Escolar Documentos
Profissional Documentos
Cultura Documentos
(1a)
(1b)
in which
velocity of wave qP; and and are the anisotropic parameters defined by Thomsen (1986).
The discretization of the derivatives of equations (1) and (2) is usually performed with the scheme of
finite differences. A second-order discretization is performed in the time derivatives as follows:
(2a)
(2b)
)]
(3a)
)]
(3b)
is the
may be written
in which
is the sampling interval in axis , N is the discretization order and
the binomial window:
(
(4)
are the coefficients defined by
(5)
( )
A way to reduce the numerical dispersion is to adopt window functions for generating optimized
coefficients for Equation 1. ALFORD et al. (1974) have proposed a window
in order to obtain optimized FD
coefficients, which may be generated as follows:
in which
(6)
(7)
(
Using a final form for the optimized FD operator, it is possible to combine window function
parts of Equation 6 as follows:
in which
with all
(8)
is the final form of the coefficients to be used in implementation and may be defined as:
(9a)
(9b)
(9c)
It is difficult to determine parameter M of such window function and at the same time such parameter
needs to be handled with caution, since it may materially affect the final result. ZHANG & YAO (2013) have
determined coefficients
by using an optimization scheme that uses the maximization of the wave number
convergence, considering an error limitation. Analysis was made concerning the peak of error between the
optimized operator in the wave number domain and in the analytical wave number domain. In order to generate
the optimized coefficients, the Simulated Annealing (SA) minimization algorithm was used in this paper.
Numerical Examples
We have used stencils 2a, 2b, 3a and 3b to solve Equations 1a and 1b. FD coefficients are determined by
Equations 9a, 9b and 9c. The first example is a modeling in a homogeneous 2d model with elliptical anisotropy.
Figure 1 shows the modeling snapshot using the conventional method (Figure 1a) and the optimized method
(Figure 1b). It is possible to see that dispersive property was more developed in the optimized modeling than in
the conventional one. Figure 2 shows the time registration of seismic traces based on offset. In such event, the
level of dispersion was lower in the optimized modeling than in the conventional one.
Conclusions
The stencil of finite differences was used for estimating the spatial derivatives of Equations 1a and 1b.
Taylor coefficients generated by the binomial window were inserted in the discretization of VTI equation
proposed by Du et al. (2008) and compared with the ones arising from the optimized window function. The
results show that the optimized coefficients were more accurate. Such accuracy was achieved with the same
computational cost of the conventional method.
Figure 1: Modeling snapshots at 0.15s by the (a) conventional and (b) optimized methods for a 2D
homogeneous, acoustic elliptically anisotropic model;
,
,
,
,
,
. Grid dimensions are
.
Figure 2: Modeling records of Figure 1 by the (a) conventional and (b) optimized methods. Receivers have the
same depth as the source.
References
ALFORD, R. M., KELLY, K. R., BOORE, D. M., 1974. Accuracy of finite difference modeling of the acoustic
wave equation: Geophysics, v. 39, no. 6, pages 834842.
CHU, C., STOFFA, P. L., 2012. Determination of finite-difference weights using scaled binomial windows:
Geophysics, v. 77, pages 57-67.
DU, X., FLETCHER, R.P., and FOWLER, P. J., 2008. A new pseudo-acoustic wave equation for VTI media.
70th EAGE Conference & Exhibition - Rome, Italy.
THOMSEN, L., 1986. Weak elastic anisotropic: Geophysics, v. 51, pages 1954-1966.
ZHANG, J. H., YAO, Z. X., 2013. Optimized finite-difference operator for broad-band seismic wave modeling:
Geophysics, v.78, pages 1-6.