融合 ANN 增强机理和 SVM 残差补偿的锡-铅-锑真空分离含量预测

谢慧 ,  李根根 ,  戴卫平 ,  余勇 ,  李建春 ,  张发祥 ,  杨春曦

有色金属(冶炼部分) ›› 2026, Vol. 0 ›› Issue (11) : 2388 -2397.

PDF (2549KB)
有色金属(冶炼部分) ›› 2026, Vol. 0 ›› Issue (11) : 2388 -2397. DOI: 10.20237/j.issn.1007-7545.2026.11.010

融合 ANN 增强机理和 SVM 残差补偿的锡-铅-锑真空分离含量预测

作者信息 +

Prediction of the Content of Tin-Lead-Antimony in Vacuum Separation Based on the Fusion of ANN Enhancement Mechanism and SVM Residual Compensation

Author information +
文章历史 +
PDF (2609K)

摘要

在锡- 铅- 锑多元合金真空蒸馏过程中,液相中锡、铅、锑的实时含量是评价分离效果与冶金效率的关键指标,但目前工业中仅能通过离线检测获取,这种数据反馈的大滞后导致真空炉无法及时调整工艺参数,从而影响生产过程的稳定性与总体经济效益。因此,本文提出一种融合人工神经网络(Artificial Neural Network,ANN)增强机理与支持向量机(Support Vector Machine,SVM)残差补偿的混合建模方法(AMM-SVM),实现上述含量的软测量。该方法首先利用ANN 优化机理模型的关键参数,以克服原有模型在复杂工况下刻画非线性动力学行为的不足;随后通过SVM 构建残差补偿机制,进一步提升预测精度。同时,采用Adam 优化器和粒子群优化算法(Particle Swarm Optimization,PSO)分别优化ANN 与SVM 的超参数。最后,以云南某企业生产的DBLL11-10 型多功能锡铅锑合金真空分离电阻炉在二十多个锡冶炼厂的实际运行数据为例,验证了所提模型的有效性。对比试验结果表明,AMM-SVM 模型在预测性能上表现最优。

Abstract

The vacuum distillation technology, with its advantages of short process, no pollution, low energy consumption and high resource utilization rate, provides an innovative solution for the green and efficient upgrading of crude tin refining. The process of vacuum distillation is based on the vapor pressure differences between other elements and tin. By heating, arsenic, antimony, lead, bismuth, etc. are preferentially volatilized and separated through condensation, thereby efficiently purifying tin metal. In the vacuum distillation process of tin-lead-antimony alloys, the real-time content of tin, lead and antimony in the volatile substances directly reflects the effect of distillation separation, and is a key indicator for evaluating the quality of vacuum metallurgy. However, in current industrial production, the content of volatile substances can only be obtained by sampling the final product and conducting offline testing. This significant delay in data feedback prevents vacuum furnaces from promptly adjusting process parameters, thereby affecting the stability of the production process and overall economic benefits. Therefore, this paper proposes a hybrid modeling method (AMM-SVM) that integrates Artificial Neural Network (ANN) enhanced mechanism and Support Vector Machine (SVM) residual compensation for soft measurement of tin, lead, and antimony contents in volatile substances. Firstly, in response to the issue that the ternary system mechanism model of tin-lead-antimony can describe the equilibrium trend, but is difficult to accurately depict the nonlinear dynamic behavior under complex working conditions, the key parameters of the mechanism model are optimized through data-driven methods using ANN. Secondly, a residual compensation model based on SVM is constructed to further model and correct the prediction residuals of the previous model, thereby effectively enhancing the overall prediction accuracy of the hybrid model. Based on this, in order to fully utilize the advantages of different optimization algorithms, the adaptive learning rate optimization algorithm (Adam) and the particle swarm optimization algorithm (PSO) were adopted to optimize the hyperparameters of the ANN and SVM models respectively. In the Adam optimizer, the learning rate α was set to 0.01, the learning rate scheduler was set to 20, the parameter ε was set to 0.5, and the total number of training rounds was 100; In the PSO optimizer, the number of particles was set to 80, the number of iteration rounds was set to 20, the inertia weight w was set to 0.8, the cognitive coefficient c1 was set to 2, and the social coefficient c2 was set to 2. The key parameters C, ε, and γ of SVM were globally optimized within the search space [1, 4], [0.01, 0.1], and [1, 4] respectively. This hybrid optimization strategy takes into account the efficient gradient update ability of Adam and the global exploration advantage of PSO, thereby systematically improving the model performance and generalization ability. The AMM-SVM hybrid soft measurement model constructed in this study effectively solves the problem of real-time monitoring of volatile substances such as tin, lead, and antimony content in the vacuum distillation process through a multi-layer structure of "mechanism prior knowledge and ANN deviation fitting and SVM residual compensation". The verification results show that this model achieves the best performance in all indicators such as MAE (0.042 7), RMSE (0.059 1), and R2 (0.978 6), significantly outperforming the pure mechanistic model (which cannot provide reasonable predictions), the pure data-driven model, and similar hybrid models. It demonstrates higher accuracy and reliability. This model can directly replace the lagging offline detection methods and provide reliable data support for the real-time regulation of process parameters and the stable improvement of product quality.

