土工基础

›› 2014, Vol. 28 ›› Issue (6): 64-66.

• 工程实录 • 上一篇    下一篇

基于改进BP神经网络之井筒结构优化研究

雷 刚   

  1. (中油辽河工程有限公司,辽宁盘锦 124010)
  • 收稿日期:2013-09-06 出版日期:2014-12-25 发布日期:2015-01-12
  • 作者简介:雷 刚(1967-)，男，高级工程师，研究方向为岩土、结构专业设计。
  • 基金资助:

    国家自然科学基金资助项目(51109035)，中央高校基本科研业务费专项资金资助项目（N110401006;N110301006）

Vertical Shaft Dimension Optimization Using Modified BP Neural Network

LEI Gang   

  1. (China Liaohe Petroleum Engineering Co., Ltd., Panjin 124010)
  • Received:2013-09-06 Online:2014-12-25 Published:2015-01-12

摘要： 在立井施工过程中，选择合适的井筒断面及尺寸对立井快速高质量的施工有重要影响。现通过数值模拟软件FLAC3D计算出井筒不同断面及尺寸的塑性区总体积、最大应力、最大位移值三个力学参数，运用改进BP神经网络建立起井筒断面及尺寸与FLAC3D计算出的力学参数之间复杂的非线性关系。最后将实际施工中能满足设计需要的力学参数输入训练好的BP神经网络，即可得到反演出的井筒断面及尺寸，为井筒结构设计作参考。

关键词: BP神经网络, 井筒断面, 结构优化, FLAC3D

Abstract: The cross section dimensions have a significant impact on the fast construction and quality of the vertical shaft. The different cross sectional sizes are obtained by the numerical analysis using FLAC3D software and their associated total volume of the plastic zone, maximum stress and maximum deflection. The nonlinear relationships between the different shaft dimensions and the mechanical parameters obtained by FLAC3D are established through the modified Back Propagation (BP) neural network. The mechanical parameters that meet the design requirements are input into the trained BP neural network and the shaft cross section dimensions can be obtained throughout the back analysis.

Key words: BP Neural Network, Vertical Shaft Cross Section, Structural Optimization, FLAC3D