论文部分内容阅读
通过收集到的长春市及周边地区各类钻孔资料,运用软件GMS建立长春及周边地区的三维地层结构可视化模型,与实际地质(地形)情况较为吻合,清晰地反映出长春地区地层结构情况,通过软件还可观察地层任意位置的剖面情况。将神经网络引入其中,当输入钻孔坐标(x,y,z)、地层厚度及地层深度时,能够较为准确地预测出对应地层的地质时代和岩性,采用结构为5-13-5的BP神经网络(单隐含层)预测结果的平均相对误差为11.12%,最小误差为7.50%、最大误差为15.71%;采用改进后的结构为5-11-7-5的BP神经网络(双隐含层),预测结果的平均相对误差为4.64%,最小误差为3.63%、最大误差为6.59%,完全满足预测精度要求。
Through the collected data of all kinds of boreholes in Changchun city and the surrounding area, using the software GMS to establish the visual model of 3D stratum structure in Changchun and the surrounding area is in good agreement with the actual geology (topography), which clearly shows the stratigraphic structure of Changchun area, Through the software can also be observed at any position of the formation profile. When the neural network is introduced, the geologic age and lithology of the corresponding strata can be predicted more accurately when the borehole coordinates (x, y, z), formation thickness and depth of formation are input. The structure of 5-13-5 The average relative error of BP neural network (single hidden layer) prediction results was 11.12%, the minimum error was 7.50% and the maximum error was 15.71%. The BP neural network with improved structure of 5-11-7-5 The hidden layer), the average relative error of the prediction results is 4.64%, the minimum error is 3.63% and the maximum error is 6.59%, which fully satisfies the prediction accuracy requirements.