The Workflow
This stage of inversion is defined by the injection of the source wavelet into the starting Earth model.
The Forward Modelling (or Shot QC) stage
When running FWI without triggering the inversion step, this expression must be added to the output.
.IOSimulationTestOnly
As a result of the convergence process, we expect to see a flipping of the polarity of this property we QC.
The Gradient
This type of FWI, commonly referred to as Conventional FWI, uses an amplitude based objective function.
Least Squares FWI
This represents the difference between the observed and predicted shot gathers, and can be output for QC during shot collection.
The Cost Function
This process is run to estimate an optimal alpha scalar and determine the magnitude of update for a model property.
Line Search process
Setting this parameter to "TRUE" will allow FWI to account for multiple and ghost energy.
The Free Surface parameter (Surface Related Effects parameter set)
Cycle skipping may occur when a time shift larger than this threshold exists between modelled and observed data.
1/2 a cycle (1/2 a wavelength)
This type of FWI differs mainly in the type of modelling performed, Born modelling.
Reflection FWI (RFWI)
These headers must be included on the input shot gathers with a negative value to avoid an FWI abend.
ELEV_SOURCE and ELEV_DETECT
This stage of the workflow, unique to each survey, must be carefully considered before running FWI to preserve low frequencies, retain acquisition geometry, and maximize the potential for matching between observed and modelled data.
Input Data Preparation
Multiply this by the Coded Frequency to control cost by limiting the maximum frequency of FWI.
the High Frequency Tuning parameter (Tuning Factor)
We evaluate consistency in amplitude and phase across this QC and exclude outliers to avoid biasing our inversion results.
the QC of Wavelets from Individual Shots
A good strategy is to begin with this type of FWI to mitigate cycle-skipping and provide a good background update.
Adjustive FWI
A key objective of FWI is to minimize this term, also known as J[m].
Data Misfit
A part of the inversion stage, its calculated from the cross-correlation of the backward propagated misfit with the forward propagated source wavefield.
The Gradient
This key property, used in the determination of acoustic impedance, can be set to a constant value or an empirically derived equation known as Gardner's relation in the FWI SFM.
Density
It is recommended to start FWI with this part of the shot record for a better background update.
The Early Arrivals
This combination amplitude and phase based FWI uses 3D pattern matching for a better handling of a complex wavefield.
Enhanced Template Matching FWI
As a rule of thumb, the maximum depth of a reliable FWI update is limited by this calculation.
A third of the maximum offset
Within each frequency band, the user must determine whether the inversion has achieved this goal before progressing to a higher frequency band.
Convergence to a global minimum
To help mitigate the influence of noise or footprint in our model update, we can apply this using a 3D Gaussian filter or an image file.
Gradient Smoothing
The Maximum Velocity Change (and Average Velocity Change) for any given iteration is related to this parameter, documented in the QC Statistics tab of the printout.
Step Length
Adjustive FWI uses a travel-time based objective function, meaning it determines model updates based on travel-time shift, also known as this type of error.
Phase error
Because of forward modeling and gradient calculation, the cost of FWI is proportional to this factor.
The number of shots in an iteration