关键词

有色金属 / 锡- 铅- 锑合金 / 真空冶金工艺 / 混合建模 / 残差补偿 / 超参数优化

Key words

nonferrous metals / tin-lead-antimony alloy / vacuum metallurgy process / hybrid modeling / residual compensation / hyperparameter optimization

引用本文

引用格式 ▾
谢慧,李根根,戴卫平,余勇,李建春,张发祥,杨春曦. 融合 ANN 增强机理和 SVM 残差补偿的锡-铅-锑真空分离含量预测[J]. 有色金属(冶炼部分), 2026, 0(11): 2388-2397 DOI:10.20237/j.issn.1007-7545.2026.11.010

登录浏览全文

4963

注册一个新账户 忘记密码

参考文献

[1]

戴永年, 杨斌 . 有色金属材料的真空冶金[M]. 北京: 冶金工业出版社, 2000: 88-103.

[2]

Dai Yongnian, Yang Bin . Vacuum metallurgy of non-ferrous materials[M]. Beijing: Metallurgical Industry Press, 2000: 88-103.

[3]

孔令鑫, 杨斌, 李一夫, 等 . MIVM在真空蒸馏分离锡锌合金中的应用及实验研究[J]. 真空科学与技术学报, 2013, 33(5): 483-489.

[4]

Kong Lingxin, Yang Bin, Li Yifu, et al. Modeling and experimental study of tin-zinc alloy separation by vacuum distillation[J]. Chinese Journal of Vacuum Science and Technology, 2013, 33(5): 483-489.

[5]

孔令鑫 . 分子相互作用体积模型在锡基合金真空蒸馏中的应用研究[D]. 昆明: 昆明理工大学, 2013: 21-22.

[6]

Kong Lingxin . Study on the application of molecular interaction volume model in vacuum distillation of tin-based alloys[D]. Kunming: Kunming University of Science and Technology, 2013: 21-22.

[7]

李一夫 . 复杂锡合金真空蒸馏分离提纯的研究[D]. 昆明: 昆明理工大学, 2015: 27-38.

[8]

Li Yifu . Study on separation and purification of complex tin alloys by vacuum distillation [D]. Kunming: Kunming University of Science and Technology, 2015: 27-38.

[9]

王安祥 . 锡锑合金真空蒸馏分离的研究[D]. 昆明: 昆明理工大学, 2015: 24-26.

[10]

Wang Anxiang . Study on vacuum distillation separation of tin-antimony alloy[D]. Kunming: Kunming University of Science and Technology, 2015: 24-26.

[11]

王安祥, 杨斌, 李一夫, 等 . 分子相互作用体积模型在锡锑合金真空蒸馏深度脱锑中的应用[J]. 真空科学与技术学报, 2015, 35(4): 502-507.

[12]

Wang Anxiang, Yang Bin, Li Yifu, et al. Simulation of removal of Sb from Sn-Sb alloy in vacuum with molecular interaction volume model[J]. Chinese Journal of Vacuum Science and Technology, 2015, 35(4): 502-507.

[13]

任佳琦, 徐俊杰, 孔令鑫, 等 . 锡基合金体系组元活度及气-液平衡的模型预测[J]. 中国有色金属学报, 2020, 30(10): 2399-2409.

[14]

Ren Jiaqi, Xu Junjie, Kong Lingxin, et al. Model prediction of activity and vapor-liquid equilibrium of tin-based alloy system[J]. The Chinese Journal of Nonferrous Metals, 2020, 30(10): 2399-2409.

[15]

伍美珍, 陈亮亮, 贾元伟, 等 . 铟-锡合金真空蒸馏气-液平衡研究[J]. 有色金属工程, 2023, 13(9): 42-47.

[16]

Wu Meizhen, Chen Liangliang, Jia Yuanwei, et al. Study on vapor-liquid equilibrium of In-Sn binary alloy in vacuum distillation[J]. Nonferrous Metals Engineering, 2023, 13(9): 42-47.

[17]

李家文, 赖华, 冯丽辉 . 基于神经网络与LabVIEW平台建立焊锡真空炉粗锡含铅量的软测量模型与在线监测研究[J]. 冶金分析, 2017, 37(2): 1-6.

[18]

Li Jiawen, Lai Hua, Feng Lihui . Research on soft measurement model and on-line monitoring of lead content in crude tin in solder vacuum furnace based on neural network and LabVIEW[J]. Metallurgical Analysis, 2017, 37(2): 1-6.

