于 雨, 王余莲, 朱益斌, 张一帆, 李克卿, 关 蕊, 孙浩然, 韩会丽, 袁志刚

(沈阳理工大学 材料科学与工程学院, 辽宁 沈阳 110159)

引用格式:

于雨, 王余莲, 朱益斌, 等. 基于 BP 神经网络的三水碳酸镁晶须制备工艺优化[J]. 中国粉体技术,2024, 30(1): 103-113.

YU Y, WANG Y L, ZHU Y B, et al. Optimization of nesquehonite whisker preparation process based on BP neural network[J].China Powder Science and Technology,2024, 30(1): 103-113

摘要: 【目的】为了得到更大长径比的三水碳酸镁晶须, 基于反向传播( back-propagation, BP)神经网络方法对三水碳酸镁的制备工艺进行优化。 【方法】以菱镁矿为原料, MgCl₂ 为添加剂, 采用正交试验和 BP 神经网络相结合的方法, 对三水碳酸镁制备工艺中的反应时间、 反应温度、 搅拌速率、 pH 和 MgCl₂ 用量等条件进行优化。 【结果】正交试验优化后的条件是反应时间为 80 min, 反应温度为 40 ℃ , 搅拌速率为 500 r/ min, pH 为 7±0. 1, MgCl₂ 质量浓度为 0. 5 g / L, 可获得长径比约为 20 的三水碳酸镁晶须; BP 神经网络优化后的条件是反应时间为 86 min, 反应温度为 44 ℃ , 搅拌速率为760 r/ min、 pH 为 7,MgCl₂ 质量浓度为 1. 6 g / L,可获得长径比约为 25 的三水碳酸镁晶须。 【结论】利用 BP 神经网络优化工艺参数能制备更大长径比的三水碳酸镁晶须,优于传统工艺方法。

关键词: 菱镁矿; 反向传播神经网络; 三水碳酸镁晶须

Abstract

Objective In order to increase the aspect ratio of nesquehonite, a single factor test or orthogonal test is usually used to determine the optimal preparation process of nesquehonite. However, these two methods suffer from the problems of errors due to random effects, complicated experimental data, and large workloads. Optimize the preparation process of nesquehonite using BP neural network can effectively improve aspect ratio of nesquehonite and avoid the above problems.

Methods Nesquehonite were prepared by a simple process using magnesite as raw material and MgCl₂ as an additive. A combination of orthogonal test and BP neural network was used to optimize five conditions in the process, including reaction time, reaction temperature, stirring rate, pH, and MgCl₂ dosage. Firstly, an orthogonal test of L₁₆(4⁵) was designed to find out the optimal process conditions, while the influence of each condition on the aspect ratio of nesquehonite was judged by the range analysis.Then, a three-layer BP neural network model was designed with the number of nodes in the hidden layer determined by Eq. (1) and the loss function as MSE. Based on the optimal process conditions of orthogonal test, the model was used to predict the relationship between each condition and the variation of the aspect ratio of nesquehonite. Moreover, the optimal preparation process optimized by BP neural network model was obtained.

Results and Discussion The results of XRD and SEM images show that nesquehonite whisker with an aspect ratio of 20 could be obtained through orthogonal test under the optimized conditions of 80 min for reaction time, 40℃ for reaction temperature, 500 r/ min for stirring rate, 7±0. 1 for pH, and 0. 5 g / L for MgCl₂ dosage. The influence of each condition on the aspect ratio of nesquehonite can be obtained according to the range analysis , and the order is MgCl₂ dosage, reaction temperature, pH, reaction time, and stirring rate in descending. From the results of orthogonal tests, it can be seen that the aspect ratio of nesquehonite is significantly increased by the addition of MgCl₂ because Mg²⁺ accelerates the nucleation rate and the growth rate of nesquehonite. Samples with high pH have a small aspect ratio and a rough surface, which may be due to the fact that it will accelerate the dissolution of nesquehonite and the transformation to hydromagnesite with a large amount of OH^- entering the solution. When the temperature is too high, nesquehonite is easily converted to hydromagnesite, resulting in a decrease in the aspect ratio. Nesquehonite whisker with an aspect ratio of 25 could be obtained by BP neural network with the optimized conditions of 86 min for reaction time, 44 ℃ for reaction temperature, 760 r/ min for stirring rate, 7 for pH, and 1. 6 g / L for MgCl₂ dosage. The crystal aspect ratio shows a tendency to increase and then decrease with the extension of reaction time. With the rise of reaction temperature, the whisker aspect ratio shows a tendency to increase and then decrease. It can be seen the stirring rate increases, that the aspect ratio of nesquehonite shows an increasing trend, and it is close to 800 r/ min, that the aspect ratio shows a weak decreasing trend. With the increase of pH, the aspect ratio of nesquehonite decreased instead. The aspect ratio of nesquehonite whisker shows a trend of enlargement and then decreased with the increase of MgCl₂ dosage.

