基于流形正则化域适应随机权神经网络的湿式球磨机负荷参数软测量

贺敏; 汤健; 郭旭琦; 阎高伟

doi:10.16383/j.aas.2018.c170289

基于流形正则化域适应随机权神经网络的湿式球磨机负荷参数软测量

doi: 10.16383/j.aas.2018.c170289

贺敏^1,,
汤健^2,,
郭旭琦^1,,
阎高伟^1, ,

1.
太原理工大学信息工程学院太原 030024
2.
北京工业大学信息部北京 100124

基金项目:

国家自然科学基金 61573364

山西省煤基重点科技攻关项目 MD 2014-07

国家自然科学基金 61450011

山西省自然科学基金 2015011052

详细信息

作者简介:
贺敏  太原理工大学硕士研究生.主要研究方向为工业大数据应用, 机器学习, 数据驱动软测量建模与迁移学习.E-mail:hemin0215@link.tyut.edu.cn

汤健  北京工业大学信息部教授, 博士, 主要研究方向为数据驱动软测量, 工业自动化控制.E-mail:tjian001@126.com

郭旭琦  太原理工大学硕士研究生, 主要研究方向为智能信息处理, 数据驱动软测量建模, 迁移学习.E-mail:guoxuqi0330@link.tyut.edu.cn

通讯作者:
阎高伟太原理工大学信息工程学院教授, 博士.主要研究方向为复杂工业控制系统, 智能控制理论及其应用, 智能信息处理等.本文通信作者.E-mail:yangaowei@tyut.edu.cn

计量
- 文章访问数: 2101
- HTML全文浏览量: 299
- PDF下载量: 533
- 被引次数: 0
出版历程
- 收稿日期: 2017-05-26
- 录用日期: 2017-08-29
- 刊出日期: 2019-02-20

Soft Sensor for Ball Mill Load Using DAMRRWNN Model

HE Min^1
,,
TANG Jian^2
,,
GUO Xu-Qi^1
,,
YAN Gao-Wei^{1
, ,}

1.
College of Information Engineering, Taiyuan University of Technology, Taiyuan 030024
2.
Information Department, Beijing University of Technology, Beijing 100124

Funds:

National Natural Science Foundation of China 61573364

Key Project of Coal Based Science and Technology in Shanxi Province MD 2014-07

National Natural Science Foundation of China 61450011

Natural Science Foundation of Shanxi Province of China 2015011052

More Information

Author Bio:
Master student at Taiyuan University of Technology. His research interest covers industrial large data applications, machine learning, data-driven soft sensor modeling, and transfer learning

Ph. D., professor in the Information Department of Beijing University of Technology. His research interest covers data-driven soft sensor modeling and industrial automatic control

Master student at Taiyuan University of Technology. Her research interest covers intelligent information processing, data-driven soft sensor modeling, and transfer learning

Corresponding author: YAN Gao-Wei Ph. D., professor at the College of Information Engineering of Taiyuan University of Technology. His research interest covers complex industrial control system, intelligent control theory and application, and intelligent information processing. Corresponding author of this paper

摘要

摘要: 针对湿式球磨机多工况运行过程中标签样本难以获取和工况改变导致的原测量模型失准问题，本文引入域适应随机权神经网络（Domain adaptive random weight neural network，DARWNN），实现待测工况中少量标签样本与原工况样本共同进行迁移学习.DARWNN网络解决了不同工况间难以共同进行机器学习的问题，但其只考虑经验风险，而未考虑结构风险，从而泛化性能较差，预测精度较低.在此基础上，本文引入流形正则化，并构建基于流形正则化的域适应随机权神经网络（Domain adaptive manifold regularization random weight neural network，DAMRRWNN），以保持数据几何结构，提高相应模型性能.实验结果表明，所提方法可以有效提高DARWNN的学习精度，解决多工况情况下湿式球磨机负荷参数软测量问题.
- 迁移学习 /
- 域适应 /
- 磨机负荷 /
- 流形正则化 /
- 软测量
Abstract: The problem of misalignment of the original measurement model is caused by the difficulty in obtaining the labeled sample and the change of working condition during the operation of the wet-type ball mill. In this paper, we introduce a domain adaptive random weight neural network (DARWNN), thus a small number of labeled samples in the working condition combined with the original working condition samples can be used to implement transfer learning. The DARWNN network can solve the problem of machine learning in different working conditions, however it considers only the empirical risk but not the structural risk. Thus the generalization performance is poor and the prediction accuracy is low. On this basis, we propose a domain adaptive manifold regularization random weight neural network (DAMRRWNN) in terms of manifold regularization to maintain data geometry structure, so as to improve the performance of the corresponding model. Experimental results indicate that the performance of the proposed methods is superior to or at least comparable with the existing benchmarking methods and that the proposed methods can effectively improve the learning accuracy of DARWNN and solve the problem of soft sensor for wet ball mill load parameters under multiple loading conditions.
- Transfer learning /
- domain adaption /
- mill load /
- manifold regularization /
- soft sensor
注释:

