基于IA-BP智能算法的初始地应力场反演研究

孙港; 王军祥; 郭连军; 寇海军; 徐景龙

doi:10.11835/j.issn.2096-6717.2021.116

您当前的位置：

首页 >

文章列表页 >

基于IA-BP智能算法的初始地应力场反演研究

土木工程 | 更新时间：2023-03-07

- 基于IA-BP智能算法的初始地应力场反演研究
- In-situ stress field inversion via IA-BP intelligent algorithm
- 土木与环境工程学报 2023年45卷第2期页码：89-99
- 作者机构：
  
  1.沈阳工业大学建筑与土木工程学院，沈阳 110870
  2.中铁十九局集团第五工程有限公司，辽宁大连 116100
- 作者简介：
  
  孙港（1997- ），男，主要从事岩石冲击与爆破研究，E-mail：1761152028@qq.com。
  王军祥（通信作者），男，副教授，博士生导师，E-mial：w.j.xgood@163.com。
- 基金信息：
  
  国家自然科学基金(51774066);辽宁省自然科学基金(2019-MS-242);辽宁省教育厅重点攻关项目(LZGD2020004);中国博士后科学基金(2018M630293)
- DOI：10.11835/j.issn.2096-6717.2021.116
  中图分类号： TU431
- 收稿：2021-01-15，
  
  纸质出版：2023-04-25
- 稿件说明：
移动端阅览
孙港, 王军祥, 郭连军, 等. 基于IA-BP智能算法的初始地应力场反演研究[J]. 土木与环境工程学报, 2023,45(2):89-99.

SUN Gang, WANG Junxiang, GUO Lianjun, et al. In-situ stress field inversion via IA-BP intelligent algorithm[J]. Journal of Civil and Environmental Engineering, 2023, 45(2): 89-99.
孙港, 王军祥, 郭连军, 等. 基于IA-BP智能算法的初始地应力场反演研究[J]. 土木与环境工程学报, 2023,45(2):89-99. DOI： 10.11835/j.issn.2096-6717.2021.116.

SUN Gang, WANG Junxiang, GUO Lianjun, et al. In-situ stress field inversion via IA-BP intelligent algorithm[J]. Journal of Civil and Environmental Engineering, 2023, 45(2): 89-99. DOI： 10.11835/j.issn.2096-6717.2021.116.

摘要

初始地应力场是地下工程设计与施工的重要依据，在实际工程中难以精准测得，为了能较准确地获得初始地应力场的分布规律，提出将免疫算法与BP神经网络相结合（IA-BP）的算法对初始地应力场进行反演研究。免疫算法优化BP神经网络就是将BP神经网络的连接权值和阈值作为免疫算法中的抗体进行编码。该混合算法既可以利用免疫算法全局寻优的特点快速搜索到全局最优解或次优解附近，又可以采用BP算法避免在最优解和次优解附近发生震荡，对其进行局部优化，从而达到快速收敛全局最优解的目的。通过COMSOL分别构建平面边坡模型及三维立体模型，对其进行正分析计算，将计算的结果作为“实测值”，对地应力进行反演分析，并将IA-BP算法反演的结果与PSO-BP算法及多元线性回归算法的反演结果进行对比。结果表明：二维边坡模型下，IA-BP算法反演结果误差更小。三维模型下，IA-BP算法所得实测值与反演值之间的相对误差的绝对值为0%~10.64%（平均为3.39%），PSO-BP算法所得实测值与反演值之间相对误差的绝对值为0%~48.39%（平均为6.93%），多元线性回归算法所得实测值与反演值之间相对误差的绝对值为0.55%~121.95%（平均为21.87%），通过对比可知，IA-BP算法整体反演结果精度最高。无论是平面模型还是三维模型，利用IA-BP算法反演出的结果与其他两种算法反演的结果相比，误差更小。将IA-BP智能算法运用到地应力场的反演研究中，可以为地下工程的建设提供依据。

Abstract

Initial in-situ stress field is an important basis for design and construction of underground engineering

while it is difficult to accurately measure the initial in-situ stress in engineering practice. In order to accurately obtain distribution law pattem of initial geostress field

the immune algorithm combined with BP neural network (IA-BP algorithm) for inversion of initial in-situ stress field is studied. The optimization of BP neural network by immune algorithm is to encode the connection weights and thresholds of BP neural network as antibodies in the immune algorithm. The hybrid algorithm can not only take advantage of the characteristics of immune algorithm as well as the global optimization quick search for the global optimal solution or near optimal solution

and can also adopt BP algorithm to avoid the near optimal and sub-optimal solutions

oscillation on the local optimization

realizing the aim of fast converge of the global optimal solution. The plane slope model and three-dimensional model were constructed by COMSOL respectively to carry out forward analysis and calculation

and the calculated results were taken as “measured values” to conduct inversion analysis of in-situ stress

but the inversion results of IA-BP algorithm were compared with those of PSO-BP and multiple linear regression algorithms. The results show that the inversion error of IA-BP algorithm is smaller under the two-dimensional (2D) slope model. Under the three-dimensional (3D) model

IA-BP algorithm from the measured values and the inversion of the absolute value of relative error between 0% and 10.64% (3.39% on average)

for PSO-BP

it is between 1% and 48.39% with average value of 6.93%

for MLR

it is between 0.55% and 121.95% with average of 21.87%. By comparison

it can be known that the overall inversion results of IA-BP algorithm have the highest accuracy. In both 2D and 3D models

the error of the inversion results is smaller than those using the other two. The application of IA-BP intelligent algorithm to the inversion of in-situ stress field can provide technical support for the construction of underground engineering.

