动态模糊RBF神经网络焊接接头力学性能预测建模

张永志; 董俊慧; 朱红玲

doi:10.11868/j.issn.1005-5053.2016.5.005

动态模糊RBF神经网络焊接接头力学性能预测建模

doi: 10.11868/j.issn.1005-5053.2016.5.005

1.
内蒙古农业大学机电工程学院, 呼和浩特 010018
2.
内蒙古工业大学材料科学与工程学院, 呼和浩特 010051
3.
内蒙古大学创业学院商学教学部, 呼和浩特 010070

基金项目:

国家自然科学基金 51165027

详细信息

通讯作者:
董俊慧(1961-), 男, 博士, 教授, 主要从事焊接结构强度与断裂、焊接应力与变形、新材料开发与连接技术、模糊神经网络在焊接中的应用等方面的研究, (E-mail)jhdong1009@163.com

中图分类号: TG407
计量
- 文章访问数: 2658
- HTML全文浏览量: 1240
- PDF下载量: 13
- 被引次数: 0
出版历程
- 收稿日期: 2016-01-11
- 修回日期: 2016-02-18
- 刊出日期: 2016-10-01

Predictive Modeling of Mechanical Properties of Welded Joints Based on Dynamic Fuzzy RBF Neural Network

1.
College of Mechanical and Electrical Engineering, Inner Mongolia Agricultural University, Hohhot 010018, China
2.
School of Materials Science and Engineering, Inner Mongolia University of Technology, Hohhot 010051, China
3.
Business Studies Teaching Department, Inner Mongolia University Business School, Hohhot 010070, China

摘要

摘要: 建立动态模糊径向基神经网络RBF（Radial Basis Function, RBF）焊接接头力学性能预测模型, 克服静态RBF和模糊神经网络（Fuzzy Neural Network, FNN）在结构辨识、动态样本训练及学习算法的不足。该模型的结构参数不再提前预设, 在训练过程中动态自适应调整, 适用动态样本数据学习, 学习算法引入分级学习和模糊规则修剪策略, 加速训练并使模型结构更加紧凑。利用三种厚度、不同工艺TC4钛合金TIG焊接试验数据对该模型进行仿真。结果表明：模型具有较高的预测精度, 适用于预测焊接接头力学性能, 为焊接过程在线控制开辟了新的途径。
- 动态模糊RBF神经网络 /
- 预测 /
- 焊接 /
- 建模 /
- 力学性能
Abstract: A dynamic fuzzy RBF neural network model was built to predict the mechanical properties of welded joints, and the purpose of the model was to overcome the shortcomings of static neural networks including structural identification, dynamic sample training and learning algorithm. The structure and parameters of the model are no longer head of default, dynamic adaptive adjustment in the training, suitable for dynamic sample data for learning, learning algorithm introduces hierarchical learning and fuzzy rule pruning strategy, to accelerate the training speed of model and make the model more compact. Simulation of the model was carried out by using three kinds of thickness and different process TC4 titanium alloy TIG welding test data. The results show that the model has higher prediction accuracy, which is suitable for predicting the mechanical properties of welded joints, and has opened up a new way for the on-line control of the welding process.
- dynamic fuzzy RBF neural network /
- prediction /
- welding /
- modelling /
- mechanical property

HTML全文

图 1 D-FRBFNN的结构

Figure 1. Structure of D-FRBFNN

下载: 全尺寸图片幻灯片

图 2 D-FRBFNN模糊规则数曲线

Figure 2. Fuzzy rules curve of D-FRBFNN

下载: 全尺寸图片幻灯片

图 3 训练完成后的均方根误差

Figure 3. RMSE after training completed

下载: 全尺寸图片幻灯片

表 1 TC4钛合金化学成分(质量分数/%)

Table 1. Chemical composition of TC4 titanium alloy (mass fraction/%)

C	O	H	N	Al	V	Fe	Si	Ti
0.10	0.15	0.015	0.05	5.5-6.8	3.5-4.5	0.30	0.15	Bal.

