акустика / gan_ws_acoustical_imaging_techniques_and_applications_for_en
.pdfStatistical Treatment of Acoustical Imaging |
151 |
As ρ changes, the correlation coefficient falls off from its maximum value of N(ξ ,0,0) at ρ = 0 to values which are effectively zero at distances ρ of the order of the correlation distance a (a = ka). Correspondingly, the variable q changes to a quantity of the order of magnitude a 2/2ξ , and since the region of appreciable values of ξ also does not exceed a , the change in q will not be less than a / 2, that is, q ≥ a . In view of this, we have
|
∂2N(r ) |
1 |
N(ξ , 0, 0) ≤ |
N(ξ , 0, 0) |
|
(8.52) |
||||||||||
|
|
|
|
|
|
|
|
|
||||||||
|
∂q2 |
|
q2 |
|
a 2 |
|
|
|||||||||
Therefore, the integral in the right-hand side of equation (8.51) is of order |
1 |
N(ξ , 0, 0). |
||||||||||||||
a |
||||||||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
For the case of large-scale inhomogeneities, a 1 and we can neglect this integral as |
||||||||||||||||
compared with the first term N(ξ , 0, 0). Then we can write |
|
|
|
|
||||||||||||
|
∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
sin q · N |
r |
|
dq = N(ξ , 0, 0) |
|
(8.53) |
||||||||||
0 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
Then the integral I1 of equation (8.46) reduces to the form |
|
|
|
|
||||||||||||
|
|
|
|
|
|
∞ |
|
|
|
|
|
|
|
|
||
|
|
I1 = 2L |
0 |
N (ξ , 0, 0) dξ |
|
|
|
(8.54) |
||||||||
Introducing the variable |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
ν = |
|
ρ2 |
|
|
|
|
|
(8.55) |
||
|
|
|
|
|
|
4L |
|
|
|
|
||||||
the integral I2 of equation (8.47) reduces to the form |
|
|
|
|
|
|||||||||||
|
|
|
|
|
∞ |
|
|
∞ |
|
|
|
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
I2 = −2L |
dξ |
|
siν · N r dν |
|
(8.56) |
||||||||||
|
|
|
0 |
|
|
0 |
|
|
|
|
|
|
Finally we obtain the following results for the mean square amplitude and phase fluctuations:
S¯2 |
μ2L |
∞ dξ N (ξ , 0, 0) |
∞ siν |
|
N(r )dν |
(8.57) |
|
|
= ¯ |
|
|
− |
· |
|
|
|
|
0 |
0 |
|
|
||
B¯2 |
μ2L |
∞ dξ N (ξ , 0, 0) |
∞ siν |
|
N(r )dν |
(8.58) |
|
|
= ¯ |
|
|
+ |
· |
|
|
|
|
0 |
0 |
|
|
Equations (8.57) and (8.58) give the general solutions of the problem of amplitude and phase fluctuation for the case of a large-scale fluctuation (ka 1) under the additional condition that the correlation distance is large compared to the scale of the inhomogeneities (L a). The forms of the solution depend in an essential way on the size of the dimensionless parameter D = 4L/ka2, defined as the rate of the size of the first Fresnal zone to the scale of the
152 |
Acoustical Imaging: Techniques and Applications for Engineers |
inhomogeneities. Equations (8.57) and (8.58) can be studied without specifying the form of the correlation coefficient N(r ) in the following two cases which correspond to limiting values of the wave parameter: (1) D 1 and (2) D 1. This study then reduces to an evaluation of the integral
∞ |
|
|
|
|
I = |
(8.59) |
|||
siν.N r |
dν |
0
in the two limiting cases mentioned.
