img

QQ群聊

img

官方微信

高级检索

黄金科学技术 ›› 2024, Vol. 32 ›› Issue (4): 675-684.doi: 10.11872/j.issn.1005-2518.2024.04.006

• 采选技术与矿山管理 • 上一篇    下一篇

基于SVR的PFC微观参数辅助标定方法研究

温晨1(),黄敏1,邱贤阳2,黄帅2   

  1. 1.紫金(长沙)工程技术有限公司,湖南 长沙 410208
    2.中南大学资源与安全工程学院,湖南 长沙 410083
  • 收稿日期:2024-03-26 修回日期:2024-04-08 出版日期:2024-08-31 发布日期:2024-08-27
  • 作者简介:温晨(1998-),男,河南安阳人,硕士研究生,从事控制爆破及岩层控制研究工作。1398448249@qq.com
  • 基金资助:
    “十四五”重点研发计划项目“特大型多金属资源高通量分选关键技术与装备”(2022YFC2904602);广西重点研发计划“复杂地表环境下地下矿山开采岩移规律及低沉降充填开采技术研究”(2022AB31023)

Auxilliary Calibration Method for Microscopic Parameters of PFC Based on SVR

Chen WEN1(),Min HUANG1,Xianyang QIU2,Shuai HUANG2   

  1. 1.Zijin(Changsha) Engineering Technology Co. , Ltd. , Changsha 410208, Hunan, China
    2.School of Resources and Safety Engineering, Central South University, Changsha 410083, Hunan, China
  • Received:2024-03-26 Revised:2024-04-08 Online:2024-08-31 Published:2024-08-27

摘要:

PFC数值模拟所需的微观参数通常通过人工试算的方式进行标定,该方法受标定人员经验的影响,效率较低,难以快速处理大量岩石试件。以平行黏结模型为例,建立微观参数正交试验表并进行数值模拟,以此为样本分别使用支持向量回归机(SVR)和BP神经网络模型进行训练,对室内测得的宏观参数进行预测,得到的微观参数进行数值模拟分析预测效果,若效果不佳则将模拟数据加入样本继续训练直至获得理想的结果。研究表明:利用数值模拟和机器学习相结合的正反演方法,可以高效标定微观参数,其中BP神经网络模型需要试算7次,而支持向量机模型仅需试算3次,标定效率更高。因此,基于正反演结合的SVR微观参数辅助标定方法不仅效率高、可重复性强、不受标定人员经验影响,而且适用于批量试件的标定工作。

关键词: 参数标定, 颗粒流, 支持向量回归机, 反向传播神经网络, 正交试验, 正反演

Abstract:

In numerical simulation studies utilizing the Particle Flow Code (PFC),the direct acquisition of microscopic parameters of discrete particles through experimental means presents a significant challenge.Traditionally,manual trial-and-error techniques are utilized,involving continuous adjustments of microscopic parameters to observe the corresponding effects on macroscopic mechanical parameters within the simulation.This iterative process is characterized by its randomness and lack of systematic approach,heavily reliant on the expertise of the calibrator,ultimately leading to diminished calibration efficiency and reproducibility.This challenge is particularly evident when calibrating microscopic parameters for extensive quantities of rock samples,requiring substantial manual labor and repetitive tasks.In order to mitigate these issues,a novel approach for calibrating microscopic parameters is essential,one that is not reliant on the calibrator’s skill level and ensures consistent reproducibility in the calibration of rock specimen parameters on a large scale.Utilizing the parallel bond model as a case study,an orthogonal experimental design table was constructed to investigate microscopic parameters,followed by numerical simulations to generate a comprehensive small-sample dataset.Support vector regression (SVR) and back propagation(BP) neural network models were separately trained on this dataset.This approach involves utilizing macroscopic parameters derived from PFC numerical simulations as the forward process,with machine learning techniques employed to predict microscopic parameters as the inverse process.If the prediction error for the macroscopic parameters measured in the laboratory is deemed inadequate,the corresponding microscopic parameters and their resultant macroscopic parameters are incorporated into the dataset for additional training until the desired outcome is attained.Studies have shown that utilizing machine learning with small-sample data,in conjunction with forward and inverse modeling,can effectively calibrate parameters.In particular,the BP model required 7 iterations,whereas the SVR model only needed 3 iterations to attain satisfactory outcomes,showcasing superior calibration efficiency.In scenarios involving numerous and highly nonlinear macro-micro parameters,the utilization of machine learning-assisted calibration presents notable benefits over traditional manual trial-and-error approaches,including enhanced efficiency,increased reproducibility,and improved generalizability.

