

The lateral grid spacing (m) used for the velocity update step. The default is 500 m. This determines the spatial resolution at which velocity corrections are applied during each inversion iteration.

Variable step Z

The depth step (m) used for the velocity update grid. The default is 200 m. This determines the vertical resolution at which velocity corrections are applied.

Ray shoot step X/Y

The lateral spacing (m) between individual ray shooting points within each cluster. The default is 100 m. Together with the cluster step, this defines the total lateral density of rays used in the inversion.

Maximum ray offset

The maximum source-receiver offset (m) at which rays are traced. The default is 2000 m. Set this to match the maximum usable offset in the seismic data. Rays beyond this offset are not included in the inversion.

Ray time step

The time step (s) used when integrating ray paths through the velocity model. The default is 0.004 s. Smaller values improve ray-tracing accuracy but increase computation time. This should typically match the seismic data sample rate.

Solver params

Container grouping the iterative solver parameters and Tikhonov regularisation weights that control convergence and the smoothness of the output interval velocity model.

Number of global iterations

The number of outer inversion iterations. The default is 10. Each global iteration re-traces rays through the updated velocity model and computes new residuals. More iterations improve convergence but increase total run time. Monitor the residual norm to determine when convergence is sufficient.

Number of local iterations

The number of inner solver iterations per global iteration. The default is 1000. These iterations refine the velocity update within each global step before rays are re-traced.

Tikhonov lambda X/Y

The Tikhonov regularisation weight applied in the lateral (inline and crossline) directions. The default is 1. Higher values produce a smoother lateral velocity variation; lower values allow the inversion to fit sharper lateral gradients. Increase this when the input VRMS model is noisy or sparsely sampled laterally.

Tikhonov lambda Z

The Tikhonov regularisation weight applied in the vertical (depth) direction. The default is 1. Higher values enforce a smoother velocity profile with depth and suppress high-frequency oscillations in the interval velocity function.

Normal weight

The weight applied to the data fit term in the inversion objective function. The default is 1. Increasing this value emphasises matching the input VRMS more closely at the potential expense of model smoothness.

V0 weight

The weight applied to the near-surface velocity constraint in the inversion. The default is 1. A higher value anchors the shallow portion of the inverted model more tightly to the specified V0 value, which can stabilise the inversion when near-surface data coverage is sparse.

Visualization params

Container grouping parameters that select which inline and crossline slices of the model are displayed in the g-Platform graphics panel during and after processing.

Inline to show

The inline number displayed in the velocity model graphics panel. Set this to a representative inline through the area of interest to monitor the inversion progress visually.

Crossline to show

The crossline number displayed in the velocity model graphics panel. Set this to a representative crossline through the area of interest to complement the inline display.

Settings

Execute on { CPU, GPU }

Selects whether the module runs on the CPU or GPU. GPU execution can accelerate the iterative ray-tracing inversion for large 3D velocity models.

Distributed execution

Enables distributed processing across multiple compute nodes. Configure node addresses and resource allocation in the cluster settings.

Bulk size

The number of data units sent to each compute node in a single work package during distributed execution. Increase for large datasets to reduce communication overhead.

Limit number of threads on nodes

When enabled, restricts the number of CPU threads used on each distributed node. Use this to leave resources available for other processes running on shared hardware.

Job suffix

An optional text label appended to the job name when running in distributed mode. Use this to distinguish multiple simultaneous jobs in the cluster queue.

Set custom affinity

When enabled, allows the user to specify which CPU cores the module may use via the Affinity mask. Leave disabled to allow the operating system to schedule threads automatically.

Affinity

The CPU core affinity mask. Active only when Set custom affinity is enabled. Specify as a bitmask or core list to pin execution to particular CPU cores.

Number of threads

The number of CPU threads used for local processing. Set to match the number of available physical cores for best performance.

Skip

When enabled, the module is bypassed during batch execution. Use this to temporarily disable the module without removing it from the processing flow.

Output data

Output Vint model

The updated depth interval velocity model produced by the ray-tracing inversion. This model reproduces the input VRMS field more accurately than a simple Dix conversion because it accounts for ray-bending effects. It can be used directly for depth imaging, further tomographic updates, or as a reference model for subsequent iterations of the inversion.

Information

Graphics

Custom actions