基于人工神经网络的金刚石微粉化学镀镍层性能预测

方莉俐; 刘韩; 姜羽飞

doi:10.13394/j.cnki.jgszz.2024.0042

基于人工神经网络的金刚石微粉化学镀镍层性能预测

doi: 10.13394/j.cnki.jgszz.2024.0042

方莉俐^{1, 2,},
刘韩^{1, 2},
姜羽飞^{1, 2}

1.
中原工学院物理与光电工程学院, 郑州 450007
2.
郑州市低维量子材料及器件重点实验室, 郑州 450007

基金项目: 2021 年河南省重点研发与推广专项（212102210488）。

详细信息

作者简介:
方莉俐，女，1965年生，博士、教授。主要研究方向：薄膜材料和薄膜物理、金刚石及其制品等。E-mail：hnzzfll@126.com

中图分类号: TQ164
计量
- 文章访问数: 51
- HTML全文浏览量: 32
- PDF下载量: 3
- 被引次数: 0
出版历程
- 收稿日期: 2024-03-06
- 修回日期: 2024-06-07
- 录用日期: 2024-06-17
- 刊出日期: 2025-04-20

Prediction of properties of electroless nickel plating with diamond powder based on artificial neural network

FANG Lili^{1, 2
,},
LIU Han^{1, 2},
JIANG Yufei^{1, 2}

1.
Physics and Optoelectronic Engineering, Zhongyuan University of Technology, Zhengzhou 450007, China
2.
Zhengzhou Key Laboratory of Low-Dimensional Quantum Materials and Devices, Zhengzhou 450007, China

摘要

摘要: 用化学镀方法在M1/2、M6/12、M20/30金刚石微粉表面镀镍，并用人工神经网络预测金刚石颗粒粒径、次亚磷酸钠浓度、镀液温度、镀液pH值等化学镀工艺参数，对镀层沉积速率、镀层密度、镀层耐腐蚀性等镀层性能的影响。结果表明：构建的BP神经网络模型和GRNN模型，经过样本数据训练学习后适用于金刚石微粉化学镀镍层性能的预测；训练完成的BP神经网络和GRNN预测值与实际样品测量值相对误差绝对值的平均值分别为9.14%和5.07%，两者都有较好的预测效果；且在金刚石微粉化学镀镍层性能预测中，GRNN的预测性能优于BP神经网络的预测性能。
- 金刚石微粉 /
- 化学镀镍 /
- 镀层性能 /
- BP神经网络 /
- GRNN
Abstract: Objectives To improve the quality of chemical plating on diamond micropowders, an experimental analysis was conducted on the influence of key process parameters on the plating quality during the chemical plating process. The experimental results were then predicted using artificial neural networks. Methods Nickel plating experiments were carried out on the surfaces of M1/2, M6/12, and M20/30 micron diamond powders using the electroless plating method. The effects of electroless plating process parameters—such as diamond particle size, concentration of sodium hypophosphite, plating solution temperature, and plating solution pH—on the coating properties were investigated. The performance of the coatings were evaluated as follows: (1) The deposition rate of the coating was expressed as the difference in the quality of diamond powder before and after electroless plating per unit time. (2) The coating density was expressed as the mass of the coating per unit volume. (3) Each group of coated diamond powders was immersed in hydrochloric acid solution with a mass fraction of 10% for 24 hours, and the corrosion weight loss of diamond powder was used to indicate the coating's corrosion resistance of the coating—where higher corrosion weight loss indicates poorer corrosion resistance. Data on the influences of process parameters, such as diamond particle size, sodium hypophosphite concentration, plating solution temperature, and plating solution pH on coating performance were used as the training set. Both BP and GRNN artificial neural networks were applied to predict the deposition rate, coating density, and corrosion resistance under four different conditions. The accuracy of the models was evaluated by comparing experimental data with predicted values. Results The BP neural network model and the GRNN model can be used to predict the coating performance of micron diamond powders after training on sample data. The absolute relative error between the predicted coating performance values of the BP neural network model and the experimental values was less than 15.00%, with an average absolute relative error of 9.14%. The absolute relative error between the predicted coating performance values and experimental values of the GRNN model was less than 10.00%, with an average absolute relative error of 5.07%. In predicting the performance of electroless nickel plating on diamond micro powders, the predictive performance of GRNN is superior to that of BP neural network. Conclusions The prediction error values of BP neural network and GRNN for the chemical plating performance of diamond micropowder are both less than 10.00%, which proves that they can be used to predict the relevant results and reduce the number of experiments to obtain optimal process parameters. And the prediction error of GRNN is smaller than that of BP neural network, which proves that the performance of GRNN in prediction experiments is better than that of BP neural network.
- diamond powder /
- chemical nickel plating /
- coating performance /
- BP neural network /
- GRNN

