The Application Of Fuzzy Neural Network In 2D Seismic Data Interpretation

seismic data for the large difference in quality,comprehensive interpretation difficult issues to Sulige gas field,Ordos Basin, the southern region, for example, the fuzzy neural network system to carry out seismic multi-attribute seismic facies analysis and prediction of the application.Fuzzy neural network combines the ability of knowledge expression of fuzzy reasoning and learning ability of neural networks,which can be better used in the qualitative and quantitative variety of information,through the fuzzy rules,fuzzy data, neural network learning process and anti-fuzzy,etc, to achieve the qualitative model identification and parameter quantitative prediction. The method is strongly integrated and more flexible operational,more suitable for identification of seismic facies model and quantitative reservoir prediction.According to the relationship between over-well seismic line and well-point sandstone,there is certain qualitative relationship between the distribution of sandstone and sedimentary facies in research areas and the seismic waveform. The latter can provide qualitative instruction of sedimentary characteristics.What's more, there lies the weak linear correlation among the thickness of sandstone,root mean square amplitude,attenuation coefficient and the average absolute amplitude. Meanwhile,Seismic attributes can be used for the quantitative prediction of sandstone thickness. The existence of these relations is the basis of the feasibility of integrated seismic interpretation. But there are two aspects of the problem: Firstly,according to the differences in the quality of seismic lines,adopt a system of targeted technical methods for interpretation; secondly, more flexible techniques should be required to combine the information such as earthquake,drilling and geological understanding in the process of integrated seismic interpretation.Through the configuration of objective function of fuzzy neural network clustering algorithm,by combining 2D seismic waveform information, geostatistical characteristics of seismic data,well-point information of sandstone facies and sandstone thickness information,to achieve a constrained clustering for well-point information on the seismic waveform, to finally realize a supervised fuzzy clustering neural network. The method has good geological implications in seismic facies.Based on the differences in quality of seismic data,a method,which is hierarchical clustering and has intersection controlling of Wells and seismic joint, can be used to realize the effective use of seismic data of different quality and the expression of uncertainty characterization of clustering results.On the basis of correlation analysis between seismic attributes and well-point in the thickness of sandstone,through the selection of the four sensitive seismic attributes, RMS amplitude, the average absolute amplitude, attenuation coefficient and wave impedance sandstone thickness,under seismic facies constraint,to get quantitative prediction of sandstone thickness.The studies show that the fuzzy neural network technology can be better used in 2D Seismic Interpretation of seismic facies analysis and quantitative prediction of sandstone thickness. Meanwhile,by combining geological statistics,probability characterization,and flexible strategy of the algorithm,can make the technology more strongly and widely used in application.