Journal of Textile Research ›› 2019, Vol. 40 ›› Issue (8): 157-163.doi: 10.13475/j.fzxb.20180606307

• Machinery & Accessories • Previous Articles     Next Articles

Prediction method of integrated piercing pressure parameters based on machine learning

YANG Jingzhao1, JIANG Xiuming1,2(), DONG Jiuzhi1,2, CHEN Yunjun2,3, MEI Baolong1   

  1. 1. School of Mechanical Engineering, Tianjin Polytechnic University, Tianjin 300387, China
    2. Advanced Mechatronics Equipment Technology Tianjin Area Major Laboratory, Tianjin Polytechnic University, Tianjin 300387, China
    3. School of Electrical Engineering and Automation, Tianjin Polytechnic University, Tianjin 300387, China
  • Received:2018-06-21 Revised:2019-04-04 Online:2019-08-15 Published:2019-08-16
  • Contact: JIANG Xiuming E-mail:jxjxm@126.com

Abstract:

In view of the problem that the spring back of carbon cloth leads to a large fluctuation range of average layer height and thus affects the performance of three-dimensional fabric during pressurized compaction process of integrated piercing, a real-time prediction method of pressure parameters based on machine learning theory was proposed. The complex modeling of the relationship between the average layer height and the pressurized parameters was transformed into multiple regression problems, and an unconstrained optimization iteration method suitable for computer operation was adopted to solve the problem. Based on the scikit-learn class library, the feature variables were selected. After comparing the predictive performance scores of the six regression models, the K nearest neighbor regression were selected as the base learners, and the prediction performance of the model was improved by using the integration algorithm. The experimental results after the prediction model was deployed to the production environment show that owing to the use of the machine learning prediction, the response speed of the pressure parameters to the average level in the integrated piercing process is improved, and the mean change amplitude is reduced. The average height fluctuation range of the experimental sample product is reduced from 12.0% to 6.8%, and the standard deviation from 0.008 3 to 0.006 6.

Key words: three-dimensional fabric, integrated piercing, control parameter prediction, machine learning

CLC Number: 

  • TP181

Fig.1

Schematic diagram of integrated piercing process"

Fig.2

Schematic diagram of pressurized compaction and integrated piercing"

Tab.1

Definition of modeling variables for machine learning"

含义 变量 说明
输入特征 x x=[x1,x2,,x11]
输出特征 y y=[x12]
训练集 D={(x(i),y(i));i=1,2,,m} i为样本序号
输入变量空间 X={x(i);i=1,2,,m} i为样本序号
输出变量空间 Y={y(i);i=1,2,,m} i为样本序号

Fig.3

Schematic diagram of supervised learning process"

Fig.4

Comparison diagram of under fitting and over fitting"

Fig.5

Multi models fitting diagram with full features"

Tab.2

MSE score table of multiple samples based on three models"

模型 性能度量标准 得分
线性回归 MSE(均方误差) 11.599
5阶多项式回归 MSE(均方误差) 3.935
30阶多项式回归 MSE(均方误差) 0.164

Fig.6

Schematic diagram of machine learning process"

Tab.3

Label encoding of data features"

特征类型名称 特征值 标签编码
助剂类型 A 0
B 1
碳布类别 I 0
II 1
III 2
IV 3
加压间隔 2层 0
4层 1
6层 2
8层 3

Tab.4

One-Hot encoding of data features"

特征类型名称 特征值 特征编码
助剂类型 A 10
B 01
碳布类型 I 1000
II 0100
III 0010
IV 0001
加压间隔 2层 1000
4层 0100
6层 0010
8层 0001

Tab.5

Removing data features with low variance"

特征名称 特征说明 方差 是否保留
CT-1 碳布类型独热编码特征1 0.237
CT-2 碳布类型独热编码特征2 0.140
CT-3 碳布类型独热编码特征3 0.140
CT-4 碳布类型独热编码特征4 0.140
AT-1 助剂类型独热编码特征1 0.140
AT-2 助剂类型独热编码特征2 0.140
PI-1 加压间隔独热编码特征1 0.300
PI-2 加压间隔独热编码特征2 0.400
PI-3 加压间隔独热编码特征3 0.458
PI-4 加压间隔独热编码特征4 0.489

Tab.6

Correlation degree classification table of Pearson coefficient"