HTML全文

图 1 BP神经网络算法流程图

Figure 1. BP neural network algorithm flowchart

下载: 全尺寸图片幻灯片

图 2 GRNN算法流程图^[19]

Figure 2. GRNN algorithm flowchart^[19]

下载: 全尺寸图片幻灯片

图 3 BP神经网络结构

Figure 3. BP neural network structure

下载: 全尺寸图片幻灯片

图 4 样本的BP神经网络训练结果

Figure 4. Training results of samples using BP neural network

下载: 全尺寸图片幻灯片

图 5 GRNN结构

Figure 5. GRNN structure

下载: 全尺寸图片幻灯片

图 6 GRNN中光滑因子寻优

Figure 6. Optimization of smoothing factor in GRNN

下载: 全尺寸图片幻灯片

图 7 2种神经网络预测结果的相对误差绝对值对比

Figure 7. Comparison of relative error absolute values between two neural network prediction results

下载: 全尺寸图片幻灯片

表 1 化学镀实验工艺参数的因素和水平

Table 1. Factors and levels of experimental process parameters for chemical plating

水平	因素
水平	次亚磷酸钠浓度 ρ₁ / (g·L⁻¹)	镀液pH值	镀液温度 θ / ℃
1	25.0	3	30
2	27.5	5	45
3	30.0	7	60
4	32.5	9	75
5	35.0	11	90

下载: 导出CSV

表 2 训练样本

Table 2. Training samples

组号	输入值				输出值
组号	粒度标记	次亚磷酸钠浓度 ρ₁ / (g·L⁻¹)	镀液pH值	镀液温度 θ / ℃	镀层沉积速率 v₁ / (g·h⁻¹)	镀层密度 ρ₂ / (g·cm⁻³)	镀层耐腐蚀性能 m / g
1	M20/30	25.0	3	30	0.0500	0.5784	0.0420
2	M20/30	25.0	5	60	0.2500	0.6864	0.0733
3	M20/30	25.0	7	90	1.0100	0.1791	0.3073
4	M20/30	25.0	9	45	0.3800	0.2220	0.1658
5	M20/30	25.0	11	75	0.6600	4.5814	0.2541
…	…	…	…	…	…	…	…
36	M6/12	30.0	3	75	0.3500	1.8224	0.0660
37	M6/12	30.0	5	30	0.1200	2.8382	0.0373
38	M6/12	30.0	7	60	0.5200	0.2157	0.3646
39	M6/12	30.0	9	90	0.5400	0.7598	0.1465
40	M6/12	30.0	11	45	0.4500	0.1699	0.0522
…	…	…	…	…	…	…	…
71	M1/2	35.0	3	45	0.2600	0.9648	0.0238
72	M1/2	35.0	5	75	2.2600	5.9523	0.2161
73	M1/2	35.0	7	30	0.7800	3.3864	0.3403
74	M1/2	35.0	9	60	1.8000	3.4721	0.4498
75	M1/2	35.0	11	90	1.2900	1.4304	0.0189

下载: 导出CSV

表 3 BP神经网络预测值与实测结果对比

Table 3. Comparison between BP neural network predicted values and measured results

