关键词: 地层破裂压力, BP神经网络, 多元回归分析, 统计模式

Abstract:

At present, a lot of the theoretical  models or equations of calculating formation fracture  pressure are presented and used, but there are few statistical models that can obviously reflect the  formation fracture  pressure and the logging rock mechanics parameter variety. In this paper, based the relation of formation fracture  pressure and rock  mechanics parameter, the four parameters which have a close connection to formation fracture  pressure and are regarded as independent variable (input  parameter), such as Young's modulus, bulk modulus, poison's ratio and depth etc. are selected. And by the BP neural network and multi-regress analysis technique,  some reasonable statistical models of formation fracture pressure are established by making full use of formation fracture  pressure data of carbonate formation,  and the prediction study is developed. All research shows, the form of multi-regress analysis is simple and visual, and easy to use, but the error is not lesser.  The BP neural network model is complex, and it's model is difficult to set up, but the estimate's formation fracture the pressure error is little, the precision  is high. The two kinds of statistical model are still practical and feasible approaches to calculate formation fracture pressure.

Key words: Formation fracture pressure, BP neural network, Multi-regress analysis, Statistical model.