This is the region of large values of wave parameter. Here the integrand in equation (8.59) is different from zero for value of r that do not exceed the correlation distance a in order of magnitude. In this region of relevant values of r , the argument ν of the integrand sine does not exceed the value ka2/ 4L , that is, it does not exceed the value 1/D
ν |
1 |
(8.60) |
D |
In the case of a large wave parameter, ν will be small. Therefore, siν − π2 , and equation (8.59) takes the form
I = − |
π |
ν |
(8.61) |
2 |
The integral I can be neglected compared to N(ξ ,0,0), then equations (8.56) and (8.57) can be written as
|
∞ |
|
|
S¯2 = B¯2 = μ¯ 2L |
0 |
N(ξ , 0, 0)dξ |
(8.62) |
or, using the dimensional variables L and ξ = ξ /k, |
|
||
|
∞ |
|
|
S¯2 = B¯2 = μ¯ 2k2L |
N(ξ , 0, 0)dξ |
(8.63) |
|
|
0 |
|
|
Thus, in the case of the farfield zone or large wave parameter (D 1) the mean square amplitude and phase fluctuation are the same and increase in proportion to the distance.
|
|
|
|
|
|
|
ξ 2 |
|
|
|
For example, from the correlation coefficient in the form N (ξ ) = e− |
|
, we obtain |
|
|||||||
a2 |
|
|||||||||
|
|
|
|
|
√ |
|
|
|
|
|
S¯2 |
= |
B¯ |
2 |
= |
π |
μ2k2aL |
|
(8.64) |
||
|
|
|
||||||||
|
2 |
|
||||||||
|
|
|
¯ |
|
|
|
8.3.3Correlation of Fluctuations
The mean square amplitude and phase fluctuations still do not give the complete characteristics of the statistical properties of the wavefield. The statistical properties of the fluctuation of the wavefield can be characterized more completely using a correlation function.
Statistical Treatment of Acoustical Imaging |
153 |
Here we will deal with the correlation of the fluctuation of the basic characteristics of the wavefield.
8.3.3.1Correlation of the Amplitude and Phase Fluctuations at the Receiver
We begin by studying the cross-correlation of the amplitude and phase fluctuations at the
receiver. To do this we determine the form of the correlation function BS, using equations (8.30) and (8.31). Multiplying these equations and averaging them, we obtain
|
|
L |
L |
∞ |
|
|
|
|
|
|
|
|
|
|
|
= μ¯ 2 |
|
L − ξ1 |
, ρ1 |
|
L − ξ2 |
, ρ2 |
N r |
dξ1dξ2dη1 |
dη2dζ1dζ2 |
||
BS |
1 |
2 |
|||||||||||
0 |
0 |
−∞ |
|
|
|
|
|
|
|
|
|
(8.65)
Considering only the case of a statistically isotopic medium, we then have that the correlation coefficient N depends only on the magnitude r. Introducing relative and centre of mass coordinates from equations (8.34) and (8.35), we obtain
|
|
|
|
|
|
|
|
|
L |
|
|
L |
|
|
∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
+ y + 2 + z |
2 |
|
|
||||||||||||||||||||||
|
|
|
|
BS = μ¯ 2 |
|
1 L − ξ1, |
|
|
|
|
|
||||||||||||||||||||||||||||||||||
|
|
|
|
|
|
|
|
|
0 |
|
|
|
0 |
|
|
|
−∞ |
|
|
|
|
|
|
|
η |
|
2 |
|
|
|
ζ |
|
|
|
|
|
|||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||
|
|
|
|
|
|
|
× 2 L − ξ2, |
2 |
− y + |
2 − z |
2 |
N r |
dξ1dξ2dηdζ dydz |
|
|
||||||||||||||||||||||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
η |
|
|
2 |
|
|
ζ |
|
|
|
|
|
|
|
|
|
|
|
||||||||||||
Carrying out the integration with respect to the variables y and z, we have |
|
|
|
|
|
||||||||||||||||||||||||||||||||||||||||
∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 L − ξ1, |
|
|
|
