前馈神经网络在水泥胶结识别中的应用

Application of Feedforward Neural Network in Cement Bond Identification

为了解决石油测井中水泥胶结质量识别误差较大的问题,采用八扇区水泥胶结测井仪进行声幅测量。仪器灵敏度变化以及泥浆对声信号的衰减所引起的误差可以综合利用首波幅度信息对其消除。通过对非线性连接权的神经网络方法的研究和阐述,克服了传统的BP学习算法过程中难以跳出局部极小值与收敛速度慢的缺点,使其具有3层BP网络的功能且提高了运行速度,优于统计识别方法。实验表明,前馈神经网络方法的应用可识别水泥胶结质量,识别正确率远高于相对幅度法,效果显著。

In order to solve the problem of big unavoidable error in cement bond logof oil casing-well engineering, the eight segmented cement bond tool is adopted to measure sonic amplitude. Comprehensive utilization of the first wave of amplitude information eliminates the inevitable errors caused by the mud on the attenuation of the acoustic signal, as well as changes in instrument sensitivity. The method of artificial neural network （ANN） with nonlinear connected weights superior to that of statistics theory is studied, which can replace three-layer error back-propagation （BP） algorithm, so the implied-layer removed, the calculating simplified, and the operated speed in- creased. Actual application example shows that the method of ANN can identify cement quality, the identification accuracy rate is much better than that of amplitude-compare method,and the application effect is very notable.