Key words: parameters calibration, particle flow, support vector regression, back propagation neural network, orthogonal experimental design, forward and inversion methods

中图分类号: 

  • TU45

图1

参数标定流程"

图2

线性平行黏结模型(Potyondy,2018)"

图3

PFC2D数值模型"

表1

正交试验因素和水平数据"

水平Eˉ*/GPaκˉ*μσˉc /MPacˉ /MPa?ˉ
1310.2151515
2620.4302030
3930.6452545
41240.8603060

表2

正交试验表"

编号Eˉ* /GPaκˉ*μσˉc/MPacˉ/MPa?ˉ/(°)E/GPaνσc/MPaσt?/MPaσc/σt
1310.21515154.820.16922.405.474.1
2310.44530454.940.14652.6711.854.4
3320.24525604.100.30542.375.927.2
4320.41520304.230.26827.575.285.2
5330.66015303.850.32436.577.614.8
6330.83030604.000.31335.517.794.6
7340.66025453.630.39943.907.625.8
8340.86020153.690.38452.9210.904.9
9610.66020459.330.14335.845.666.3
10610.83025159.480.12838.536.785.7
11620.63030307.760.27141.948.265.1
12620.86015607.900.25023.963.866.2
13630.21520606.470.38118.643.485.4
14630.44525306.820.36648.679.924.9
15640.24530156.010.43852.3010.764.9
16640.41515456.280.40419.933.615.5
17910.260253012.380.17337.315.996.2
18910.430206012.890.14924.734.725.2
19920.230154510.510.30521.804.934.4
20920.460301510.930.27547.8411.884.0
21930.615251510.150.33816.063.614.5
22930.845204510.200.33632.686.744.9
23940.64515609.350.38121.763.366.5
24940.81530309.010.37818.623.555.3
251210.615306016.000.15219.955.623.6
261210.845153016.380.13326.664.665.7
271220.645201513.770.25234.414.837.1
281220.815254514.170.24320.613.775.5
291230.260304511.380.36447.459.025.3
301230.430151511.920.31324.454.705.2
311240.230203010.420.42127.925.235.3
321240.460256010.990.41031.415.026.3

图4

SVR模型示意图"

图5

K折交叉验证(K=5)"

表3

室内试验宏观参数"

编号E/GPaνσc?/MPaσt?/MPa
18.400.29029.82.4
28.870.22126.22.4
316.710.27047.39

表4

第一组参数标定结果"

次数微观参数(预测)宏观参数(数值模拟)误差百分比/%
Eˉ*/GPaκˉ*μσˉc/MPacˉ/MPa?ˉ/(°)E/GPaνσc/MPaσt/MPaσc /σtEνσcσt
16.521.730.51625.315.741.48.660.24827.45.64.93.114.58.1132
27.042.270.47828.519.142.18.640.28530.15.95.12.91.71144
待标定参数8.400.29029.82.412.4

图6

第一组预测参数PFC数值模拟结果"

表5

第二组参数标定结果"

次数微观参数(预测)宏观参数(数值模拟)误差百分比/%
Eˉ*/GPaκˉ*μσˉc/MPacˉ/MPa?ˉ/(°)E/GPaνσc/MPaσt/MPaσc /σtEνσcσt
16.691.840.62425.917.035.68.740.24529.05.05.81.510.910.72.7
26.711.810.59222.717.736.48.860.25126.44.75.60.113.60.810.4
36.701.720.59123.117.539.98.970.22924.05.14.71.13.68.42.9
46.691.780.56322.517.137.98.730.22427.94.955.61.61.46.54.8
待标定参数8.870.22126.25.25.0

