基于机器学习的开孔加载金属腔电磁屏蔽效能评估

刘筝阳; 闫丽萍; 赵翔

doi:10.11884/HPLPB201931.190079

基于机器学习的开孔加载金属腔电磁屏蔽效能评估

doi: 10.11884/HPLPB201931.190079

四川大学电子信息学院, 成都 610065

基金项目:

国家自然科学基金项目 61877041

详细信息

作者简介:
刘筝阳(1993—), 男, 硕士研究生, 从事电磁兼容方面研究, sculiuzhengyang@163.com

通讯作者:
闫丽萍(1972—), 女, 教授, 主要从事电磁兼容建模分析、电磁效应评估研究, liping_yan@scu.edu.cn

中图分类号: O441.4
计量
- 文章访问数: 1376
- HTML全文浏览量: 309
- PDF下载量: 89
- 被引次数: 0
出版历程
- 收稿日期: 2019-03-22
- 修回日期: 2019-05-27
- 刊出日期: 2019-08-15

Evaluation of electromagnetic shielding effectiveness for loaded metallic enclosures with apertures based on machine learning

School of Electronics and Information Engineering, Sichuan University, Chengdu 610065, China

摘要

摘要: 利用全波分析方法计算了不同电路板加载、不同孔缝和尺寸的开孔金属腔在0!5GHz范围内的屏蔽效能(SE), 获得共计5250个样本。进而利用机器学习中的随机森林回归算法, 对其中4200个样本数据进行训练, 获得了可以根据开孔腔物理尺寸、加载物材料及电磁特性和位置、频率等共计16个输入参数快速评估开孔加载金属腔屏蔽效能的机器学习模型。利用其余的1050个样本进行模型验证, 结果表明该模型可以快速准确地计算加载腔的电磁屏蔽效能。该模型具有随时根据样本量增加不断训练提高其普适性的特点, 可为实际工程中加载开孔腔的屏蔽设计及SE评估提供高效途径。
- 加载金属腔 /
- 屏蔽效能 /
- 随机森林 /
- 机器学习
Abstract: A machine learning based evaluation method for shielding effectiveness (SE) of loaded metallic enclosures with apertures under electromagnetic wave radiation is proposed.The SEs of a variety of metallic enclosures loaded with different printed circuit boards (PCBs) is calculated using full wave analysis simulation in the frequency range of 0-5 GHz, and 5250 samples are obtained.The random forest model which is one of the popular machine learning aggression algorithms is employed to train stochastically the selected 4200 samples.Consequently, the model capable to fast predict the SE for loaded shielding enclosures characterized by 16 parameters is implemented.The rest 1050 samples are used to verify the proposed random forest model.Resultsshow that the proposed model can quickly predict the electromagnetic shielding effectiveness of the enclosure loaded with PCBs.
- loaded metallic enclosures /
- shielding effectiveness /
- random forest /
- machine learning

HTML全文

图 1 开孔加载腔示意图

Figure 1. Schematic diagram of loaded metallic enclosure with apertures

下载: 全尺寸图片幻灯片

图 2 不同加载金属腔中心点屏蔽效能

Figure 2. Shielding effectivenese (SE) at center of loaded metallic enclosure

下载: 全尺寸图片幻灯片

图 3 随机森林算法的流程图

Figure 3. Flow chart of random forest algorithm

下载: 全尺寸图片幻灯片

图 4 机器学习模型预测的SE与全波分析计算结果的对比

Figure 4. Comparison of SE predicted by machine learning model to SE calculated using full wave analysis

下载: 全尺寸图片幻灯片

表 1 加载有耗介质的尺寸及位置

Table 1. Dimension size and position of loaded lossy substrates

	a_s/mm	b_s/mm	c_s/mm	center point of PCB
load 1	100	10	100	(150, 70, 60)
load 2	10	100	10	(115, 60, 175)
load 3	10	100	100	(105, 60, 150)
load 4	100	20	100	(150, 60, 150)
load 5	20	100	100	(150, 60, 150)
load 6	100	100	100	(150, 60, 150)

下载: 导出CSV

表 2 各物理量的参数变化范围

Table 2. Value variation range of physical parameters

a_e，b_e，c_e	a_s，b_s，c_s	σ/(S·m^-1)	ε_r	r/mm	d/mm	N_x，N_y	x_s，y_s，z_s	f/GHz
10~50 cm	10~20 cm	0~50	1~4	5 ~8	5 ~10	1~10	varies in cavity	0~5

