| The geophysics well logging is an important means of exploring oil field and perceiving geology visage.Logging curve features pattern recognition technology is a new aspect of logging explanation field.This paper uses various mathematics and computer technologhy such as cluster analysis,wavelet transformation and neural networks pattern recognition to form a complete logging curve pattern recognition system which had an effect on stratigraphic correlation.The system mainly includes three parts:1.Use cluster analysis to confirm shale line and design proportion function which is used to make computer's automatic zonation logging curve come ture. 2.De-noising signal of logging curve by wavelet transforms.After wavelet decomposition on logging curve,eliminate some noise and form smooth curve by log curve reconstruction in order to prepare recognition for curve state.3.Separate logging curve into bell-shaped,box type,funnel shaped,dentate and fiat curve.Use BP neural networks technology to recognize pattern and to judge the conformation of logging curve.This paper uses the system to analyse stratigraphic correlation on neighborhood well.Separate logging data of neighborhood well,make it smooth,use PNN probability neural networks to judge if the two parts of curve are the same stratum,record top and bottom depth and compare trend to stratum transformation. Finally it comes to some conclusion on the research and practice about the system. |
