目的 建立基于MATLAB的BP神经网络模型的淀粉离散元接触参数标定方法。方法 建立单球型颗粒,并通过颗粒缩放分析将粒径放大至0.8 mm,以待标定的5种型号淀粉(高预胶化淀粉:S-1、低预胶化淀粉:S-2、直压型预胶化淀粉:S-3、水溶性淀粉:S-4、玉米淀粉:S-5)离散元接触参数(颗粒-颗粒恢复系数:A, 颗粒-颗粒静摩擦系数:B,颗粒-颗粒滚动摩擦系数:C, 颗粒-不锈钢恢复系数:D,颗粒-不锈钢静摩擦系数:E,颗粒-不锈钢滚动摩擦系数:F,表面能(J·m-2):G)为输入层,两种测量方法(提升缸法和剪切盒法)休止角为输出层,MATLAB随机抽样50组进行离散元仿真模拟。BP神经网络算法对模拟结果进行训练,得到理想的神经网络模型后分别对以上淀粉的接触参数进行预测,并进行物理实验验证。结果 当BP神经网络中隐含层中神经元个数为11个时,训练样本与测试样本的决定系数R2分别为0.999 9和0.940 9,拟合较好,所建立神经网络的预测输出可以达到期望输出。获得的参数组合休止角模拟值与实测值相对误差均小于2.5%,表明预测准确。结论 BP神经网络预测药用辅料淀粉的离散元接触参数可靠,可为后续固体制剂的仿真模拟提供可靠的数据支持。
Abstract
OBJECTIVE To build a calibration method for the discrete element method contact parameters of starch based on BP neural network model. METHODS Single spherical granules were created and the particle size was scaled up to 0.8 mm by particle scaling analysis. The discrete element contact parameters(particle-particle recovery coefficient A, particle-particle static friction coefficient B, particle-particle rolling friction coefficient C, particle-stainless steel recovery coefficient D, particle-stainless steel static friction coefficient E, particle-stainless steel rolling friction coefficient F, particle-particle surface energy(J·m-2) G) of the five starch(high pregelatinized starch: S-1, low pregelatinized starch: S-2, direct compression pregelatinized starch S-3, water soluble starch S-4, maize starch S-5) to be calibrated are used as the input layer, and the angle of repose obtained by the two measurement methods(hoist cylinder method and shear box method) is used as the output layer to build a BP neural network.The contact parameters were randomly sampled by MATLAB for 50 groups, and then the discrete element simulation was carried out according to the combination of parameters. The BP neural network algorithm was used to train the simulation results, and then the contact parameters of the starch were predicted, which were verified by physical experiments. RESULTS When the number of neurons in the hidden layer in the BP neural network is 11, the coefficient of determination R2 for the training and test samples was 0.999 9 and 0.940 9, respectively, which are a good fit, so the established neural network can be proved to be reliable. The fractional error between the simulated and measured values of the obtained parameter combinations of rest angle is less than 2.5%, indicating that the predictions are accurate. CONCLUSION The discrete element contact parameters of starch obtained using BP neural network predictions are reliable and can provide reliable data support for subsequent simulations of solid formulations.
关键词
淀粉 /
离散元 /
BP神经网络 /
颗粒缩放 /
参数标定 /
休止角
{{custom_keyword}} /
Key words
starch /
discrete element method /
BP neural network /
scale-up particle /
parameter calibration /
repose angle
{{custom_keyword}} /
中图分类号：
R94
{{custom_clc.code}}
({{custom_clc.text}})
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] LI M T, LIU Q, ZHANG Y Q, et al. Preparation and characterization of ternary indomethacin solid dispersion based on quality by designd QbD concept[J]. Chin Pharm J(中国药学杂志), 2020, 55(17): 1450-1455.
[2] NING J T, SUN J M, WANG Z M, et al. Design and in vitro release study of valsartan bilayer tablets based on quality by design(QbD)concept[J]. Chin Pharm J(中国药学杂志), 2021, 56(3): 210-217.
[3] ZENG J, HUANG T, FENG L L, et al. Progress in numerical simulation technology for fluidized bed granulation[J]. Chin J New Drugs(中国新药杂志), 2019, 28(11): 1330-1342.
[4] CHEN K K. Investigation into calibration method of discrete element model parameters for bulk material and its application[D].Wuhan: Wuhan University of Technology, 2017.
[5] COETZEE J C. Calibration of the discrete element method[J]. Powder Technol, 2017, 310: 104-142.
[6] GRIMA A P,WYPYCH P W.Investigation into calibration of discrete element model parameters for scale-up and validation of particle-structure interactions under impact conditions[J]. Powder Technol,2011,212(1): 198-209.
[7] LI Y X, LI F X, XU X M, et al. Parameter calibration of wheat flour discrete element method simulation based on particle scaling[J]. Trans Chin Soc Agric Eng(农业工程学报), 2019, 35(16): 320-327.
[8] SHI C F, YANG M R, TANG Z X, et al. Calibration of discrete element simulation parameters of Chinese medicine extract powders[J]. Chin Tradit Herb Durgs(中草药), 2020, 51(24): 6205-6212.