下载: 导出CSV

表 3 部分样本的参数及SE值

Table 3. Parameters value and SE of some samples

a_e /mm	b_e /mm	c_e /mm	r /mm	d /mm	a_s /mm	c_s /mm	b_s /mm	N_x	N_y	x_s	y_s	z_s	σ /(S·m^-1)	ε_r	f /GHz	SE_min /dB	SE_mean /dB
120	300	440	7.5	5	100	150	5	3	5	70	42.5	125	0.1	4	0.1	31.5	62.3
200	120	300	7.5	4	100	80	5	4	5	60	32.5	90	0.3	4	0.2	44.7	70.9
200	400	500	5.0	5	100	200	5	1	5	70	62.5	120	0.9	3	5	20.4	34.0
250	200	250	7.5	5	200	100	10	3	5	70	105	120	21	4	0.9	32.5	51.4
300	360	150	5.0	5	150	100	5	4	7	125	102.5	70	1	4	3.1	16.9	27.0
100	150	200	7.5	5	70	150	5	1	5	55	72.5	95	0.4	4	2.4	25.4	45.4
100	150	200	7.5	5	70	150	5	4	5	55	72.5	95	0.4	4	3	20.8	31.4
150	200	260	7.5	5	100	150	5	2	5	70	82.5	95	0.3	4	1.8	31.3	42.1
150	200	260	7.5	5	100	150	5	4	5	70	82.5	95	0.3	4	4	21.4	3.1
160	300	100	7.5	10	100	80	5	1	5	70	62.5	50	0.3	4	2.8	28.5	42.1

下载: 导出CSV

表 4 验证模型的参数取值

Table 4. Parameters value of the loaded metallic enclosure for validation

a_e /mm	b_e /mm	c_e /mm	r /mm	d /mm	a_s /mm	c_s /mm	b_s /mm	N_x	N_y	x_s	y_s	z_s	σ /(S·m^-1)	ε_r	f /GHz
240	320	460	5	6	80	100	120	5	7	140	150	160	0.6	2.6	0~5

下载: 导出CSV

表 5 样本量对随机森林模型训练精度的影响

Table 5. Effect of sample number on accuracy of radom forest model

sample number	RMSE/dB		R²
sample number	training set	testing set	training set	testing set
2000	0.478	1.185	0.997	0.982
3000	0.382	0.953	0.998	0.988
4200	0.319	0.840	0.998	0.990

下载: 导出CSV

参考文献(17)

