基于BO-SVR-BO的机制砂除云母工艺参数优化方法
Research on process parameter optimization of mica removal from manufactured sand based on BO-SVR-BO
关键词:
机制砂;云母去除;机器学习;支持向量机回归(SVR);贝叶斯优化(BO);工艺参数优化
Key words:
findings can provide a technical reference for the intelligent regulation and quality optimization of the mica removal process in manufactured sand production.
刊期:
2026年第5期
DOI:
机制砂;云母去除;机器学习;支持向量机回归(SVR);贝叶斯优化(BO);工艺参数优化
文章编号:
基金项目:
Research on process parameter optimization of mica removal from manufactured sand based on BO-SVR-BO
作者:
胡嘉靖1,2,张 明1,2,杨大兵1,2,*
单位:
1.武汉科技大学 资源与环境工程学院,湖北 武汉 430081;2.冶金矿产资源高效利用与造块湖北省重点实验室,湖北 武汉 430081
摘要
摘 要:为解决机制砂生产中因原料云母含量波动大、工艺参数调控依赖经验造成其云母含量超标问题,本文提出了一种基于双层贝叶斯优化支持向量机回归(BO-SVR-BO)的云母去除工艺参数优化方法。在第一层优化中,采用贝叶斯优化(Bayesian optimization,BO)对支持向量机回归(Support vector regression,SVR)的超参数进行全局寻优,构建针对云母含量、细度模数和产品率的高精度预测模型,并与反向传播(Backpropagationk,BP)神经网络、长短期记忆网络(Long short-term memory,LSTM)、门控循环单元(Gated recurrent unit,GRU)和多层感知机(Multilayer perceptron,MLP)模型进行对比;在第二层优化中,以训练好的BO-SVR为代理模型,结合BO实现风选风速与磁场强度的多目标全局优化。结果表明:在同一测试集下,BO-SVR模型在云母含量、细度模数和产品率的预测中表现出更高的精度与稳定性;在工艺参数优化方面,BO-SVR-BO能够在确保选别后机制砂云母含量≤2%、细度模数符合相关标准要求的前提下,显著提升产品率。研究成果可为机制砂除云母工艺的智能调控和质量优化提供技术参考。
Abstract: To address the problems of excessive mica content in finished manufactured sand caused by large fluctuations in the content of mica in raw material and experience-dependent regulation of process parameters during manufactured sand production, this paper proposes a process parameter optimization method for mica removal based on double-layer Bayesian optimization-support vector regression (BO-SVR-BO). In the first layer optimization, Bayesian optimization (BO) is used to globally optimize the hyperparameters of support vector regression (SVR) and construct a high-precision prediction model for mica content, fineness modulus, and product yield. The prediction performance is compared with that of backpropagation (BP) neural network, long short-term memory (LSTM), gated recurrent unit (GRU), and multilayer perceptron (MLP) model. In the second layer optimization, the trained BO-SVR is used as the surrogate model, and is combined with BO to achieve multi-objective global optimization of classification wind speed and magnetic field strength. The results show that on the same test set, the BO-SVR model exhibits higher prediction accuracy and stability in predicting mica content, fineness modulus, and product yield. In terms of process parameter optimization, the BO-SVR-BO method can significantly improve product yield while ensuring that the mica content of the manufactured sand after mica removal is ≤2% and that the fineness modulus meets the requirements of relevant standards. The findings can provide a technical reference for the intelligent regulation and quality optimization of the mica removal process in manufactured sand production.
结论
(1)本文结合SVR对小样本非线性问题的建模优势与BO高效全局寻优能力,构建了双层贝叶斯优化支持向量机回归(BO-SVR-BO)工艺参数决策模型。该模型通过第一层BO优化SVR超参数,建立高精度预测模型;再以预测模型为代理,利用第二层BO在约束条件下对风选风速和磁场强度进行全局优化,从而实现机制砂生产过程中云母含量、细度模数与产品率的协同调控。
(2)BO-SVR模型与BP神经网络模型、LSTM模型、GRU模型、MLP模型在同一测试集下的对比结果表明,BO-SVR模型在云母含量、细度模数和产品率的预测中具备更高的精度与稳定性。
(3)工艺参数优化结果表明,BO-SVR-BO模型实现了机制砂生产工艺的智能建模与参数优化,能够在确保选别后机制砂云母含量≤2%、细度模数符合相关标准要求的前提下,显著提升产品率。该机器学习方法为机制砂除云母工艺的智能调控和质量优化提供了可靠的技术路径,具备在实际工程中推广应用的潜力。
参考文献
[1] 林天乐.机制砂特性对水泥基材料物理力学性能的影响[D].杭州:浙江大学,2022.
