关键词: 井壁稳定性, 安全泥浆密度窗口, 地震记录, 钻前预测, 神经网络

Abstract: Seismic records and sonic log data can all reflect acoustic property of formation.According to seismic wave dynamic mechanics,the reasonable nonlinear mapping relationship between seismic records and log data was established.The model of log data was constructed by using seismic records,on the bases of BP artificial neural network theory.The input layer of BP network is composed of autoregressive coefficient,fractal dimension,maximum Lyapunov index and mutation parameter.The acoustic wave velocity and density of the formation being drilled can be acquired with the log data model,and the elastic property and strength can also be estimated.Based on the nonlinear function curve simulation method,the safe mud weight range model for predicting borehole stability was established and successfully verified in YT101 Well.The prediction model was adapted to predict borehole stability of pre-drilling formation by using the log data of drilled part for the first wild cat well.

Key words: borehole stability, mud weight range, seismic record, drilling prediction, neural network