[19]

申孟亚 . 基于LabVIEW的焊锡真空炉粗锡质量的软测量与模糊控制研究[D]. 昆明: 昆明理工大学, 2015: 37-43.

[20]

Shen Mengya . Research on soft sensing and fuzzy control of crude tin quality in solder vacuum furnace based on LabVIEW[D]. Kunming: Kunming University of Science and Technology, 2015: 37-43.

[21]

蒲正浩, 韩继标, 李一夫, 等 . 响应曲面优化粗锡真空蒸馏[J]. 过程工程学报, 2018, 18(3): 496-502.

[22]

Pu Zhenghao, Han Jibiao, Li Yifu, et al. Optimization of crude tin vacuum distillation by response surface methodology[J]. The Chinese Journal of Process Engineering, 2018, 18(3): 496-502.

[23]

何宁 . 焊锡真空脱铅炉计算机优化控制策略研究[D]. 昆明: 昆明理工大学, 2001: 39-55.

[24]

He Ning . Research on computer-optimized control strategy for solder vacuum lead-removal furnaces[D]. Kunming: Kunming University of Science and Technology, 2001: 39-55.

[25]

冯丽辉 . 多台焊锡真空蒸馏脱铅炉的优化控制与管理[J]. 工业仪表与自动化装置, 2002(6): 45-47.

[26]

Feng Lihui . The optimization control and management of multiple soldering vaccum furnaces based on DCS[J]. Industrial Instrumentation & Automation, 2002(6): 45-47.

[27]

冯丽辉, 樊雪萍 . 内热式多级连续真空炉实时最优控制[J]. 有色金属, 1999, 51(3): 67-69.

[28]

Feng Lihui, Fan Xueping . On-line optimizing control of inner-heating, multi-stage continuous vacuum furnace[J]. Nonferrous Metals, 1999, 51(3): 67-69.

[29]

速斌, 邵丹, 马祥亚. 一种锡铅锑砷合金真空蒸馏分离锡的方法: 201510060060.7[P]. 2015-05-27.

[30]

Su Bin, Shao Dan, Ma Xiangya, et al. A method for separating tin from tin-lead-antimony-arsenic alloy by vacuum distillation: CN 201510060060.7[P]. 2015-05-27.

[31]

孔令鑫 . Pb-Sn-Sb三元合金气液相平衡研究[D]. 昆明: 昆明理工大学, 2016: 68-69.

[32]

Kong Lingxin . Study on gas-liquid equilibrium of Pb-Sn-Sb ternary alloy[D]. Kunming: Kunming University of Science and Technology, 2016: 68-69.

[33]

余东原, 罗跃, 梁大业, 等 . 基于机器学习方法预测地浸过程中铀浸出金属量的变化[J]. 有色金属(冶炼部分), 2024(2): 92-98.

[34]

Yu Dongyuan, Luo Yue, Liang Daye, et al. Predicting the variation of uranium leaching metal content in ground-leaching process based on machine learning methods[J]. Nonferrous Metals (Extractive Metallurgy), 2024(2): 92-98.

[35]

马朝君, 彭巨擘, 袁海滨, 等 . 基于灰狼优化支持向量机回归与SHAP值的锡冶炼能耗预测[J]. 有色金属(冶炼部分), 2024(2): 1-7.

[36]

Ma Chaojun, Peng Jobo, Yuan Haibin, et al. Energy consumption prediction of tin smelting based on grey wolf optimized support vector machine regression and SHAP values[J]. Nonferrous Metals (Extractive Metallurgy), 2024(2): 1-7.

[37]

He Juanxia, Huang Liwen, Xiao Yao, et al. Prediction model of continuous discharge coefficient from tank based on KPCA-DE-SVR[J]. Journal of Loss Prevention in the Process Industries, 2024, 89: 105316. DOI: 10.1016/j.jlp.2024.105316.

[38]

Kinga D, Ba J . Adam: a method for stochastic optimization[C]// Proceedings of the 3rd International Conference on Learning Representations (ICLR 2015). California, 2015. DOI: 10.48550/arXiv.1412.6980.

[39]

Cunha L, Zhang Li, Sowan B, et al. Video deepfake detection using Particle Swarm Optimization improved deep neural networks[J]. Neural Computing and Applications, 2024, 36(15): 8417-8453.

[40]

Wu Long, Yang Chunxi, Zhang Xiufeng, et al. Prediction of cobalt ions concentration in zinc smelting purification section Ⅱ based on classification modeling[C]// 2025 IEEE 14th Data Driven Control and Learning Systems, Wuxi, China. IEEE, 2025: 1054-1059.

基金资助

云南省重大科技专项项目(202302AD080005)

云南省基础研究计划项目(202301BE070001-064)

AI Summary AI Mindmap
PDF (2549KB)

0

访问

0

被引

详细

导航
相关文章

AI思维导图

/

〈 〉