图7

第二组预测参数PFC数值模拟结果"

表6

第三组参数标定结果"

次数微观参数(预测)宏观参数(数值模拟)误差百分比/%
Eˉ*/GPaκˉ*μσˉc/MPacˉ/MPa?ˉ/(°)E/GPaνσc/MPaσt/MPaσc /σtEνσcσt
112.62.110.52657.124.331.814.40.27243.48.05.413.80.78.210.1
213.02.260.50359.426.432.014.40.28546.510.14.613.85.61.711.1
313.52.250.48056.227.927.015.30.29049.39.85.08.47.44.29.1
414.22.140.49561.926.530.515.90.27950.39.25.54.83.36.32.3
待标定参数16.70.27047.39.05.3

图8

第三组预测参数PFC数值模拟结果"

表7

BP神经网络参数标定结果"

次数微观参数(预测)宏观参数(数值模拟)误差百分比/%
Eˉ*/GPaκˉ*μσˉc/MPacˉ/MPa?ˉ/(°)E/GPaνσc/MPaσt/MPaσc /σtEνσcσt
18.032.260.51136.521.638.59.620.30238.18.74.48.536.745.466.5
26.461.470.55341.518.549.68.140.21833.05.65.93.01.426.06.5
36.241.860.66041.119.353.68.430.25334.46.25.65.014.531.319.6
46.121.500.53241.020.356.98.510.21631.55.85.44.00.520.211.1
56.821.920.65944.616.659.69.150.25626.84.56.03.215.82.313.5
66.811.670.60241.018.952.49.210.22930.55.85.33.83.616.411.9
75.851.630.58142.616.655.58.110.22527.04.85.68.61.83.16.9
86.351.620.63746.518.055.98.610.21627.75.05.55.72.92.34.6
待标定参数8.870.22126.25.25.0
Abi E D, Zheng Yingren, Feng Xiating,et al,2018.Relationship between particle micro and macro mechanical parameters of parallel-bond model [J].Rock and Soil Mechanics,39(4):1289-1301.
Benvenuti L, Kloss C, Pirker S,2016.Identification of DEM simulation parameters by artificial neural networks and bulk experiments[J].Powder technology,291:456-465.
Chen Pengyu, Kong Ying, Yu Hongming,2018.Research on the calibration method of microparameters of a uniaxial compression PFC2D model for rock[J].Chinese Journal of Underground Space and Engineering,14(5):1240-1249.
Cundall P A, Strack O D L,1979.A discrete numerical model for granular assembles[J].Geotechnique,29(1):47-65.
Cundall P A,1971.The Measurement and Analysis of Acceleration in Rock Slopes[D].London:Imperial College of Science and Technology.
Deng Shuxin, Zheng Yonglai, Feng Lipo,et al,2019.Application of design of experiments in microscopic parameter calibration for hard rocks of PFC3D model[J].Chinese Journal of Geotechnical Engineering,41(4):655-664.
Feng Haotian, Chen Junzhi, Ren Chunfang,et al,2022.Study on calibration method of cemented structure plane parameters in numerical test[J].Chinese Journal of Underground Space and Engineering,18(6):1824-1833.
Feng Xiating, Zhang Zhiqiang, Yang Chengxiang,et al,1999.Study on genetic-neural network method of displacement back analysis [J].Chinese Journal of Rock Mechanics and Engineering,18(5):529-533.