下载: 导出CSV

表 2 D-FRBFNN模型初始化参数

Table 2. Initialization parameters of D-FRBFNN model

Parameter	Value	Meaning
d_max	27.200	The maximum Euclidean distance between the input variables
d_min	0.300	The minimum length of the modeling
β	0.300	The boundary of convergence constant
γ	0.350	Error attenuation coefficient
e_max	10.000	Predefined maximum system error
e_min	0.200	Desired accuracy of the model
σ₀	13.600	The first radial unit width
k	1.050	Coefficient overlapping of radial basis units
k_w	10.000	The adjustment coefficient of Radial Basis unit width
k_err	0.002	Threshold of importance fuzzy rules

下载: 导出CSV

表 3 D-FRBFNN模型焊接接头力学性能仿真结果

Table 3. Simulation results of D-FRBFNN model of mechanical properties of welded joints

No	Thickness/mm	Welding voltage/V	Weldingcurrent/A	Arg gas flow/(L·min^-1)	Welding speed/(m·h^-1)	Tensile strength/MPa		Elongation/%		Weld hardness/HRC		HAZ hardness/HRC
No	Thickness/mm	Welding voltage/V	Weldingcurrent/A	Arg gas flow/(L·min^-1)	Welding speed/(m·h^-1)	Experi-ment	Prediction	Experi-ment	Prediction	Experi-ment	Prediction	Experi-ment	Prediction
1	2.0	15	0	8.0	14.0	841	780	4.0	4.2	38.5	42.3	42.0	28.5
2	2.0	15	0	7.0	15.0	881	847	7.0	6.1	43.7	35.4	44.3	38.6
3	2.5	11	0	9.0	11.0	901	756	2.5	1.8	37.1	39.7	36.9	36.4
4	2.5	11	0	7.0	16.0	880	736	2.4	3.0	40.5	33.9	35.9	34.9
5	3.0	18	0	15.0	7.5	901	911	1.8	2.1	44.2	44.9	38.4	38.8

下载: 导出CSV

参考文献(16)