[2] 刘正义,刘红旭.花岗岩铀成矿作用的模拟实验[J].地学前缘,2009,16(1):99-113.
[3] 杨占峰,李顺凯,赵欢.机制砂中云母含量对C50混凝土力学和耐久性能的影响[J].公路,2025,70(8):350-354.
[4] LIU H,YANG X,JIANG L,et al.Study on the effect of interfacial modification on the properties of super standard mica sand cement-based materials[J].Buildings,2024,14(6):1665.
[5] 覃靖,李顺凯,廖巧玲,等.磁选去除机制砂云母对混凝土性能影响研究[J].公路,2022,67(4):327-330.
[6] CHI H,WANG C,TIAN Y,et al.Unraveling polycarboxylate superplasticizer (PCE) compatibility in muscovite-blended cement paste through aggregation mechanisms[J].Journal of Building Engineering,2024,95:110133.
[7] LV C,WU H,SHI M,et al.Experimental study on the mechanical strength, deformation behavior and infiltration characteristics of coral sand[J].Sustainability,2024,16(8):3479.
[8] XUE H L,HAN C Q,CHEN M L,et al.Improving mechanical properties of manufactured sand concrete with high biotite content: Application of magnetic separation process and equipment optimization[J].Construction and Building Materials,2022,350:128861.
[9] TANG Y,QIU W,LIU D,et al.Experimental study on the properties of mortar and concrete made with tunnel slag machine-made sand[J].Materials,2022,15(14):4817.
[10] WANG X,ZHANG Y,ZHANG W,et al.Research on improving the performance and mechanism of manufactured sand concrete based on vibration-coupled mixing[J].Journal of Building Engineering,2025,108:112971.
[11] WU H,LIN S,CHENG X,et al.Comparative study of strontium adsorption on muscovite, biotite and phlogopite[J].Journal of Environmental Radioactivity,2020,225:106446.
[12] 吴海军,杨大兵,罗作仟,等.一种花岗岩机制砂除云母的方法及系统:CN202411616484.2[P].2025-02-28.
[13] 冯艳虎.老挝孟昆褐铁矿焙烧—重磁联合选矿工艺研究[D].昆明:昆明理工大学,2023.
[14] HUANG F Z,YANG J H,FANG H Y,et al.Research on online measurement method of fineness modulus of manufactured sand based on deep learning[J].Construction and Building Materials,2024,414:134922.
[15] 郭耀雄,史星祥,徐志军,等.基于图像处理技术的机制砂形貌及机制砂混凝土性能研究[J].混凝土与水泥制品,2023(9):1-6,11.
[16] 邵建峰,倪卫红,岳晓伟,等.基于BP神经网络的制砂工艺参数控制研究[J].混凝土,2020(11):154-156.
[17] 张云升,王习,张文华,等.机制砂绿色制备工艺仿真模拟及其参数优化[J].建筑材料学报,2025,28(7):591-597,707.
[18] 陶松涛,延永东,郑玉龙,等.机制砂混凝土抗压强度分析及预测[J].混凝土与水泥制品,2022(8):96-100,104.
[19] TONG H,CHEN D R,YANG F.Support vector machines regression with unbounded sampling[J].Applicable Analysis,2018,98(9):1626-1635.
[20] YANG K,LIU L,WEN Y.The impact of Bayesian optimization on feature selection[J].Scientific Reports,2024,14:3948.
[21] MOTOYAMA Y,TAMURA R,YOSHIMI K,et al.Bayesian optimization package: PHYSBO[J].Computer Physics Communications,2022,278:108405.
[22] 袁力.面向SMT锡膏印刷的关键工艺参数优化技术研究[D].成都:电子科技大学,2025.
引用本文
胡嘉靖,张明,杨大兵.基于BO-SVR-BO的机制砂除云母工艺参数优化方法[J].混凝土与水泥制品,2026,53(5):47-53. HU J J,ZHANG M,YANG D B.Research on process parameter optimization of mica removal from manufactured sand based on BO-SVR-BO[J].China Concrete and Cement Products,2026,53(5):47-53 (in Chinese).
下一个: 试论混凝土智能制造中的三大关键技术
相关文件