The Methods Research Of Lithology Identification Based On Logging Data

With the development of the logging technology, the comprehensive application of allsorts of logging data has became the main way for getting information of lithology rapidlyand accurately.Lithology identification is the most key ring of logging datainterpretation.The traditional methods have slow efficiency,low accuracy, and humanfactors so that not conducive to the actual engineering application.Therefore, constructing aspeed, high recognition rate and strong generalization ability lithology identification methodhas been the pressing needs in interpretation field of logging datas.Around the purpose of lithology identification, this paper mainly completes three areas:logging data pretreatment, log facies analysis, the methods research of lithologyidentification. logging data pretreatment is a basic work before completing of logging datainterpretation,and is for eliminating the influences of unformation factors. Log faciesanalysis based on logging curves layering of the multiple information fusion is a necessaryprocess for constructing log facies-lithology database in the whole regional. The methodsresearch of lithology identification complete algorithm construction and application contrastof three neural networks,which contain GA-BP neural network,FCM_PNN neural network,and genetic optimized RBPNN neural network. The results show: the applications of threemethods in lithology identification area are feasible, but there exit slow speed to GA-BP andrelatively low recognition accuracy to FCM_PNN.Thus they can not achieve the standard ofpractical engineering application.So structuring genetic optimized RBPNN networkmodel.It is searched for the optimized hidden-center vectors and matching kernel-functioncontrol parameters of the RBPNN by using the genetic algorithm,which must be satisfiedminimum error of RBPNN training. Relative to the BP network and FCM_PNNnetwork,This method can not noly guarantee convergence speed and the accuracy,but alsocan identify lithology informations more quickly and more accuratly.And it has simplenetwork structure,strong promotion and generalization ability.That make it can be uesed in logging data interpretation field.The methods research of lithology identification based on logging data providesscientific method support and theory dependence for logging data interpretation, and hasimportant practical significance to the areas of geological exploration.