组号	网络输入值				网络输出值			实测结果			相对误差 δ / %
组号	金刚石粒度标记	次亚磷酸钠浓度 ρ₁ / (g·L⁻¹)	镀液 pH值	镀液温度 θ / ℃	镀层沉积速率 v₁ / (g·h⁻¹)	镀层密度 ρ₂ / (g·cm⁻³)	镀层耐腐蚀性能 m₁ / g	镀层沉积速率 v₂ / (g·h⁻¹)	镀层密度 ρ₃ / (g·cm⁻³)	镀层耐腐蚀性能 m₂ / g	镀层沉积速率	镀层密度	镀层耐腐蚀性能
P1	M4/8	30.0	11	45	0.603 0	2.601 0	0.194 0	0.570 0	2.288 0	0.178 0	5.79	13.68	8.99
P2	M4/8	35.0	5	75	0.599 0	1.063 0	0.131 0	0.560 0	1.149 0	0.116 0	6.96	−7.49	12.93
P3	M8/12	30.0	11	45	0.485 0	0.752 0	0.165 0	0.510 0	0.821 0	0.189 0	−4.90	−8.40	−12.70
P4	M8/12	35.0	5	75	0.580 0	1.689 0	0.160 0	0.540 0	1.842 0	0.182 0	7.41	−8.31	−12.09

下载: 导出CSV

表 4 GRNN预测值与实测结果对比

Table 4. Comparison between GRNN predicted values and measured results

组号	网络输入值				网络输出值			实测结果			相对误差 δ / %
组号	金刚石粒度标记	次亚磷酸钠浓度 ρ₁ / (g·L⁻¹)	镀液 pH值	镀液温度 θ / ℃	镀层沉积速率 v₁ / (g·h⁻¹)	镀层密度 ρ₂ / (g·cm⁻³)	镀层耐腐蚀性能 m₁ / g	镀层沉积速率 v₂ / (g·h⁻¹)	镀层密度 ρ₃ / (g·cm⁻³)	镀层耐腐蚀性能 m₂ / g	镀层沉积速率	镀层密度	镀层耐腐蚀性能
P1	M4/8	30.0	11	45	0.528 0	2.459 0	0.182 0	0.570 0	2.288 0	0.178 0	−7.37	7.47	2.25
P2	M4/8	35.0	5	75	0.595 0	1.079 0	0.121 0	0.560 0	1.149 0	0.116 0	6.25	−6.09	4.31
P3	M8/12	30.0	11	45	0.532 0	0.808 0	0.185 0	0.510 0	0.821 0	0.189 0	4.31	−1.58	−2.12
P4	M8/12	35.0	5	75	0.581 0	1.922 0	0.195 0	0.540 0	1.842 0	0.182 0	7.59	4.34	7.14

下载: 导出CSV

参考文献(23)

