The XWITM optimisation toolbox powering Full Waveform Inversion now integrates the AWI and RWI 2.0 cost functions. This introduces increased sensitivity to the deepest and most subtle reflection moveout invisible to standard FWI formulations.
AWI redefines the cost function with filter coefficients penalised away from zero lag. RWI redefines the cost function with a virtual-source generated scattered wavefield. AWI avoids cycle-skipped solutions. RWI increases the penetration depth of macro-model updates.
In this demonstration (SEG14 blind synthetic) we generate a superior fit between predicted and observed data by using AWI and RWI to update the macro model below the diving wave zone.
Original Starting Model
Original starting velocity model
Simplified Starting Model
Change relative to original start
Simplified starting velocity model
X-Wave
No AWI. No RWI.
Cumulative update after 1st pass
Velocity model after 1st pass
Cumulative update after 2nd pass
Velocity model after 2nd pass
Cumulative update after 3rd pass
Velocity model after 3rd pass
X-Wave + AWI
Deeper penetrating update
Cumulative update after 1st pass
Velocity model after 1st pass
Cumulative update after 2nd pass
Velocity model after 2nd pass
Cumulative update after 3rd pass
Velocity model after 3rd pass
X-Wave + AWI + RWI 2.0
Deepest penetrating update
Cumulative update after 1st pass
Velocity model after 1st pass
Cumulative update after 2nd pass
Velocity model after 2nd pass
Cumulative update after 3rd pass
Velocity model after 3rd pass
Turning the remaining residual into macromodel updates
Enhanced convergence from fitting moveout curvature - deep reflected events in the predicted data shifting into position through the inversion.
Interleaved predicted (start) and field data
Interleaved predicted (iteration 40) and field data
Observed field data