|
|
2 + y |
|
+ |
|
|
2 |
+ z |
|
2 |
|
2 L − ξ2, |
2 |
− y + |
|
|
2 − z |
2 |
|
||||||||||||||||||||||
|
|
|
|
|
|
|
|
|
|
|
dydz |
||||||||||||||||||||||||||||||||||
|
|
|
|
|
|
|
|
|
|
η |
|
2 |
|
ζ |
|
|
|
|
|
|
|
|
|
|
|
|
η |
|
2 |
ζ |
|
|
|||||||||||||
−∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||
= |
1 |
( 2 2L − |
ξ1 |
|
+ ξ2 , ρ + 2 (ξ1 |
− ξ2, ρ ) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
(8.66) |
|||||||||||||||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||||||||||||||||||||||||||
2 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||||||||||||||||||||||||
where ρ = |
|
|
|
|
. Then |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||||||
η2 + ζ 2 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||||||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
L |
L |
∞ |
|
2 2L − ξ1 + ξ2 |
|
|
, ρ N r |
dξ1dξ2dηdζ |
|
|
|||||||||||||||||||
|
|
|
|
|
|
BS = 2 μ¯ 2 |
|
|
|
|
|
||||||||||||||||||||||||||||||||||
|
|
|
|
|
|
|
|
|
1 |
|
0 |
0 |
|
−∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
L |
L |
∞ |
2 ξ1 − ξ2, ρ N(r )dξ1dξ2dηdζ |
|
|
|
|
|
||||||||||||||||||||
|
|
|
|
|
|
|
|
|
− 2 μ¯ 2 |
|
|
|
|
|
|||||||||||||||||||||||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
0 |
|
0 −∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
154 |
Acoustical Imaging: Techniques and Applications for Engineers |
For L a, equation (8.66) can be further simplified. Introducing the coordinates 2x = ξ1 + ξ2 and ξ = ξ1 − ξ2 we can then integrate with respect to ξ between the limits –∞ and +∞. Equation (8.66) can be written as
|
|
|
|
L |
∞ |
2 2L − 2x, ρ N r dξ dηdζ |
|
||
BS = 2 μ¯ 2 |
dx |
|
|||||||
|
1 |
|
|
0 |
|
−∞ |
|
|
|
|
|
|
|
|
|
|
|
||
|
− |
2 |
μ¯ 2 |
L |
∞ |
|
2 (ξ , ρ ) N(r )dξ dηdζ |
(8.67) |
|
|
|
dx |
|||||||
|
|
|
1 |
|
|
|
|
|
|
0−∞
Since N(r ) is an even function of ξ , while 2(ξ , ρ) is an odd function of ξ , the second integral vanishes. In the first integral, we can integrate with respect to x. Thus, the problem is reduced
to calculating the integral
L |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
L |
1 |
|
|
|
|
|
|
|
|
|
ρ2 |
|
|
|||||
|
2 |
2L |
− 2x, ρ dx = |
|
|
|
|
|
|
cos |
|
|
dx |
|||||||||||||||||||||
2π (2L |
− |
2x) |
2(2L |
− |
2x) |
|||||||||||||||||||||||||||||
0 |
|
|
ρ2 |
|
|
|
|
|
|
|
|
|
|
0 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||||
The substitution z = |
|
|
|
reduces this to the form |
|
|
|
|
|
|
|
|
|
|
||||||||||||||||||||
2(2L −2x) |
|
|
|
|
|
|
|
|
|
|
|
|||||||||||||||||||||||
|
|
|
|
|
1 |
|
|
∞ cos z |
|
|
|
1 |
|
ci |
ρ2 |
|
|
|
|
|
||||||||||||||
|
|
|
|
|
|
|
|
|
|
2 |
|
|
|
|
dz |
= − |
|
|
|
|
|
|
|
|
||||||||||
|
|
|
|
|
|
|
4π |
|
z |
|
4π |
4L |
|
|
|
|
||||||||||||||||||
|
|
|
|
|
|
|
|
|
|
ρ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
|
|
|
|
|
|
|
|
4L |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ν |
|
|
ν |
|
dν |
|
|
|
|
|
|
|||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ci ν = |
|
|
|
|
|
|
|
|
|
||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
cos ν |
|
|
|
|
|
|
|||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Equation (8.67) then takes the form |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||||||||