[9] LU S, YUAN Q X, GOUDA S, et al. Parameters calibration of vermicomposting nursery substrate with discrete element method based on JKR contact model[J]. Trans Chin Soc Agric Mach(农业机械学报), 2018, 49(4): 343-350.
[10] WU L, QU F Z, LI S J. Inversion on discrete element model parameters of conditioned soil of earth pressure balance shield machine[J]. J Dalian Univ Technol(大连理工大学学报), 2010, 5(6): 860-866.
[11] CHENG R, LIU J, HAN X, et al. Reverse calculation technique of concrete material dynamic constitutive parameters by stages[J]. Explos Shock Waves(爆炸与冲击), 2014, 34(3): 315-321.
[12] WANG Q J, LU P. Research on application of artificial intelligence in computer network technology[J]. Int J Pattern Recogn, 2019, 33(5): 1959015.Doi.org/10.1142/S0218001419590158.
[13] FEI Q, ZHANG W H, YAN L J, et al. Parameter optimization of corn ear drying device based on BP neural network[J]. China South Agric Mach(南方农机),2021,52(19):46-47,57.
[14] WU S H, ZHOU Y, LI Z X, et al. Novel approach for simulation optimization of time-varying parameters system based on neural network[J/OL]. J Syst Eng Elect(系统工程与电子技术), 2021:1-11,[2021-11-22]. http://kns.cnki.net/kcms/detail/11.2422.TN.20211101.1512.002.html.
[15] ZHOU X P, XU Q, ZHAO K Y, et al. Research on calibration method of discrete element mesoscopic parameters based on neural network landslide in Heifangtai,Gansu as an example[J]. Chin J Rock Mech Eng(岩石力学与工程学报),2020,39(Suppl 1):2837-2847.
[16] HE Y, MENG Z Y, XU H, et al. A dynamic model of evaluating differential automatic method for solving plane problems based on BP neural network algorithm[J]. Physica A(Amsterdam, Neth.), 2020, 556: 124845.Doi:10.1016/j.physa.2020.124845.
[17] LIU G D. Preparation and release behavior of gelatin slow release excipient starch[D]. Wuxi: Jiangnan University, 2017.
[18] LEMIEUX M, GOSSELIN P, MATEESCU M A. Carboxymethyl high amylose starch as excipient for controlled drug release:mechanistic study and the influence of degree of substitution[J]. Int J Pharm, 2009, 382(1-2): 172-182.
[19] ZHOU L H. EDEM simulation and experiment of Vertical screw conveying[D]. Hangzhou: Zhejiang University of Technology, 2017.
[20] THAKUR S C, OOI J Y, AHMADIAN H. Scaling of discrete element model parameters for cohesionless and cohesive solid[J]. Powder Technol, 2016, 293: 130-137.
[21] LI J C, QIN G J, WEN X S,et al. Over-fitting in neural network learning algorithms and its solving strategies[J]. Vibration Test Diagnosis(振动、测试与诊断), 2002,(4): 16-20,76.
[22] REN J L, ZHOU L H, HAN L,et al. Discrete simulation of vertical screw conveyor based on particle scaling theory[J]. Chin J Process Eng(过程工程学报), 2017, 17(5): 936-943.
[23] FENG Y T, OWEN D R J. Discrete element modelling of large scale particle systems-Ⅰ: exact scaling laws[J]. Computational Particle Mechanics, 2014, 1(2): 159-168.
[24] JUARZE E E, OLIVAS G I, ZAMUDIO F P B, et al. Effect of water content on the flowability of hygroscopic powders[J]. J Food Eng, 2017,205:12-17.
[25] JOZEF H, MAREK M. Parameters and contact models for DEM simulations of agricultural granular materials: A review[J]. Biosyst Eng,2016,147:206-225.
[26] BENVENUTI L,KLOSS C,PIRKER S. Identification of DEM simulation parameters by artificial neural networks and bulk experiments[J]. Powder Technol, 2016, 291: 456-465.
[27] LANG J, HE Q Q, FAN X M,et al. Prediction of airflow classification effect of wet coal based on BP neural network[J/OL]. J China Coal Soc(煤炭学报),2021:1-10[2021-11-9]. http://kns.cnki.net/kcms/detail/11.2190.TD.20211116.1.
[28] ZHANG H, GUO Y, LI M. Review of development and application of artificial neural network models[J]. Comput Eng Appl(计算机工程与应用), 2021, 57(11):57-69.
[29] KATTERFELD A, COETZEE C, DONOHUE T,et al. Calibration of DEM Parameters for Cohesionless Bulk Materials under Rapid Flow Conditions and Low Consolidation[M/OL]. Research Gate,2019[2021-11-9]. https://www.researchgate.net/publication/334129169.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
基金
国家科技重大新药创制专项资助(2017ZX09101-001-006);重庆市技术创新与应用发展专项面上项目资助(cstc2020jscx-msxmX0048)
{{custom_fund}}
PDF(1915 KB)