Conclusion In this work, a BP neural network model for the relationship between the preparation process and whisker aspect ratio of nesquehonite based on orthogonal test data is established, and the error between the experimental value and the predicted value is less than 7. 2%, which indicates that the neural network is relatively accurate. On the basis of orthogonal test, the preparation process of nesquehonite is further optimized by using BP neural network, that the aspect ratio is increased by 25% and the workload is reduced. It indicates that it is feasible to optimize the preparation process of nesquehonite by using BP neural network.

Keywords: magnesite; back-propagation neural network; nesquehonite whisker

参考文献(References):

[1]KLOPROGGE T, MARTENS W, NOTHDURFT L, et al. Low temperature synthesis and characterization of nesquehonite[J]. Journal of Materials Science Letters, 2003, 22(11): 825-829.

[2]闫平科, 程书林, 卢智强, 等. 油酸钠对三水碳酸镁晶须稳定性的研究[J]. 硅酸盐通报, 2016, 35(3): 732-735.

YAN P K, CHENG S L, LU Z Q, et al. Effect of sodium oleate on stability of nesquehonite whisker[ J]. Bulletin of the Chinese Ceramic Society, 2016, 35(3): 732-735.

[3]时天骄, 王余莲, 刘珈伊, 等. 共沉淀法制备三水碳酸镁晶体及生长机理[ J]. 中国粉体技术, 2021, 27( 5):120-127.

SHI T J, WANG Y L, LIU J Y, et al. Preparation and growth mechanism of nesquehonite crystals with coprecipitation method[J]. China Powder Science and Technology, 2021, 27(5): 120-127.

[4]DING W J, OUYANG J, YANG H M. Synthesis and characterization of nesquehonite (MgCO₃·3H₂O) powders from natural talc[J]. Powder Technology, 2016, 292(9): 169-175.

[5]欧龙, 刘家祥. 重镁水中通入空气制备三水碳酸镁晶须[J]. 硅酸盐学报, 2016, 44(1): 104-111.

OU L, LIU J X. Preparation of nesquehonite whiskers through bubbling air into magnesium bicarbonate solution[J]. Journal of the Chinese Ceramic Society, 2016, 44(1): 104-111.

[6]周品. MATLAB 神经网络设计与应用[M]. 北京: 清华大学出版社, 2013: 68-89. 

ZHOU P. MATLAB neural network design and application[M]. Beijing: Tsinghua University Press, 2013: 68-89.

[7]张晓威, 冯玉杰, 陈宇, 等. 均匀共沉淀法制备 NiO 包覆 Al 粉体及性能研究[ J]. 人工晶体学报, 2012, 41(2):479-485.

ZHANG X W, FENG Y J, CHEN Y, et al. Preparation and properties of NiO coated Al powders by homogeneous coprecipitation method[J]. Journal of Synthetic Crystals, 2012, 41(2): 479-485.

[8]姚尚锋. 基于正交试验的神经网络设计[J]. 数学的实践与认识, 2008(13): 116-122.

YAO S F. Design of neural network based on experiment of orthogonality [ J]. Mathematics in Practice and Theory,2008(13): 116-122.

[9]周开利, 康耀红. 神经网络模型及其 MATLAB 仿真程序设计[M]. 北京: 清华大学出版社, 2005: 9-16.

ZHOU K L, KANG Y H. Neural network models and their MATLAB simulation program design[M]. Beijing: Tsinghua University Press, 2005: 9-16.