|
|
|
|
|
1 |
|
|
|
|
|
+∞ |
|
ci |
ρ2 |
N r |
|
|
|
|
|
|
|||||||||||||
|
|
|
BS |
= − |
|
μ¯ 2 |
|
|
|
|
dξ dηdζ |
(8.68) |
||||||||||||||||||||||
|
|
|
8π |
4L |
||||||||||||||||||||||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
−∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Transforming to polar coordinates (ρ, ) in the (η, ζ ) plane, we obtain
∞∞
|
|
|
1 |
|
μ¯ 2 |
|
dξ |
ci |
|
ρ2 |
· N r ρdρ |
(8.69) |
|
|
BS = − |
||||||||||||
|
2 |
|
4L |
||||||||||
|
|
|
|
|
|
0 |
0 |
|
|
|
|
|
|
Finally, introducing the variable ν, |
|
|
|
|
|
|
|
|
|
||||
|
|
|
|
|
|
|
ν = |
|
ρ2 |
|
|
|
(8.70) |
we obtain |
|
|
|
|
|
4L |
|
|
|
||||
|
|
|
|
∞ |
|
∞ |
|
|
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
|
BS |
= −μ¯ 2L dξ ciν · N r dν |
(8.71) |
00
Equation (8.71) gives the general solution of the problem of the correlation of amplitude and phase fluctuations at the receiver.
Statistical Treatment of Acoustical Imaging |
155 |
If we assume that the correlation coefficient has the form
|
|
|
|
|
|
|
|
2 |
|
η2 |
|
|
2 |
|
|
|
|
|
|
|
|||
N r = exp − |
ξ |
|
+ + ζ |
|
|
|
|
|
|
(8.72) |
|||||||||||||
|
|
|
|
a 2 |
|
|
|
|
|
||||||||||||||
|
|
|
= exp − |
ξ 2 |
|
|
4L ν |
|
|
|
|
|
|
|
|
|
|||||||
|
|
|
|
|
− |
|
|
|
|
|
|
|
|
|
(8.73) |
||||||||
|
|
|
a 2 |
|
a 2 |
|
|
|
|
|
|
||||||||||||
then the problem reduces to calculating the integral |
|
|
|
|
|
|
|
|
|
|
|
|
|
||||||||||
|
|
∞ |
|
a |
2 ν dν |
|
|
|
|
|
|
|
|
|
|
|
|||||||
|
0 |
ciν exp |
|
|
|
|
|
|
|
|
|
|
|
||||||||||
|
|
|
|
|
|
4L |
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
From the tables [17], we found |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
∞ ciν exp |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||
4L ν |
dω |
|
|
|
−a 2 |
|
log |
|
|
1 |
|
4L |
|
2 |
|
||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||||||||
− a 2 |
|
|
|
|
|
|
|
|
A 2 |
|
|
||||||||||||
0 |
|
= |
|
4L |
|
|
|
|
|
+ |
|
|
Finally, we have
√
BS = π μ¯ 2k3a3log(1 + D2 ) (8.74) 16
In probability theory [10], the correlation coefficient is defined as the ratio of the correlation function of the fluctuation to their r.m.s. value. Applied to our case, the correlation coefficient Rbs of the amplitude and phase correlation then has the form
Rbs = |
|
BS |
|
(8.75) |
|||
√ |
|
|
|
|
|||
|
|
S¯2 |
|
||||
B¯2 |
|||||||
Using equation (8.74) for the correlation function |
and equations (8.76) and (8.77) for the mean |
||||||
|
|
|
|
|
values of the amplitude and phase fluctuations, we finally obtain the equations for the region of intermediate value of the wave parameter. Substituting equation (8.73) into equations (8.57) and (8.58), we reduce the problem to evaluating the tabulated integrals [17],
∞ |
|
|
|
|
|
|
|
∞ |
|
ξ 2 |
|
|
|
|
|
√ |
|
|
|
|
|
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
π |
|
|
|
|
|
||||||||||
0 |
N(ξ , 0, 0)dξ = 0 |
|
e− a 2 dξ = |
a |
|
|
|
|
||||||||||||||||||
|
|
|
|
|
|
|
||||||||||||||||||||
2 |
|
|
|
|
||||||||||||||||||||||
∞ |
4L ν |
|
|
|
1 |
|
|
|
|
|
|
|
|
1 |
|
|
||||||||||
|
siν e |
−a 2 |
dν = − |
|
|
arc tan 4L a 2 |
= − |
|
arc tan D |
|
||||||||||||||||
4L a |
|
2 |
D |
|
||||||||||||||||||||||
0 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