[1]	刘金婷, 贺瑞, 刘瑞莹, 等. 碳化硅半导体材料应用及发展前景 [J]. 科技创新导报,2019,16(25):74-76. doi: 10.16660/j.cnki.1674-098X.2019.25.074 LIU Jinting, HE Rui, LIU Ruiying, et al. Application and development prospect of silicon carbide semiconductor materials [J]. Science and Technology Innovation Herald,2019,16(25):74-76. doi: 10.16660/j.cnki.1674-098X.2019.25.074
[2]	王宝玉, 宋为, 张太超. 电镀金刚线技术探讨 [J]. 金属制品,2018,44(3):10-13. doi: 10.3969/j.issn.1003-4226.2018.03.003 WANG Baoyu, SONG Wei, ZHANG Taichao. Discussion on electroplating diamond wire technology [J]. Metal Products,2018,44(3):10-13. doi: 10.3969/j.issn.1003-4226.2018.03.003
[3]	李振兴. 半导体晶体线锯切割工艺研究 [J]. 红外,2019,40(11):29-34. doi: 10.3969/j.issn.1672-8785.2019.11.006 LI Zhenxing. Study on wire sawing technology of semiconductor crystal [J]. Infrared,2019,40(11):29-34. doi: 10.3969/j.issn.1672-8785.2019.11.006
[4]	方莉俐, 刘韩, 姜羽飞. 线锯用金刚石微粉表面化学镀镍工艺及性能研究 [J]. 工具技术,2024,58(4):54-58. doi: 10.3969/j.issn.1000-7008.2024.04.009 FANG Lili, LIU Han, JIANG Yufei. Research on electroless nickel plating process and performance of diamond powder surface for wire saws [J]. Tool Engineering,2024,58(4):54-58. doi: 10.3969/j.issn.1000-7008.2024.04.009
[5]	张凤林, 袁慧, 周玉梅, 等. 硅片精密切割多线锯研究进展 [J]. 金刚石与磨料磨具工程,2006(6):14-18. doi: 10.3969/j.issn.1006-852X.2006.06.005 ZHANG Fenglin, YUAN Hui, ZHOU Yumei, et al. Progress of multi-wire saw for precision slicing of silicon wafer [J]. Diamond & Abrasives Engineering,2006(6):14-18. doi: 10.3969/j.issn.1006-852X.2006.06.005
[6]	常龙. 多线切割机高线速下振动特性及其优化控制研究 [D]. 秦皇岛: 燕山大学, 2023. CHANG Long. Research on vibration characteristics and optimal control of multi-wire saw at high wire speed [D]. Qinhuangdao: Yanshan University, 2023.
[7]	李玉伟, 郑坤坤, 卞林芝, 等. 一种电镀金刚线及其制备方法: CN202211516260.5 [P]. 2022-11-30. LI Yuwei, ZHENG Kunkun, BIAN Linzhi, et al. An electroplated diamond wire and its preparation method: CN202211516260.5 [P]. 2022-11-30.
[8]	黄世玲. 金刚石化学镀镍工艺研究及电化学分析 [D]. 郑州: 郑州大学, 2014. HUANG Shiling. Research and electrochemical analysis of electroless nickel plating on diamond [D]. Zhengzhou: Zhengzhou University, 2014.
[9]	RIEDEL W. Electroless nickel plating [M]. England: Finishing Publication Ltd, 1991.
[10]	周琼宇, 谢蔚, 王小芬, 等. 基于人工神经网络预测Ni-W合金镀层的硬度和耐腐蚀性能 [J]. 表面技术,2016,45(12):140-146. doi: 10.16490/j.cnki.issn.1001-3660.2016.12.023 ZHOU Qiongyu, XIE Wei, WANG Xiaofen, et al. Artificial neural network-based prediction of hardness and corrosion resistance of Ni-W alloy coating [J]. Surface Technology,2016,45(12):140-146. doi: 10.16490/j.cnki.issn.1001-3660.2016.12.023
[11]	LIU X, TIAN S, TAO F, et al. A review of artificial neural networks in the constitutive modeling of composite materials [J]. Composites Part B: Engineering,2021(1):109152.1-109152.15. doi: 10.1016/j.compositesb.2021.109152
[12]	JANG D P, FAZILY P, YOON J W. Machine learning-based constitutive model for J2-plasticity [J]. International Journal of Plasticity,2021,138(1):102919. doi: 10.1016/j.ijplas.2020.102919
[13]	WU Y, SHEN B, LUI L, et al. Artificial neural network modelling of plating rate and phosphorus content in the coatings of electroless nickel plating [J]. Journal of Materials Processing Technology,2008,205(1/2/3):207-213. doi: 10.1016/j.jmatprotec.2007.11.095