绝对值范围 相关性程度 是否保留
0.800~1.000 极强相关
0.600~0.800 强相关
0.400~0.600 中等强度相关
0.200~0.400 弱相关
0.000~0.200 极弱相关或无相关

Tab.7

Processing results of univariate feature selection"

特征 皮尔逊系数 F回归 是否保留
加压循环 -0.946 42 897.173
当前层数 0.981 130 261.670
压力增加 0.093 43.738
时间增加 0.014 1.003
高度增加 -0.704 4 915.491
当前高度 0.981 126 421.006
平均高度 -0.265 377.436
加压时间 0.999 2 068 838.071

Tab.8

Evaluation matrices of six models under three kinds of criteria"

评价
准则
回归
模型
得分
平均值
得分
标准差
是否
保留
MAE 线性回归 -3.280×10-2 2.000×10-5
MAE 岭回归 -3.235×10-2 1.000×10-5
MAE 套索回归 -8.695×10-1 0
MAE 弹性网络回归 -4.992×10-1 2.400×10-4
MAE K近邻回归 -2.117×10-2 1.880×10-3
MAE 支持向量机回归 -4.186×10-2 2.620×10-3
MSE 线性回归 -1.180×10-3 0
MSE 岭回归 -1.180×10-3 0
MSE 套索回归 -1 0
MSE 弹性网络回归 -3.276×10-1 4.200×10-4
MSE K近邻回归 -6.100×10-4 1.100×10-4
MSE 支持向量机回归 -2.530×10-3 3.000×10-4
R2 线性回归 9.988×10-1 0
R2 岭回归 9.988×10-1 0
R2 套索回归 0 0
R2 弹性网络回归 6.724×10-1 4.200×10-4
R2 K近邻回归 9.994×10-1 1.100×10-2
R2 支持向量机回归 9.975×10-1 3.000×10-4

Fig.7

Box Line chart of two models under three kinds of criteria"

Tab.9

Improved results of model prediction performance using integrated algorithms"

评价
准则
基模型 得分
平均值
得分
标准差
MAE K近邻回归 -2.092×10-2 3.140×10-3
MSE K近邻回归 -5.900×10-4 1.200×10-4
R2 K近邻回归 9.994×10-1 1.200×10-4

Tab.10

Process parameters of integrated piercing pressure prediction"

层数
区间
层数 加压间
隔/层
加压循
环/次
压力
范围/N
时间范
围/s
0~320 320 8 40 980~1 666 10~30
320~500 180 6 30 1 666~2 352 30~50
500~580 80 4 20 2 352~2 744 50~60
580~600 20 2 10 2 744~2 940 60~70

Fig.8

Parameter prediction system of integral piercing pressure. (a) System architecture;(b) Machine"

Fig.9

Comparison of average layer height before and after machine learning prediction model"

Fig.10

Comparison of average layer height before and after machine learning prediction model"