where D = 4L a 2 = |
4L |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||
is the wave parameter. Equations (8.57) and (8.58) then become |
||||||||||||||||||||||||||
ka2 |
||||||||||||||||||||||||||
|
|
|
|
S¯2 = |
|
|
|
|
|
1 + D arc tan D |
(8.76) |
|||||||||||||||
|
|
|
|
2 |
μ¯ 2k2aL |
|||||||||||||||||||||
|
|
|
|
|
|
|
|
π |
|
|
|
|
|
|
|
|
|
|
1 |
|
|
|
|
|
||
|
|
|
|
B¯2 = |
|
|
|
|
1 − D arc tan D |
(8.77) |
||||||||||||||||
|
|
|
|
2 |
μ¯ 2k2aL |
|||||||||||||||||||||
|
|
|
|
|
|
|
|
π |
|
|
|
|
|
|
|
|
|
1 |
|
|
|
|
|
156 |
Acoustical Imaging: Techniques and Applications for Engineers |
Equations (8.76) and (8.77) were obtained by Obukhov [9]. The following expression for the correlation coefficient Rbs can be obtained as
Rbs = |
1 |
|
|
log(1 + D2 ) |
(8.78) |
|
|
|
2D2 − (arc tan D)2
At small distances (D < C1), when the ray approach is appropriate, equation (8.78) gives
Rbs 1 3 0.6
22
At large distances (D 1), equation (8.78) takes the form Rbs = logDD , that is, the correlation coefficient falls off with distance and approaches zero. Thus the correlation between the amplitude and phase fluctuations that exist at small distances vanishes at large distances.
8.3.4Quasi-static Condition
We have considered the distribution of inhomogeneities as static, neglecting their change as a result of heat conduction, convection, drift and diffusion. This change can be neglected only if the propagation time t = L/c is small compared to the characteristic time scale of change in the inhomogeneities. However, if this conduction is not met, then in calculating the amplitude (or phase) fluctuation at time t, we have to take into account the refractive index fluctuation at time t = t − cr , where r is the distance from the scattering element to the observation point. That is, a relativistic effect has to be considered and the coordinate transformations here are Galilean transformations: x = x, y = y, z = z, t = t − cr .
In this case the basic formulae become
L |
∞ |
|
|
S L , 0, 0, t = |
|
1 (L − ξ , ρ )μ(ξ , η , η , t )dξ dη dζ |
(8.79) |
0 |
−∞ |
|
|
L |
∞ |
|
|
B L , 0, 0, t = |
|
2 (L − ξ , ρ )μ(ξ , η , ζ , t )dξ dη dζ |
(8.80) |
0 |
−∞ |
|
|
where t = t − cr . Squaring and averaging equations (8.79) and (8.80) we obtain
|
L |
L |
|
∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
S2 = μ2 |
1 L − ξ1, ρ1 |
|
1 |
L − ξ2 |
, ρ2 |
|
|
|
|
|||||||||
|
0 |
0 |
|
−∞ |
|
|
|
|
|
|
|
|
|
|
|
|
||
× |
N(ξ |
ξ |
, η |
η , ζ |
− |
ζ |
, t |
− |
t )dξ dξ dη |
dη |
dζ |
dζ |
(8.81) |
|||||
|
1 − |
2 |
1 − |
2 1 |
|
2 |
|
1 2 |
1 |
|
2 1 2 |
|
||||||
|
L |
L |
|
∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
B2 = μ2 |
2 L − ξ1, ρ1 |
|
2 |
L − ξ2 |
, ρ2 |
|
|
|
|
|||||||||
|
0 |
0 |
|
−∞ |
|
|
|
|
|
|
|
|
|
|
|
|
||
× N ξ1 − ξ2, η1 − η2, ζ1 |
− ζ2, t |
− t dξ1dξ2dη1dη2dζ1dζ2 |
(8.82) |
Statistical Treatment of Acoustical Imaging |
|
|
|
|
157 |
|
where |
|
|
|
|
|
|
N ξ1 − ξ2, η1 − η2, ζ1 − ζ2, t − t = |
1 |
|
|
|
|
|
|
μ |
ξ1, η1, ζ1, t μ |
ξ2, η2, ζ2, t |
|||
|
|
|||||
μ2 |
Here, we assume that the correlation coefficient N depends on the coordinate and time difference, that is, we assume that the process is stationary in time and homogeneous in space.