Huang Yisheng, Xia Xiaodan,2021.Calibration method of mesoscopic parameters for parallel bonding model of sandstone particle flow[J].Journal of China Three Gorges University(Natural Sciences),43(4):7-12.
Jiang Mingjing, Fang Wei, Sima Jun,2015.Calibration of micro-parameters of parallel bonded model for rocks[J].Journal of Shandong University(Engineering Science),45(4):50-56.
Jiang Yue, Zhou Wendong,2023.A study on the correlation of macro and microstructural parameters of hollow cylindrical grey sandstone based on PFC 3D[J/OL].Coal Science and Technology,1-13[2024-08-15]..
Potyondy D O,2018.Material-Modeling Support in PFC[M].Minneapolis:ITASCA.
Potyondy D O, Cundall P A,2004.A bonded-particle model for rock[J].International Journal of Rock Mechanics and Mining Sciences,41(8):1329-1364.
Ren Junqing, Xiao Ming, Liu Guoqing,2023.Lightweight analysis method for rock macro-meso parameters based on improved BP algorithm[J].Journal of Hunan University(Natural Sciences),50(9):207-218.
Wang Hongbo, Ma Zhe,Wulantuya,et al,2022.Calibration method of mesoscopic parameters using BP neural network and Burgers model[J].Transactions of the Chinese Society of Agricultural Engineering,38(23):152-161.
Wang Zhaoyang, Lin Peng, Xu Zhenhao,et al,2022.A transversely isotropic rocks integrated microparameter calibration method for flat joint model and smooth joint model[J].Journal of Central South University (Science and Technology),53(6):2211-2223.
Wu Changyou,2007.The Research and Application on Nerual Network[D].Harbin:Northeast Agricultural University.
Wu Luyuan, Zhu Yongheng, Bai Haibo,et al,2023.Study on the correlation of macro and meso parameters of parallel bond model sandstone[J].Journal of Mining Science and Technology,8(4):487-501.
Zhao Guoyan, Dai Bing, Ma Chi,2012.Study of effects of microparameters on macroproperties for parallel bonded model [J].Chinese Journal of Rock Mechanics and Engineering,31(7):1491-1498.
Zhong Weiliang, Ding Hao, Fan Lifeng,2023.Research on mesoscopic parameters calibration of geopolymer concrete upon BP neural network[J/OL].Engineering Mechanics,1-10[2024-08-15]..
Zhou Xiaopeng, Xu Qiang, Zhao Kuanyao,et al,2020.Research on calibration method of discrete element mesoscopic parameters based on neural network landslide in Heifangtai,Gansu as an example[J].Chinese Journal of Rock Mechanics and Engineering,39(Supp.1):2837-2847.