[1] ROINCIK P G . Properties and application of mod-3 pierced fabric reinforced carbon/carbon compo-sites[J]. Carbon, 1973,11(6):690.
[2] 朱建勋 . 细编穿刺织物的结构特点及性能[J]. 宇航材料工艺, 1998(1):41-43.
ZHU Jianxun . The structural characteristics and prop-erties of fine weave pierced fabric[J]. Aerospace Materials & Technology, 1998(1):41-43.
[3] 朱建勋 . 碳布整体穿刺织物编织工艺与结构参数优化[D]. 南京: 东南大学, 2004: 17-24.
ZHU Jianxun . Carbon fiber integrated piercing fabrics braiding technology and parameter optimization[D]. Nanjing: Southeast University, 2004: 17-24.
[4] 邢晓熙, 徐澄 . 台达PLC与触摸屏在细编穿刺机上的应用[J]. 机械设计与制造工程, 2013,42(2):84-85.
XING Xiaoxi, XU Cheng . The application of delta's PLC and touch screen in fine weave pierced ma-chine[J]. Machine Design and Manufacturing Engeering, 2013,42(2):84-85.
[5] 董九志, 蒋秀明, 杨建成 , 等. 立体织物整体穿刺钢针阵列布放装置研制及实验验证[J]. 纺织学报, 2015,36(3):115-120.
DONG Jiuzhi, JIANG Xiuming, YANG Jiancheng , et al. Development and experimental verification of steel needle array placement device for solid fabric punc-ture[J]. Journal of Textile Research, 2015,36(3):115-120.
[6] 陈盛洪 . 细编穿刺C/C复合材料热物理性能的模拟研究[D]. 哈尔滨: 哈尔滨工业大学, 2008: 53-56.
CHEN Shenghong . Simulation on the physical properties of fine weave pierced C/C composite[D]. Harbin: Harbin Institute of Technology, 2008: 53-56.
[7] 周钰博, 李艳霞, 李敏 , 等. 细编穿刺织物的数学建模理论与实例[J]. 材料工程, 2017,45(4):102-107.
ZHOU Yubo, LI Yanxia, LI Min , et al. Mathematical modeling theory and example of fine woven fabric[J]. Journal of Material Engineering, 2017,45(4):102-107.
[8] LEE S D, JUN C H . Fast incremental learning of logistic model tree using least angle regression[J]. Expert Systems with Applications, 2018(97):137-145.
[9] AMIN M J, RIZA N A . Machine learning enhanced optical distance sensor[J]. Optics Communications, 2018(407):262-270.
[10] ARUN M K, GHANSHYAM P, RAMPI R . Critical assessment of regression-based mahine learning methods for polymer dielectrics[J]. Computational Materials Science, 2016(125):123-135.
[11] SUNIL K J, ZULFIQAR A . Soil microbial dynamics prediction using machine learning regression methods[J]. Computers and Electronics in Agriculture, 2018(147):158-165.
[12] 朱建勋, 何建敏, 王海燕 , 等. 基于纤维弯曲伸长模式的Z向钢针针尖形态优化[J]. 中国工程科学, 2003,5(9):18-21.
ZHU Jianxun, HE Jianmin, WANG Haiyan , et al. Optimization of Z oriented needle tip shape based on fiber bending elongation model[J]. Strategic Study of CAE, 2003,5(9):18-21.
[13] 朱建勋, 何建敏, 王海燕 . 正交叠层机织布整体穿刺工艺的纤维弯曲伸长机理[J]. 中国工程科学, 2003,5(5):59-62.
ZHU Jianxun, HE Jianmin, WANG Haiyan . The mechanism of fiber bending and elongation in the Integrated piercing process of orthogonal laminated woven fabrics[J]. Strategic Study of CAE, 2003,5(5):59-62.
[14] KRISHNAN N M A, MANGALATHU S, SMEDSKJAER M M , et al. Predicting the dissolution kinetics of silicate glasses using machine learning solids[J]. Journal of Non-Crystalline Solids, 2018(487):37-45.
[15] PENG Xinjun, CHEN De . PTSVRs:regression models via projection twin support vector machine[J]. Information Sciences, 2018(435):1-14.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
[1] . [J]. JOURNAL OF TEXTILE RESEARCH, 2004, 25(01): 48 -49 .
[2] . [J]. JOURNAL OF TEXTILE RESEARCH, 1983, 4(10): 45 -46 .
[3] HUANG Chen;ANGLi′na;AN Xiaojian;U Zhaofang;E Yindi. Preparation and characterization of cotton nonwovens dipped by sericin[J]. JOURNAL OF TEXTILE RESEARCH, 2007, 28(11): 52 -55 .
[4] XU Yun-hui;LIN Hong;CHEN Yu-yue . Aggregated structure of cotton fiber oxidised selectively with sodium periodate[J]. JOURNAL OF TEXTILE RESEARCH, 2006, 27(11): 1 -5 .
[5] . [J]. JOURNAL OF TEXTILE RESEARCH, 2002, 23(03): 32 -33 .
[6] . [J]. JOURNAL OF TEXTILE RESEARCH, 1986, 7(06): 34 -38 .
[7] . [J]. JOURNAL OF TEXTILE RESEARCH, 2001, 22(05): 25 -26 .
[8] . [J]. JOURNAL OF TEXTILE RESEARCH, 2001, 22(05): 60 -62 .
[9] ZHANG Jianfei;SHAO Gaiqin. Effect of ultraviolet on Phanerochaete chrysosporium for decomposing caprolactam and degrading polyamide 6[J]. JOURNAL OF TEXTILE RESEARCH, 2008, 29(1): 9 -13 .
[10] . [J]. JOURNAL OF TEXTILE RESEARCH, 1990, 11(07): 4 -7 .