Analysis Of Predictive Deconvolution And The Related Issues

Predictive deconvolution is the most essential and commonly used method of deconvolution. It calculates the part of pure noise according to the first order reflection and noise of the seismic record, and then uses this seismic record which includes the first order reflection and noise to subtract the part of pure noise. If the noise that predicted were multiple waves (especially the whole course multiple waves), it is useful to decay the energy of multiple wave by using this method of prediction filter. However, the multiple waves are not only one in the actual seismic section or a seismic trace gathers, usually there will be many groups of multiple waves which are caused by different interface. The groups of multiple waves generally have different apparent period of itself, it is also possible to have the same apparent period of different multiple wave. At present, in the application of the predictive deconvolution, we usually suppose the predictive distance is a constant, and just use one predictive distance , so we can only subtract the multiple wave of one period, or the subtraction effect is not obvious. As a result, we can not achieve the expect purpose to subtract the multiple wave. In order to decay the many groups of multiple waves, it needs to run the program of predictive deconvolution many times. In this paper, we put forward the multiple-prediction-step deconvolution, which is, taking many prediction-steps, subtracting the multiple waves step by step which points at the multiple wave of many periods in the seismic record. Coping with the remaining multiple wave, we reproduce correlation analysis and define the new parameters of deconvolution, repeated in this way, until the many groups of multiple wave is subtracted all.By model data test and subtracting multiple waves of actual near-surface reflection seismic data, we prove that the method of multiple-prediction-step deconvolution is practicable and functional; besides, this method also achieves better results in subtracting multiple waves of VSP data.