阿比尔的,郑颖人,冯夏庭,等,2018.平行黏结模型宏细观力学参数相关性研究[J].岩土力学,39(4):1289-1301.
陈鹏宇,孔莹,余宏明,2018.岩石单轴压缩PFC2D模型细观参数标定研究[J].地下空间与工程学报,14(5):1240-1249.
邓树新,郑永来,冯利坡,等,2019.试验设计法在硬岩PFC3D模型细观参数标定中的应用[J].岩土工程学报,41(4):655-664.
冯豪天,陈俊智,任春芳,等,2022.数值试验中胶结结构面参数标定方法的研究[J].地下空间与工程学报,18(6):1824-1833.
冯夏庭,张治强,杨成祥,等,1999.位移反分析的进化神经网络方法研究[J].岩石力学与工程学报,18(5):529-533.
黄宜胜,夏晓丹,2021.砂岩颗粒流平行黏结模型细观参数标定方法研究[J].三峡大学学报(自然科学版),43(4):7-12.
姜玥,邹文栋,2023.基于PFC3D的空心圆柱灰砂岩宏细观参数相关性研究[J/OL].煤炭科学技术,1-13[2024-08-15]..
蒋明镜,方威,司马军,2015.模拟岩石的平行黏结模型微观参数标定[J].山东大学学报(工学版),45(4):50-56.
任俊卿,肖明,刘国庆,2023.基于改进BP算法的岩石宏细观参数轻量化分析方法[J].湖南大学学报(自然科学版),50(9):207-218.
王朝阳,林鹏,许振浩,等,2022.横观各向同性岩体平直—光滑节理双模型细观参数联合标定方法[J].中南大学学报(自然科学版),53(6):2211-2223.
王洪波,马哲,乌兰图雅,等,2022.采用BP神经网络和Burgers模型的细观参数标定[J].农业工程学报,38(23):152-161.
吴昌友,2007.神经网络的研究及应用[D].哈尔滨:东北农业大学.
吴禄源,朱永恒,白海波,等,2023.砂岩颗粒流平行黏结模型宏细观参数关联性研究[J].矿业科学学报,8(4):487-501.
赵国彦,戴兵,马驰,2012.平行黏结模型中细观参数对宏观特性影响研究[J].岩石力学与工程学报,31(7):1491-1498.
钟惟亮,丁昊,范立峰,2023.基于BP神经网络的地聚物混凝土细观参数标定研究[J/OL].工程力学,1-10[2024-08-15]..
周小棚,许强,赵宽耀,等,2020.基于神经网络的离散元细观参数标定方法研究——以甘肃黑方台黄土滑坡为例[J].岩石力学与工程学报,39(增1):2837-2847.
[1] 许云美, 袁利伟, 龙皓楠. 干堆尾矿库稳定性影响因素的敏感性分析[J]. 黄金科学技术, 2023, 31(6): 1014-1022.
[2] 荣光旭, 李宗洋. CNN-LSTM模型在边坡可靠度分析中的应用[J]. 黄金科学技术, 2023, 31(4): 613-623.
[3] 王卫华, 黄瑞新, 罗杰. 应力波在非线性变形节理处传播规律的颗粒流模拟研究[J]. 黄金科学技术, 2023, 31(4): 580-591.
[4] 章逸锋,李洪超,张智宇,李恒. 基于正交试验的小断面巷道掏槽爆破参数确定[J]. 黄金科学技术, 2023, 31(2): 331-339.
[5] 高峰,艾浩泉,梁耀东,罗增武,熊信,周科平,杨根. 基于NSGA-Ⅱ算法的废石及尾砂混合充填料配比优化[J]. 黄金科学技术, 2022, 30(1): 46-53.
[6] 张美道,饶运章,徐文峰,王文涛. 全尾砂膏体充填配比优化正交试验[J]. 黄金科学技术, 2021, 29(5): 740-748.
[7] 黄仁东,李哲. 基于正交试验的细尾砂—分级尾砂充填体强度研究[J]. 黄金科学技术, 2021, 29(2): 256-265.
[8] 王卫华,罗杰,刘田,韩震宇. 节理粗糙度对应力波传播及试样破坏影响的颗粒流模拟[J]. 黄金科学技术, 2021, 29(2): 208-217.
[9] 胡建华,庞乐,王学梁,郑明华. 基于正交试验的过断层软破段巷道支护参数优化[J]. 黄金科学技术, 2020, 28(6): 859-867.
[10] 张雷, 郭利杰, 李文臣. 基于铜镍冶炼渣制备充填胶凝材料试验研究[J]. 黄金科学技术, 2020, 28(5): 669-677.
[11] 梁昌金, 马传净. 碘—氨浸出体系用于废旧印刷线路板中金的浸取[J]. 黄金科学技术, 2019, 27(5): 784-790.
[12] 徐文峰,饶运章,李尚辉,许威. 含膨润土充填料浆泌水特性分析[J]. 黄金科学技术, 2019, 27(3): 433-439.
[13] 蓝志鹏,王新民,张钦礼,陈秋松. 磁化水对膏体料浆管道输送摩阻损失影响试验研究[J]. 黄金科学技术, 2018, 26(6): 811-818.
[14] 王新民,胡一波,王石,刘吉祥,陈宇,卞继伟. 超细全尾砂充填配比优化正交试验研究[J]. 黄金科学技术, 2015, 23(3): 45-49.
[15] 曾庆田,刘科伟,严体,王李管. 基于多数值模拟方法联合的自然崩落法开采研究[J]. 黄金科学技术, 2015, 23(1): 66-73.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!