[1]	ADEL M, MOHAMED B.Sizing of stand-alone photo voltaic systems using neural network adaptive model[J].Desalination, 2007, 209(6):64-72.
[2]	张永志, 董俊慧, 张艳飞.基于径向基神经网络焊接接头力学性能预测[J].焊接学报, 2008, 29(7):81-84. http://www.cnki.com.cn/Article/CJFDTOTAL-HJXB200807020.htm ZHANG Y Z, DONG J H, ZHANG Y F.Prediction of mechanical properties of titanium alloy welding joints based on RBF neural network[J].Transactions of the China Welding Institution, 2008, 29(7):81-84. http://www.cnki.com.cn/Article/CJFDTOTAL-HJXB200807020.htm
[3]	唐正魁, 董俊慧, 张永志.基于RBF神经网络预测接头力学性能的优化算法研究[J].材料科学与工艺, 2009, 17(增刊2):93-97. http://www.doc88.com/p-409566150701.html TANG Z K, DONG J H, ZHANG Y Z.Research on optimization based of RBF neural network to predict mechanical properties of joints[J].Materials Science and Technology, 2009, 17 (Suppl 2):93-97. http://www.doc88.com/p-409566150701.html
[4]	张永志, 董俊慧.两种预测焊接接头力学性能的模糊神经网络[J].焊接学报, 2011, 32(11):104-107. http://www.cnki.com.cn/Article/CJFDTOTAL-HJXB201111028.htm ZHANG Y Z, DONG J H.Research on two fuzzy neural networks to predict mechanical properties of welded joints[J].Transactions of the China Welding Institution, 2011, 32(11):104-107. http://www.cnki.com.cn/Article/CJFDTOTAL-HJXB201111028.htm
[5]	张永志, 董俊慧.基于FCM聚类的模糊RBF神经网络预测焊接接头力学性能建模[J].机械工程学报, 2014, 50(12):58-64. doi: 10.3901/JME.2014.12.058 ZHANG Y Z, DONG J H.Modeling fuzzy RBF neural network to predict of mechanical properties of welding joints based on fuzzy C-meaning cluster[J].Journal of Mechanical Engineering, 2014, 50(12):58-64. doi: 10.3901/JME.2014.12.058
[6]	吕其兵, 谭克利, 骆德阳, 等.基于RBF神经网络的钢轨交流闪光焊接头灰斑面积预测[J].焊接学报, 2008, 29(2):93-99. http://www.cnki.com.cn/Article/CJFDTOTAL-HJXB200802025.htm LYU Q B, TAN K L, LUO D Y, et al.Prediction of area of gray-spots flaw in alternate rail flash butt welded joint based on RBF neural network[J].Transactions of the China Welding Institution, 2008, 29(2):93-99. http://www.cnki.com.cn/Article/CJFDTOTAL-HJXB200802025.htm
[7]	王清, 那月, 孙东立, 等.GH99合金TIG焊接接头拉伸性能的人工神经网络预测[J].焊接学报, 2010, 31(3):77-82. http://www.cnki.com.cn/Article/CJFDTOTAL-HJXB201003022.htm WANG Q, NA Y, SUN D L, et al.Prediction of tensile property of TIG welding joints in GH99 alloy by artificial neural network[J].Transactions of the China Welding Institution, 2010, 31(3):77-82. http://www.cnki.com.cn/Article/CJFDTOTAL-HJXB201003022.htm
[8]	张旭明, 吴毅雄, 徐滨士, 等.BP神经网络及其在焊接中的应用[J].焊接, 2003, 2:9-14. http://www.cnki.com.cn/Article/CJFDTOTAL-HAJA200302003.htm ZHANG X M, WU Y X, XU B S, et al.BP neural network and its application to welding [J].Welding, 2003(2):9-14. http://www.cnki.com.cn/Article/CJFDTOTAL-HAJA200302003.htm
[9]	FATHI A, AGHAKOUCHAK A A.Prediction of fatigue crack growth rate in welded tubular joints using neural network[J].International Journal of Fatigue, 2007, 29:261-275. doi: 10.1016/j.ijfatigue.2006.03.002
[10]	WU S Q, MENG J E.Dynamic fuzzy neural networks—a novel approach to function approximation[J].IEEE Transactions on Systems, Man and Cybernetics B, 2000, 30(2):358-364. doi: 10.1109/3477.836384
[11]	倪志盛, 王明彦.基于动态模糊神经网络的多余力矩抑制方法[J].哈尔滨工业大学学报, 2012, 44(10):79-83. http://www.cnki.com.cn/Article/CJFDTOTAL-HEBX201210016.htm NI Z S, WANG M Y.A novel method for restraining the redundancy torque based on DFNN[J].Journal of Harbin Institute of Technology, 2012, 44(10):79-83. http://www.cnki.com.cn/Article/CJFDTOTAL-HEBX201210016.htm
[12]	马草原, 刘建峰, 毛家松, 等.基于D-FNN的电网混沌振荡系统的自适应控制[J].电测与仪表, 2012, 49(4):51-54. http://www.cnki.com.cn/article/cjfdtotal-dcyq201204015.htm MA C Y, LIU J F, MAO J S, et al.Adaptive control of power grid chaotic oscillation system based on dynamic fuzzy neural network[J].Electrical Measurement & Instrumentation, 2012, 49(4):51-54. http://www.cnki.com.cn/article/cjfdtotal-dcyq201204015.htm
[13]	WU S Q, MENG J E, GAO Y.A fast approach for automatic generation of fuzzy rules by generalized dynamic fuzzy neural networks[J].IEEE Transactions on Fuzzy Systems, 2001, 9(4):578-594. doi: 10.1109/91.940970
[14]	伍世虔, 徐军.动态模糊神经网络——设计与应用[M].北京:清华大学出版社, 2008:27-36. WU S Q, XU J.Dynamic fuzzy neural network design and application[M].Beijing:Tsinghua University Press, 2008:27-36.
[15]	CHEN S, COWAN C, GRANT P.Orthogonal least squares learning algorithm for radial basis function network[J].IEEE Transactions on Neural Networks, 1991, 2(2):302-309. doi: 10.1109/72.80341
[16]	DAYA R, LAXMI S, MANJAREE P, et al.Corrective action planning using RBF neural network[J].Applied Soft Computing, 2007, 7:1055-1063. doi: 10.1016/j.asoc.2006.10.007