The time difference is
|
− |
|
= |
|
− c |
− |
|
− c |
= |
c |
|
|||
t |
|
t |
|
t |
|
r1 |
|
t |
|
r2 |
|
(r2 − r1 ) |
|
(8.83) |
|
|
|
|
|
|
|
|
|
The correlation coefficient N is different from zero if r1 − r2 does not exceed the correlation distance a in order of magnitude. This means that the time difference t − t does not exceed the quantity ac in order of magnitude, that is,
a |
|
t − t ≈ c |
(8.84) |
The quantity ac defines the time it takes the wave to go a distance equal to the correlation distance (the scale of the inhomogeneities). If this time is small compared to the correlation time of the refractive index, then the time difference t − t can be set equal to zero and any explicit time dependence in the above equation vanishes. Consequently the quasi-static condition takes the form of the inequality
a/c T |
(8.85) |
The quasi-static condition (equation (8.85)) for average quantities is much weaker than the quasi-static condition for unaveraged quantities, which can clearly be written as
L/C T |
(8.86a) |
In all actual cases, the quasi-static condition (equation (8.85)) for averaged quantities is evidently met with a margin. This justifies the quasi-static assumptions that we have made from the beginning.
8.3.5The Time Autocorrelation of the Amplitude Fluctuations
The change of the inhomogeneities in time produces a change in the frequency of the scattered waves and broadens the frequency bandwidth of the incident radiation. The nature of this broadening can be inferred from the form of the time autocorrelation function of the amplitude and phase fluctuations.
An appreciable weakening of the autocorrelation of the amplitude fluctuations can be expected after a time interval τ commensurate with the correlation time T of the refractive index. If the quasi-static condition is met, then the time interval τ is also large compared to the time ac ; that is,
τ |
a |
(8.86b) |
|
c
158 Acoustical Imaging: Techniques and Applications for Engineers
Writing equation (8.86b) for the time t1 and t2, assuming that they are separated by the interval
τ :
B L , 0, 0, t1 |
L |
∞ |
|
|
L − ξ , ρ μ ξ , η , ζ , t dξ dη dζ |
|
|||
= |
|
2 |
(8.87) |
||||||
|
0 |
−∞ |
|
|
|
|
|
|
|
B L , 0, 0, t2 |
L |
∞ |
|
|
L − ξ , ρ μ ξ , η , ζ , t dξ dη dζ |
|
|||
= |
|
2 |
(8.88) |
||||||
|
0 |
−∞ |
|
|
|
|
|
|
|
where |
|
|
|
|
|
|
|
|
|
|
t |
= t1 − |
r |
and t = t2 − |
r |
(8.89) |
|||
|
|
|
|
||||||
|
c |
c |
Denoting the time autocorrelation fluctuation of the amplitude by F(τ ), we have, by definition
F (τ ) = B (L , 0, 0, t1 ) B(L , 0, 0, t2 )
Multiplying with equations (8.87) and (8.88) and averaging, we obtain
|
L |
L |
|
|
∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
F (τ ) = μ¯ 2 |
2 (L − ξ1, ρ1 ) 2 (L − ξ2, ρ2 ) |
|
|
|
|
|||||||||||||
|
0 |
0 |
|
|
−∞ |
|
|
|
|
|
|
|
|
|
|
|
|
|
× |
N(ξ |
ξ |
|
, η |
η , ζ |
− |
ζ |
|
, t |
− |
t )dξ |
dξ dη |
dη |
dζ |
dζ |
|
(8.90) |
|
|
1 − |
|
2 |
1 − |
2 1 |
|
2 |
|
|
1 2 1 |
2 1 2 |
|
From (8.89) we have