1) 本文责任编委贺威

HTML全文

图 1 M个目标域DARWNN算法结构

Fig. 1 M target domain DARWNN algorithm structure

下载: 全尺寸图片幻灯片

图 2 基于流形正则化域适应随机权神经网络的软测量策略

Fig. 2 Soft sensor strategy based on DAMRRWNN

下载: 全尺寸图片幻灯片

图 3 不同设备的数据分布差异图

Fig. 3 Data distribution diagram for difierent devices

下载: 全尺寸图片幻灯片

图 4 光谱数据集实验结果对比

Fig. 4 Comparison of experimental results for spectral data sets

下载: 全尺寸图片幻灯片

图 5 数据预处理流程

Fig. 5 Data preprocessing process

下载: 全尺寸图片幻灯片

图 6 目标域为工况2时RWNN和DAMRRWNN的负荷参数预测结果

Fig. 6 Prediction results of RWNN and DAMRRWNN on load parameter, when the condition 2 is the target domain

下载: 全尺寸图片幻灯片

图 7 目标域为工况2时三种对比算法对负荷参数浓度的预测结果

Fig. 7 Prediction results of three algorithms on load parameter PD, when the condition 2 is the target domain

下载: 全尺寸图片幻灯片

图 8 DAMRRWNN负荷参数预测结果

Fig. 8 DAMRRWNN load parameter prediction results

下载: 全尺寸图片幻灯片

图 9 目标域为工况2时建模方法对比

Fig. 9 The target domain is the working condition 2 and the prediction results are compared

下载: 全尺寸图片幻灯片

图 10 三种负荷参数预测误差对比

Fig. 10 Comparison of prediction errors of three load parameters

下载: 全尺寸图片幻灯片

表 1 m5油脂含量预测结果

Table 1 Prediction result of oil content in m5

评价标准	RWNN	Bagging	JITL
RMSE	0.0490	0.1615	0.0803
NRMSE	0.0655	0.2171	0.1079

下载: 导出CSV

表 2 m5为源域的不同算法实验结果对比

Table 2 Comparison of experimental results of difierent algorithms for m5 as source domain

方法	为目标域				为目标域
	油脂含量		淀粉含量		油脂含量		淀粉含量
	RMSE	NRMSE	RMSE	NRMSE	RMSE	NRMSE	RMSE	NRMSE
RWNN	0.7723	1.0380	1.1142	0.3056	0.8612	0.1575	1.1701	0.3209
Bagging	0.2346	0.3153	0.9127	0.2503	0.2253	0.3028	0.9707	0.2662
JITL	0.5866	0.7844	2.7080	0.7427	0.8021	1.0781	3.3141	0.9090
DARWNN	0.1047	0.1407	0.5066	0.1389	0.1090	0.1465	0.5294	0.1452
DAMRRWNN	0.0908	0.1220	0.4660	0.1278	0.0911	0.1224	0.4958	0.1360

下载: 导出CSV

表 3 不同工况振动信号采集次数

Table 3 Acquisition times of vibration signals under difierent working conditions

MFR	0.3	0.35	0.4	0.45	0.5
次数	139	103	88	95	102

下载: 导出CSV

表 4 磨机负荷参数预测结果对比(RMSE)