[10]LIN J. Prediction for dilution rate of AlCoCrFeNi coatings by laser cladding based on a BP neural network[ J]. Coatings,2021, 11(11): 1402.

[11]XING S, JU J, XING J. Research on hot-rolling steel products quality control based on BP neural network inverse model [J]. Neural Computing and Applications, 2019, 31(5): 1577-1584.

[12]LIU C, DING W, LI Z, et al. Prediction of high-speed grinding temperature of titanium matrix composites using BP neural network based on PSO algorithm[ J]. The International Journal of Advanced Manufacturing Technology, 2016, 89( 5): 2277-2285.

[13]崔万顺, 文伟翔, 高玉娟, 等. L-天门冬氨酸镁制备高热稳定性三水碳酸镁晶体[ J]. 北京化工大学学报(自然科学版), 2023, 50(2): 63-69.

CUI W S, WEN W X, GAO Y J, et al. Preparation of nesquehonite crystals with high thermal stability from magnesium L-aspartate[J]. Journal of Beijing University of Chemical Technology (Natural Science Edition), 2023, 50(2): 63-69.

[14]陈娟, 黄志良, 陈常连, 等. 反应温度和反应时间对三水碳酸镁晶须长径比的影响[J]. 武汉工程大学学报, 2017,29(1): 54-58. 

CHEN J, HUANG Z, CHEN C L, et al. Influence of temperature and time on aspect ratios of magnesium carbonate trihydrate whiskers[J]. Journal of Wuhan Institute of Technology, 2017, 39(1): 54-58.

[15]BHATTACHARJYA D, SELVAMANI T, MUKHOPADHYAY I. Thermal decomposition of hydromagnesite[ J]. J Therm Anal Calorim, 2012, 107(2): 439-445.

[16]赵航毅, 王余莲, 时天骄, 等. 氯化铵辅助三水碳酸镁水热法合成无水碳酸镁及其形成机理[ J]. 中国粉体技术,2022, 28(6): 49-57.

ZHAO H Y, WANG Y L, SHI T J, et al. Synthesis and formation mechanism of anhydrous magnesium carbonate by hydrothermal method of nesquehonite with assistance of ammonium chloride[J]. China Powder Science Technology, 2022, 28(6): 49-57.

[17]王丽, 孙宝昌, 周海军, 等. 超重力法制备三水碳酸镁晶须[J]. 北京化工大学学报(自然科学版), 2014, 41(2):13-18.

WANG L, XUN B C, ZHOU H J, et al. The preparation of nesquehonite whiskers by high gravity technology[J]. Journal of Beijing University of Chemical Technology (Natural Science Edition), 2014, 41(2): 13-18.

[18]SIRINE G, JULIEN M, LAN-HUONG T, et al. A study of hydromagnesite and nesquehonite precipitation in indirect aqueous carbonation of thermally-activated serpentine in a batch mode[J]. Journal of Crystal Growth, 2022, 584(8): 126540.

[19]王素平, 肖殷, 王世荣, 等. 高镁转化率三水碳酸镁晶须生长条件的研究[ J]. 功能材料, 2016, 47(2): 2116-2124, 2129. 

WANG S P, XIAO Y, WANG S R, et al. Crystallization processes of nesquehonite whisker with high conversion of magnesium[J]. Journal of Functional Materials, 2016, 47(2): 2116-2124, 2129.

[20]王余莲, 印万忠, 张夏翔, 等. 大长径比三水碳酸镁晶须的制备及晶体生长机理[J]. 硅酸盐学报, 2018, 46(7): 938-945.

WANG Y L, YIN W Z, ZHANG X X, et al. Preparation and growth mechanism of nesquehonite whiskers with large aspect ratio[J]. Journal of the Chinese Ceramic Society, 2018, 46(7): 938-945.

[21]王勇, 赵风云, 胡永琪, 等. 晶型控制剂对沉淀碳酸钙晶型、 形态的影响[J]. 无机盐工业, 2006(3): 5-8.

WANG Y, ZHAO F Y, HU Y Q, et al. Influence of chemical additives on the crystal polymorphs and shapes of precipitated calcium carbonate[J]. Inorganic Chemicals Industry, 2006(3): 5-8.