[1]	王楠, 陈贵齐, 赵延安, 等. 复杂城市环境电磁态势的场数值分析与测量对比[J]. 电波科学学报, 2018, 33(6): 671-676. https://www.cnki.com.cn/Article/CJFDTOTAL-DBKX201806006.htm Wang Nan, Chen Guiqi, Zhao Yan'an, et al. Numerical analysis and measured comparison on electromagnetic problems in complex city environments. The Chinese Journal of Radio Science, 2018, 33(6): 671-676 https://www.cnki.com.cn/Article/CJFDTOTAL-DBKX201806006.htm
[2]	王天乐, 闫丽萍, 赵翔, 等. 包含非线性组件的系统级电磁效应分析方法[J]. 强激光与粒子束, 2014, 26: 073204. doi: 10.11884/HPLPB201426.073204 Wang Tianle, Yan Liping, Zhao Xiang, et al. System-level analysis method of electromagnetic effects on an electronic system containing nonlinear components[J]. High Power Laser and Particle Beams, 2014, 26: 073204 doi: 10.11884/HPLPB201426.073204
[3]	IEEE Standard 299, Method for measuring the shielding effectiveness of enclosures and boxes having all dimensions between 0.1 m and 2 m[S].
[4]	孙勇, 刘强, 闫丽萍, 等. 开孔金属腔屏蔽效能快速算法适用性评估[J]. 强激光与粒子束, 2017, 29: 083203. doi: 10.11884/HPLPB201729.170022 Sun Yong, Liu Qiang, Yan Liping, et al. Applicability evaluation of fast algorithm for shielding effectiveness prediction of metallic enclosures with apertures. High Power Laser and Particle Beams, 2017, 29: 083203 doi: 10.11884/HPLPB201729.170022
[5]	安静, 王泽锴, 韩承江, 等. 孔缝对内置带MSL电路板壳体耦合特性研究[J]. 微波学报, 2016, 32(1): 36-40. https://www.cnki.com.cn/Article/CJFDTOTAL-WBXB201601009.htm An Jing, Wang Zekai, Han Chen jiang, et al. Study of slot on the coupling characteristics for the shelled of loaded PCB with MSL. Journal of Microwaves, 2016, 32(1): 36-40 https://www.cnki.com.cn/Article/CJFDTOTAL-WBXB201601009.htm
[6]	路宏敏, 刘国强, 余志勇, 等. 加装印刷电路板的圆孔阵矩形机壳屏蔽效能[J]. 强激光与粒子束, 2009, 21(1): 108-112 http://www.hplpb.com.cn/article/id/3841 Lu Hongmin, Liu Guoqiang, Yu Zhiyong, et al. Shielding effectiveness of PCB loaded rectangular enclosure with circular-aperture array. High Power Laser and Particle Beams, 2009, 21(1): 108-112 http://www.hplpb.com.cn/article/id/3841
[7]	汪柳平, 高攸纲, 沈远茂, 等. 装有PCB有孔矩形腔屏蔽效能的传输线法分析[J]. 电波科学学报, 2008, 23(4): 740-744. doi: 10.3969/j.issn.1005-0388.2008.04.029 Wang Liuping, Gao Yougang, Shen Yuanmao, et al. Analysis of shielding effectiveness of rectangular cavity of loaded PCB with aperture by transmission line method[J]. The Chinese Journal of Radio Science, 2008, 23(4): 740-744 doi: 10.3969/j.issn.1005-0388.2008.04.029
[8]	郝建红, 公延飞, 范杰清, 等. 一种内置条状金属板的双层金属腔体屏蔽效能的理论模型[J]. 物理学报, 2016, 65: 044101. doi: 10.7498/aps.65.044101 Hao Jianhong, Gong Yanfei, Fan Jieqing, et al. An analytical model for shielding effectiveness of double layer rectangular enclosure with inner strip-shaped metallic plate. Acta Physica Sinica, 2016, 65: 044101 doi: 10.7498/aps.65.044101
[9]	尚宇炜, 马钊, 彭晨阳, 等. 内嵌专业知识和经验的机器学习方法探索(二): 引导学习的应用与实践[J]. 中国电机工程学报, 2017, 37(20): 5852-5861. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGDC201720003.htm Shang Yuwei, Ma Zhao, Peng Chenyang, et al. Study of a novel machine learning method embedding expertise(part Ⅱ): Applications and practices of guiding learning. Proceedings of the Chinese Society for Electrical Engineering. 2017, 37(20): 5852-5861 https://www.cnki.com.cn/Article/CJFDTOTAL-ZGDC201720003.htm
[10]	Medico R, Lambrecht N, Pues H, et al. Machine learning based error detection in transient susceptibility tests[J]. IEEE Trans Electromagnetic Compatibility, 2019, 61(2): 352-360. doi: 10.1109/TEMC.2018.2821712
[11]	Trinchero R, Manfredi P, Stievano I S, et al. Machine learning for the performance assessment of high-speed links[J]. IEEE Trans Electromagnetic Compatibility, 2018, 60(6): 1627-1634.
[12]	Br'eard A, Moulla R, Vollaire C. Metamodel of power electronic converters using learning SVR method coupling with wavelet compression[J]. IEEE Trans Electromagnetic Compatibility, 2016, 58(2): 588-598.
[13]	Lozano A, Robinson M P, Diaz A, et al. Evaluation and optimization of an equivalent model for printed circuit boards inside metallic enclosures[C]//XXIX General Assembly Int. Union Radio Science. 2008: EBp6.
[14]	Marvin A C, Dawson J F, Ward S, et al. A proposed new definition and measurement of the shielding effect of equipment enclosures[J]. IEEE Trans Electromagnetic Compatibility, 2004, 46(3): 459-468.
[15]	李航. 统计学习方法[M]. 北京: 清华大学出版社, 2012: 55-73. Li Hang. Statistical learning method. Beijing: Tsinghua University Press, 2012: 55-73
[16]	Breiman L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32.
[17]	Horvath L, Kokoszka P. A bootstrap approximation to a unit root test statistic for heavy-tailed observations[J]. Statistics and Probability Letters, 2003, 63(2): 163-173.