t
since
|
|
t |
|
t |
|
|
r1 |
− |
t |
|
r2 |
|
t |
t |
|
|
(r2 − r1 ) |
|
τ |
|
(r2 − r1 ) |
|
τ (8.91) |
|
− |
|
|
− c |
− c |
|
+ |
c |
= |
+ |
c |
|
|||||||||||||
|
|
= 1 |
2 |
= |
1 − |
2 |
|
|
||||||||||||||||
|
(r2−r1 ) |
a |
, therefore equation (8.90) becomes |
|
|
|
|
|
|
|||||||||||||||
|
c |
|
|
c |
|
|
|
|
|
|
L |
L |
∞ |
|
F (τ ) = μ¯ 2 |
2 (L − ξ1, ρ1 ) 2 (L − ξ2, ρ2 ) × N(r , τ )dξ1dξ2dη1dη2dζ1dζ2 |
||
0 |
0 |
−∞ |
(8.92) |
|
|
|
Mintzer [11] started with the assumption that the correlation coefficient N(r ,t) separates into two factors, one of which depends only on the coordinates and the other only on the time – that is, the correlation coefficient can be represented in the form
|
|
|
|
|
N(r )M(τ ) |
(8.93) |
||
then (8.92) becomes |
|
|
|
|
|
|
|
|
|
L |
L |
|
∞ |
|
|
|
|
F (τ ) = M(τ )μ¯ 2 |
2 (L − ξ1, ρ1 ) 2 (L − ξ2, ρ2 ) |
|
||||||
|
0 |
|
0 |
|
−∞ |
|
|
|
× |
N(r )dξ |
dξ dη |
dη dζ |
dζ |
(8.94) |
|||
|
1 |
2 |
1 |
2 |
1 |
2 |
|
Statistical Treatment of Acoustical Imaging |
|
|
159 |
Denoting the amplitude autocorrelation coefficient by R(τ ), we have |
|
||
R (τ ) = |
F (τ ) |
(8.95) |
|
|
|
||
B¯2 |
|||
by definition. Then, by (8.94) and (8.33), we obtain |
|
||
R (τ ) = M(τ ) |
(8.96) |
This simple result shows that the time autocorrelation component R(τ ) of the amplitude coincides with M(τ ), the time autocorrelation coefficient of the refractive index.
It is possible that Mintzer’s assumption [11], that the time and space coordinates can be separated, is justified in the absence of any motion of the inhomogeneities which are produced by drift and convection. In this case, a change of the inhomogeneities in time might be produced by, for example, heat conduction, diffusion and turbulence . However, such a separation cannot be valid when drift or convection is present. In this regard, it is of interest to examine the case when a change in the inhomogeneities in time is caused by their motion.
Assume that all the inhomogeneities move with the same velocity v, as a result of the drift, and that a change in the inhomogeneities results exclusively from the drift, while other factors (heat conduction, diffusion, turbulence) play no important role, that is, changes produced by these factors proceed much more slowly. Then in the coordinate system moving with the flow, the correlation coefficient depends only on the coordinates, that is, it has the form
|
|
N(x1 − x2, y1 − y2, z1 − z2 ) |
|
Using the Galilean coordinates transformation formulae, |
|
||
x1 |
− x2 |
= ξ1 − ξ2 − vxτ |
(8.97) |
y1 |
− y2 |
= η1 − η2 z1 − z2 = ζ1 − ζ2 − vzτ |
(8.98) |
We obtain N (ξ1 − ξ2 − vxτ , η1 − η2, ζ1 − ζ2 − vzτ ) to the coordinate system ξ , η, ζ |
fixed |
||
with respect to the receiver. |
|
|
The coordinate systems are orientated in such a way that the flow velocity v lies in the plane x0z and ξ 0ζ . It can be seen from (8.98) that in the case under consideration there is no separation of the space and time coordinates.