[14]	郭宝会, 邱友绪, 李海龙. 人工神经网络在钛合金表面Ni-SiC复合电镀工艺中的应用 [J]. 中国腐蚀与防护学报,2017,37(4):389-394. doi: 10.11902/1005.4537.2016.064 GUO Baohui, QIU Youxu, LI Hailong. Application of artificial neural network for preparation process of Ni-SiC composite coatings on Ti-alloy [J]. Journal of Chinese Society for Corrosion and Protection,2017,37(4):389-394. doi: 10.11902/1005.4537.2016.064
[15]	邓羽, 张杰, 彭中波, 等. 基于人工神经网络的Ni-ZrO₂纳米镀层耐腐蚀性能预测 [J]. 装备环境工程,2022,19(2):98-105. doi: 10.7643/issn.1672-9242.2022.02.016 DENG Yu, ZHANG Jie, PENG Zhongbo, et al. Prediction of corrosion resistance of Ni-ZrO₂ nano-plating based on artificial neural network [J]. Equipment Environmental Engineering,2022,19(2):98-105. doi: 10.7643/issn.1672-9242.2022.02.016
[16]	陈佳兵, 吴自银, 赵荻能, 等. 基于粒子群优化算法的PSO-BP海底声学底质分类方法 [J]. 海洋学报,2017,39(9):51-57. doi: 10.3969/j.issn.0253-4193.2017.09.005 CHEN Jiabing, WU Ziyin, ZHAO Dineng, et al. Back propagation neural network classification of sediment seabed acoustic sonar images based on particle swarm optimization algorithms [J]. Acta Oceanologica Sinica, 2017, 39(9): 51-57. doi: 10.3969/j.issn.0253-4193.2017.09.005
[17]	HAYAJNEH M, HASSAN A M, ALRASHDAN A, et al. Prediction of tribological behavior of aluminum–copper based composite using artificial neural network [J]. Journal of Alloys & Compounds,2009,470(1/2):584-588. doi: 10.1016/j.jallcom.2008.03.035
[18]	XIAO L, WANG J, HOU R, et al. A combined model based on data pre-analysis and weight coefficients optimization for electrical load forecasting [J]. Energy,2015,82:524-549. doi: 10.1016/j.energy.2015.01.063
[19]	LIN Y H, GU S, WU W S, et al. Analysis and prediction of overloaded extra-heavy vehicles for highway safety using machine learning [J]. Mobile Information Systems, 2020(1):1-20. doi: 10.1155/2020/6667897
[20]	简益梅, 许承东, 王倚文, 等. 基于改进GRNN的电离层VTEC误差模型 [J]. 计算机仿真,2022,39(8):45-50. doi: 10.3969/j.issn.1006-9348.2022.08.009 JIAN Yimei, XU Chengdong, WANG Yiwen, et al. Ionospheric VTEC error model based on improved GRNN [J]. Computer Simulation,2022,39(8):45-50. doi: 10.3969/j.issn.1006-9348.2022.08.009
[21]	侯克鹏, 包广拓, 孙华芬. 改进的MVO-GRNN神经网络岩爆预测模型研究 [J]. 安全与环境学报,2024,24(3):923-932. doi: 10.13637/j.issn.1009-6094.2023.0341 HOU Kepeng, BAO Guangtuo, SUN Huafen. Research on improved MVO-GRNN neural network rockburst prediction model [J]. Journal of Safety and Environment,2024,24(3):923-932. doi: 10.13637/j.issn.1009-6094.2023.0341
[22]	吴荣华, 王江安, 任席闯, 等. 不同波长近红外激光大气消光特性的泛化回归神经网络反演算法 [J]. 红外与毫米波学报,2010,29(6):461-464. WU Ronghua, WANG Jiangan, REN Xichuang, et al. Real time meafurement of atmospheric optical properties [J]. Journal of Infrared and Millimeter Waves,2010,29(6):461-464.
[23]	张宝磊, 熊艺文, 王为庆, 等. 高速铣削TC4表面粗糙度预测模型研究 [J]. 组合机床与自动化加工技术,2015,(3):108-110. doi: 10.13462/j.cnki.mmtamt.2015.03.029 ZHANG Baolei, XIONG Yiwen, WANG Weiqing, et al. Research on surface roughness prediction model for high-speed milling TC4 [J]. Modular Machine Tool and Automatic Machining Technique,2015,(3):108-110. doi: 10.13462/j.cnki.mmtamt.2015.03.029