We describe a new method of automatic normal move-out (NMO) correction
and velocity analysis that combines a feedforward neural network (FNN
) with a simulated annealing technique known as very fast simulated an
nealing (VFSA). The task of the FNN is to map common midpoint (CMP) ga
thers at control locations along a 2-D seismic Line into seismic veloc
ities within predefined velocity search limits. The network is trained
while the velocity analysis is performed at the selected control loca
tions. The method minimizes a cost function defined in terms of the NM
O-corrected data. Network weights are updated at each iteration of the
optimization process using VFSA. Once the control CMP gathers have be
n properly NMO corrected, the derived weights are used to interpolate
results at the intermediate CMP locations. Ln practical situations in
which lateral velocity variations are expected, the method is applied
in spatial data windows, each window being defined by a separate FNN.
The method is illustrated with synthetic data and a real marine data s
et from the Carolina Trough area.