Table 4 Comparison of prediction results of mill load parameters (RMSE)

建模算法	负荷参数	1-1	1-2	1-3	1-4	1-5
RWNN	MBVR	0.0279	0.6219	1.0044	1.2182	1.1312
	PD	0.0040	0.1546	0.0656	0.1763	0.2541
	CVR	0.0021	0.0817	0.3063	0.2516	0.3607
Bagging	MBVR	0.0923	0.1618	0.3379	0.3919	0.2346
	PD	0.0157	0.0470	0.0897	0.0880	0.1263
	CVR	0.0086	0.0808	0.1270	0.1395	0.2012
JITL	MBVR	0.0829	0.3688	0.4490	0.7043	0.9147
	PD	0.0138	0.0695	0.1149	0.3406	0.3467
	CVR	0.0094	0.0908	0.1428	0.1777	0.2471
DARWNN	MBVR	$-$	0.1349	0.0854	0.0843	0.0709
	PD	$-$	0.0270	0.0170	0.0153	0.0178
	CVR	$-$	0.0159	0.0114	0.0084	0.0081
DAMRRWNN	MBVR	$-$	0.1058	0.0693	0.0457	0.0365
	PD	$-$	0.0219	0.0149	0.0120	0.0162
	CVR	$-$	0.0141	0.0108	0.0074	0.0072

下载: 导出CSV

参考文献(21)

