The Study Of Pattern Recognition In The Deposition Facies Of Petroleum

Depositional facies, a term of petroleum geology, is a combination of sedimentary rocks produced by a certain depositional environment. The research of the pattern recognition of depositional facies is still a beginner. In practice, the manual recognition is the main method.On the base of experiences of experts, in this dissertation, we construct an integrated auto system to recognize the pattern of depositional facies combining technologies of the computer, the pattern recognition and the geology. With the system accomplished, the efficiency and the velocity of the pattern recognition will be increased and the various results produced by the system will be lesser than that produced by manual work.In this dissertation, we set up our task from two hands. The one is scientific research, the other is project exploitation. Making use of the well-logs, we describe in detail the data preprocess, the extraction and choice of features, the design of classification, the design and realization of system, etc. The linear regression referred in the data preprocess strengthen the tide movement. In the inversion of well-log and the extraction of features, the Walsh function and Sequence Analysis is introduced, which raises the vertical resolution, protrudes the limit of layers and extracts more effective features. When we design the classification, we combine the tree classification and the Support Vector Machines in order to improve the ability of combining experiences and performance of generalization. The technology of Object-Oriented offer us a power tool to design and to exploit the project of our system. With all of above, we realize a pattern recognition system of depositional facies which can work in the real task.Not only the petroleum geology principle and the concrete handle details of oil field work can not be left in the establishment of whole system , but also the basis of machine cognition that is computer pattern recognition technology.