Sei sulla pagina 1di 2

c

Multiple from the ocean bottom and water free surface is major error source in marine data
acquisition.

To remove the multiple from the data is challenging problem in marine seismic data processing.
These are removed on the basic of properties of the multiple

1. Multiple have periodicity in amplitude (Predictive)


2. Primary reflection and multiples have different normal move out (CMP gathering and
staking and F-K filtering).
3. Predictability as the auto-convolution of the primaries (Surface Related Multiples).

O 
 


This method uses the periodicity of the multiples to remove from


multiples from the data. Periodicity of the multiple is exploited to design an operator that
identifies and removes the predictable part of the wavelet (multiple), leaving only non-
predictable part (signal). The key assumption is that genuine reflections come from an earth
reflectivity series that can be considered random and therefore not predictable. This method
can be used for attenuation of short period multiples (most notably reverberation form the
relatively flat shallow water bottom). The predictive Deconvolution is a linear square inverse
filter has prediction distance greater than unity. The predictive technique allows one to control
the length of the desired output wavelet, and hence to specify the desired degree of
resolution. This is minimum phase filter designed by the autocorrelation of the signal.
Removing the predictable part of the signal leave so called prediction error which is taken to
be the meaningful signal in reflection seismology.

    

 


     
 
 
 
 

Predictive convolution is used for the signal when we know the signals (input & output)
spectral characteristics or auto-correlation and cross-correlation of the input and output signal.

In seismic processing generally Butterworth filter (band-pass) is used after predictive


Deconvolution because predictive Deconvolution increases the frequency content of the
signal. The signal vibroseis frequency sweep form (30 to 300 Hz).

      
 Primaries and multiples exhibit hyperbolic moveout in CMP but
their curvature is different. After NMO correction with velocity of the primaries ideally primaries
exhibit flat moveout. Whereas the residual move out remain is multiples can be approximated
as hyperbola or hyperbolas. This different in moveout can be exploited to separate the
primaries from the multiples either in frequency or in K-tau (Radon) domain.

The performance in frequency domain (f-filter) in suppressing multiples strongly depends on


primary and multiples are mapped to separate region of the frequency plane. This is general in
the case on far-offset traces for which the difference in moveout can be large, but not for short
off-set for which the different in moveout is small. Therefore moveout method is poor for small
offset even if underline geology is not complex. In such case of short offset radon
transformation can be used. For complex geology like salt dome the hyperbolic or infact any
NMO approximation break down than in that case after NMO primary are unlikely to be flat.
Furthermore, the residual moveout of multiples are unlikely to be approximated by hyperbola
or parabola. So in complex area radon transformation for the multiple attenuation is not good
option.

ï !   " #



(ï!")

 This method uses the recorded seismic data


to predict and iteratively subtract the multiples series. The key advantage of SRME is that it
needs no subsurface information whatsover. The multiples are completely predicted from the
data. 2D SRME can deal with all kinds of surface related 2D multiples, provided all relevant
data are recorded within the aperture and offset limitation of the survey line. In predicting 3D
multiple with 2D SRME is hazardous because accuracy of the prediction depends upon on the
amount of crossline dip. 

Potrebbero piacerti anche