Study On Repeated Weak Signal Processing Methods In Active Source Experiment

Extracting the effective but weak signals from repeated seismic records with low signal-noise-ratio is the first step to explore the subsurface deep structure using active source experiment,which would directly affect the accuracy further data processing and interpretation afterwards.In this paper,we explored the effects and evaluated the efficiency of the weak signal extraction by stacking methods,including linear stacking,semblance weighted stacking,time-domain phase weighted stacking,time-frequency domain phase weighted stacking and proposed time-frequency domain phase weighted stacking,as well as direct separation methods,such as singular value decomposition and blind source separation,and further applied these methods to effective signal extraction in the real active source experiment.First,we constructed the synthetic traces using Klauder wavelet and air-gun source wavelet respectively to explore the efficiency of different methods on extracting effective signals.We also evaluated the results by quantitative measurements of travel time,amplitude,frequency content and so on.Then,we synthetized seismic profile data with a 1-D velocity model and real noise level,in order to find the suitable processing flow and extraction methods for the two seismic source.At last,we processed real data in active source experiment to further test the reliability and the efficiency.The processing on synthetics and real data obtained the following conclusions.(1)When the numbers of shots with active-source are small,the signals extracted by non-linear stacking method are better than that extracted by linear stacking method;while if the shot numbers are large enough,the linear stacking method on signal extractions are better,and it also has minor influences on the final effective signals.(2)Zero-phase processing reduce the effects on travel time and waveforms during the wavelets extraction.(3)During the vibroseis source signal processing process,cross-correlations would compress the wavelets and avoid the unwanted stacks between different phases after pre-processing of the raw waveforms,then the effective signals can be further extracted.(4)Direct stacking of the pre-processed signals from air-gun source effects accuracy of the arrive time information.Accordingly,we proposed that the zero-phase processing after pre-processing of the raw data reduce the effects on travel time and waveform information.Combinations of the above two process-flows provide valuable reference for travel-time picking.