关键词

Keywords

references

王金安 , 李飞 . 复杂地应力场反演优化算法及研究新进展 [J]. 中国矿业大学学报 , 2015 , 44 ( 2 ): 189 - 205 .

WANG J A LI F . Review of inverse optimal algorithm of in situ stress filed and new achievement [J]. Journal of China University of Mining & Technology , 2015 , 44 ( 2 ): 189 - 205 . (in Chinese)

杨德林 . 岩土工程问题的反演理论与工程实践 [M]. 北京 : 科学出版社， 1996 : 1 - 27 .

YANG D L . Inversion theory and enjineering practice for geotechnical enjineening problem [M]. Beijing : Science press . 1996 : 1 - 27 . (in Chinese)

LI Y , ZHANG Y Y , GAO R J , et al . In situ stress measurement and inversion in deep roadway [C]// 2nd International Conference on Energy and Environmental Protection . 2013 .

YU R G , TIAN Y . Application of hybrid genetic algorithm in ground stress inversion [J]. Applied Mechanics and Materials , 2011 , 90-93 : 337 - 341 .

LI X P , ZHOU X J , XU Z X , et al . Inversion method of initial in situ stress field based on BP neural network and applying loads to unit body [J]. Advances in Civil Engineering , 2020 , 2020 : 1 - 15 .

曹鹏 , 李科 , 李长江 , 等 . 五女峰隧道隧址区地应力场反演与岩爆风险分析 [J]. 现代隧道技术 , 2018 , 55 ( 5 ): 45 - 52 .

CAO P , LI K , LI C J , et al . Back-calculation of ground stress field and rockburst analysis for Wunvfeng tunnel [J]. Modern Tunnelling Technology , 2018 , 55 ( 5 ): 45 - 52 . (in Chinese)

汪波 , 何川 , 吴德兴 , 等 . 苍岭特长公路隧道地应力场反演分析 [J]. 岩土力学 , 2012 , 33 ( 2 ): 628 - 634 .

WANG B , HE C , WU D X , et al . Inverse analysis of in situ stress field of Cangling super-long highway tunnel [J]. Rock and Soil Mechanics , 2012 , 33 ( 2 ): 628 - 634 . (in Chinese)

代聪 , 何川 , 陈子全 , 等 . 超大埋深特长公路隧道初始地应力场反演分析 [J]. 中国公路学报 , 2017 , 30 ( 10 ): 100 - 108 .

DAI C , HE C , CHEN Z Q , et al . Inverse analysis of initial ground stress field of deep embedded and extra long highway tunnel [J]. China Journal of Highway and Transport , 2017 , 30 ( 10 ): 100 - 108 . (in Chinese)

张乐文 , 张德永 , 邱道宏 . 径向基函数神经网络在地应力场反演中的应用 [J]. 岩土力学 , 2012 , 33 ( 3 ): 799 - 804 .

ZHANG L W , ZHANG D Y , QIU D H . Application of radial basis function neural network to geostress field back analysis [J]. Rock and Soil Mechanics , 2012 , 33 ( 3 ): 799 - 804 . (in Chinese)

张社荣 , 胡安奎 , 王超 , 等 . 基于SLR-ANN的地应力场三维智能反演方法研究 [J]. 岩土力学 , 2017 , 38 ( 9 ): 2737 - 2745 .

ZHANG S R , HU A K , WANG C , et al . Three-dimensional intelligent inversion method for in situ stress field based on SLR-ANN algorithm [J]. Rock and Soil Mechanics , 2017 , 38 ( 9 ): 2737 - 2745 . (in Chinese)

涂图 , 王建 , 张梦迪 . 基于改进粒子群算法的BP神经网络在初始地应力场反演优化中的应用 [J]. 水电能源科学 , 2017 , 35 ( 12 ): 123 - 126, 139 .

TU T , WANG J , ZHANG M D . Application of BP neural network based improved PSO in initial geostress field inversion optimization [J]. Water Resources and Power , 2017 , 35 ( 12 ): 123 - 126, 139 . (in Chinese)

王志豪 . 基于免疫高斯过程算法的岩体物理力学参数位移反分析方法研究 [D]. 北京 : 北京交通大学 , 2018 .

WANG Z H . Displacement back analysis method for physical and mechanical parameters of rock mass based on immune and Gaussian process coupling algorithm [D]. Beijing : Beijing Jiaotong University , 2018 . (in Chinese)

焦李成 , 杜海峰 . 人工免疫系统进展与展望 [J]. 电子学报 , 2003 , 31 ( 10 ): 1540 - 1548 .

JIAO L C , DU H F . Development and prospect of the artificial immune system [J]. Acta Electronica Sinica , 2003 , 31 ( 10 ): 1540 - 1548 . (in Chinese)

蔺雅娴 . 基于IA-BP智能算法的应力-渗流耦合蠕变多参数反演方法及工程应用 [D]. 沈阳 : 沈阳工业大学 , 2019 .

LIN Y X . Multi-parameter inversion method of stress-seepage coupling creep based on IA-BP intelligent algorithm and its engineering application [D]. Shenyang : Shenyang University of Technology , 2019 . (in Chinese)

葛红 , 毛宗源 . 免疫算法几个参数的研究 [J]. 华南理工大学学报(自然科学版) , 2002 , 30 ( 12 ): 15 - 18 .

GE H , MAO Z Y . Research on parameters of immune algorithm [J]. Journal of South China University of Technology (Natural Science) , 2002 , 30 ( 12 ): 15 - 18 . (in Chinese)

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

高铁站候车厅多维物理环境交互作用对旅客满意度的影响机制