It can also be seen that, in the case of a homogeneous flow, the problem of time correlation reduces to the problem of space correlation which has already been solved. To see this, it is sufficient to use the principle of relativity to go over to the coordinate system x, y, z moving with the flow. The receiver moves with a velocity −v with respect to this system. We assume that, at time t, the receiver is located at point A and, at time t + τ , it is located at point E where
|
|
= vτ . We denote the amplitude |
the receiver displacement of AE = l satisfies the condition l |
fluctuation at point A at time t by B(A, t) and the amplitude fluctuation at point E at time t + τ by B(E, t + τ ). Since the distribution of inhomogeneities can be considered static in the coordinate system moving the flow, then, inasmuch as we are dealing with a time interval of order τ , the amplitude at any point does not change with time, that is, B (E, t + τ ) = B (E, t ).
Therefore, |
|
B (A, t ) B (E, t + τ ) = B (A, t ) B(E, t ) |
(8.99) |
160 |
Acoustical Imaging: Techniques and Applications for Engineers |
where the left-hand side represents the time correlation function F(τ ) in the coordinate system fixed with respect to the observer, while the right-hand side represents the space correlation
function F1 ( ) l .
We then obtain
|
(8.100) |
F (τ ) = F1 (l) |
8.3.6Experimental Verification
The above consideration of moving inhomogeneities in acosutical imaging is the first application of relativity to acoustical imaging. In this case is the Galilean relativity. Experimental verification of the application of either the fluctuation theory or the statistical theory to acoustical imaging is as follows.
So far, to the author’s knowledge, there have been no experiments on the application of fluctuation theory or statistical theory to the acoustical imaging of solids. Experiments have been performed only for sound propagation in the atmosphere and in the ocean.
The final experimental studies of amplitude and phase fluctuations for sound propagation in the atmosphere are due to Krasilnikov and Ivanov-Shyts [12, 13]. They compared their experimental data with the theoretical results of Krasilnikov [14] using the ray approach. They discovered that the theoretical result concerning the phase fluctuation was in satisfactory agreement with experiment, while the theoretical formula for the amplitude fluctuations was not supported by the experimental data. Figure 8.1 shows a logarithmic plot of the dependence of the mean square phase fluctuations (over various time intervals) on the distance between the
|
|
|
|
|
|
1 |
|
transmitter and the receiver. The line L 2 is drawn through the experimental point corresponding |
|||||||
to the distance L |
= |
22 m. It is seen from the figure that the theoretical law, according to which |
|||||
|
|
1 |
|
|
|
|
|
|
|
|
|
|
|
||
|
S¯2 ≈ L 2 , agrees satisfactorily with the experimental data. The dependence of the mean |
||||||
square amplitude fluctuation √ |
B2 |
on the distance L is shown in Figure 8.2. The location of |
|||||
|
|
|
|
¯ |
1 |
3 |
|
the experimental points with respect to the curves L 2 |
, L, L 2 gives an idea of the behaviour of |
√ ¯2
B as the distance changes. As can be seen from the figure, the amplitude fluctuations, like
log √S2
1.3
~L½
1.2
1.1
1.0
0.9
22 |
45 |
67 m |
|
log L |
|
¯2 = =
Figure 8.1 Dependence of S on L for the frequency ν 3 kcps (logarithmic scales): , t 0.04 s; 0, t = 0.08 s; , t = 0.2 s (Krasilnikov and Ivanov-Shyts [12])