[1]	汤健, 赵立杰, 柴天佑, 岳恒.基于振动频谱的磨机负荷在线软测量建模.信息与控制, 2012, 41(1):123-128 http://d.old.wanfangdata.com.cn/Periodical/xxykz201201020 Tang Jian, Zhao Li-Jie, Chai Tian-You, Yue Heng. On-line soft-sensing modelling of mill load based on vibration spectrum. Information and Control, 2012, 41(1):123-128 http://d.old.wanfangdata.com.cn/Periodical/xxykz201201020
[2]	汤健, 田福庆, 贾美英, 李东.基于频谱数据驱动的旋转机械设备负荷软测量.北京:国防工业出版社, 2015. Tang Jian, Tian Fu-Qing, Jia Mei-Ying, Li Dong. Soft Sensing of Rotating Machinery Equipment Load Based on Spectrum Data Drive. Beijing:National Defense Industry Press, 2015.
[3]	汤健, 赵立杰, 岳恒, 柴天佑.湿式球磨机筒体振动信号分析及负荷软测量.东北大学学报(自然科学版), 2010, 31(11):1521-1524 http://d.old.wanfangdata.com.cn/Periodical/dbdxxb201011001 Tang Jian, Zhao Li-Jie, Yue Heng, Chai Tian-You. Analysis of vibration signal of wet ball mill shell and soft sensoring for mill load. Journal of Northeastern University (Natural Science), 2010, 31(11):1521-1524 http://d.old.wanfangdata.com.cn/Periodical/dbdxxb201011001
[4]	Huang G B, Zhu Q Y, Siew C K. Extreme learning machine:theory and applications. Neurocomputing, 2006, 70(1-3):489-501 doi: 10.1016/j.neucom.2005.12.126
[5]	Tang J, Deng C, Huang G B. Extreme learning machine for multilayer perceptron. IEEE Transactions on Neural Networks and Learning Systems, 2017, 27(4):809-821 http://ieeexplore.ieee.org/document/7103337
[6]	Maass W. Liquid state machines: motivation, theory, and applications. Computability in Context: Computation and Logic in the Real World. Hackensack, NJ: Imperial College Press, 2009. 275-296
[7]	Zhang M, Liu X G, Zhang Z Y. A soft sensor for industrial melt index prediction based on evolutionary extreme learning machine. Chinese Journal of Chemical Engineering, 2016, 24(8):1013-1019 doi: 10.1016/j.cjche.2016.05.030
[8]	汤健, 柴天佑, 余文, 赵立杰.在线KPLS建模方法及在磨机负荷参数集成建模中的应用.自动化学报, 2013, 39(5):471-486 http://www.aas.net.cn/CN/abstract/abstract17934.shtml Tang Jian, Chai Tian-You, Yu Wen, Zhao Li-Jie. On-line KPLS algorithm with application to ensemble modeling parameters of mill load. Acta Automatica Sinica, 2013, 39(5):471-486 http://www.aas.net.cn/CN/abstract/abstract17934.shtml
[9]	Shao W M, Tian X M, Wang P. Supervised local and non-local structure preserving projections with application to just-in-time learning for adaptive soft sensor. Chinese Journal of Chemical Engineering, 2015, 23(12):1925-1934 doi: 10.1016/j.cjche.2015.11.012
[10]	Pan S J, Yang Q. A survey on transfer learning. IEEE Transactions on Knowledge and Data Engineering, 2010, 22(10):1345-1359 doi: 10.1109/TKDE.2009.191
[11]	Zhang L, Zhang D. Domain adaptation extreme learning machines for drift compensation in E-nose systems. IEEE Transactions on Instrumentation and Measurement, 2015, 64(7):1790-1801 doi: 10.1109/TIM.2014.2367775
[12]	Tenenbaum J B, de Silva V, Langford J C. A global geometric framework for nonlinear dimensionality reduction. Science, 2000, 290(5000):2319-2323 doi: 10.1126-science.290.5500.2319/
[13]	Liu B, Xia S X, Meng F R, Zhou Y. Manifold regularized extreme learning machine. Neural Computing and Applications, 2016, 27(2):255-269 http://d.old.wanfangdata.com.cn/Periodical/txxb201611007
[14]	Schmidt W F, Kraaijveld M A, Duin R P W. Feedforward neural networks with random weights. In: Proceedings of the 11th IAPR International Conference on Pattern Recognition Vol.Ⅱ Conference B: Pattern Recognition Methodology & Systems. The Hague, Netherlands: IEEE, 1992. 1-4
[15]	Liu X, Lin S B, Fang J, Xu, Z B. Is extreme learning machine feasible? A theoretical assessment (Part Ⅰ). IEEE Transactions on Neural Networks and Learning Systems, 2014, 26(1):7-20 http://ieeexplore.ieee.org/document/6862852/
[16]	韩敏, 李德才.基于替代函数及贝叶斯框架的1范数ELM算法.自动化学报, 2011, 37(11):1344-1350 http://www.aas.net.cn/CN/abstract/abstract17624.shtml Han Min, Li De-Cai. A norm 1 regularization term ELM algorithm based on surrogate function and Bayesian framework. Acta Automatica Sinica, 2011, 37(11):1344-1350 http://www.aas.net.cn/CN/abstract/abstract17624.shtml
[17]	Deng W Y, Zheng Q H, Chen L. Regularized extreme learning machine. In: Proceedings of the 2009 IEEE Symposium on Computational Intelligence and Data Mining, 2009. CIDM'09. Nashville, TN, USA: IEEE, 2009. 389-395
[18]	Tomar V S, Rose R C. Manifold regularized deep neural networks. In: Proceedings of the 15th Annual Conference of the International Speech Communication Association. Singapore: ISCA, 2014. 348-352
[19]	Guan N Y, Tao D C, Luo Z G, Yuan B. Manifold regularized discriminative nonnegative matrix factorization with fast gradient descent. IEEE Transactions on Image Processing, 2011, 20(7):2030-2048 doi: 10.1109/TIP.2011.2105496
[20]	徐嘉明, 张卫强, 杨登舟, 刘加, 夏善红.基于流形正则化极限学习机的语种识别系统.自动化学报, 2015, 41 (9):1680-1685 http://www.aas.net.cn/CN/abstract/abstract18741.shtml Xu Jia-Ming, Zhang Wei-Qiang, Yang Deng-Zhou, Liu Jia, Xia Shan-Hong. Manifold regularized extreme learning machine for language recognition. Acta Automatica Sinica, 2015, 41(9):1680-1685 http://www.aas.net.cn/CN/abstract/abstract18741.shtml
[21]	Amar M, Gondal I, Wilson C. Vibration spectrum imaging:a novel bearing fault classification approach. IEEE Transactions on Industrial Electronics, 2015, 62(1):494-502 doi: 10.1109/TIE.2014.2327555