基于局部线性嵌入的测地线流式核多工况软测量建模方法

doi:10.11949/0438-1157.20190934

摘要/Abstract

摘要：

针对复杂工业过程在多工况条件下缺乏标记样本无法进行软测量建模，而原有模型失准问题，研究了一种局部线性嵌入（locally linear embedding, LLE）和测地线流式核（geodesic flow kernel, GFK）相结合的无监督软测量建模方法。该方法首先通过局部线性嵌入提取各个工况间的公共模式信息，然后将已知工况数据和未知工况数据的公共模式信息投影到流形空间，利用测地线流式核框架在流形空间上实现域迁移，以减小不同工况间数据的分布差异。最后用偏最小二乘回归法建立软测量模型，得到主导变量的软测量值。通过对TE过程中不同工况下的成分变量软测量和不同工况下的球磨机负荷参数软测量结果，验证了所提算法的实用性和有效性。

关键词: 迁移学习, 测地线流式核, 软测量, 算法, 预测, 过程控制

Abstract:

For the complex industrial process, the lack of labeled samples under multiple working conditions cannot be used for soft sensor modeling, and the original model is out of alignment. In this paper, an unsupervised soft sensor modeling method combining locally linear embedding (LLE) and geodesic flow kernel (GFK) is studied. The method firstly extracts the common feature between each working condition by locally linear embedding, and then projects the common feature of the known working condition data and the unknown working condition data into the manifold space, and uses the geodesic flow kernel frame on the manifold space for domain transfer. Finally, the soft-measurement model is established by partial least squares regression method, and the soft-measurement value of the dominant variable is obtained. The practicality and effectiveness of the proposed algorithm are verified by soft sensor of component variables under different working conditions in TE process and soft sensor results of ball mill load parameters under different working conditions.

Key words: transfer learning, geodesic flow kernel, soft sensor, algorithm, prediction, process control

中图分类号:

TP 29

杜宇浩, 阎高伟, 李荣, 王芳. 基于局部线性嵌入的测地线流式核多工况软测量建模方法[J]. 化工学报, 2020, 71(3): 1278-1287.

Yuhao DU, Gaowei YAN, Rong LI, Fang WANG. Multiple working conditions soft sensor modeling method of geodesic flow kernel based on locally linear embedding[J]. CIESC Journal, 2020, 71(3): 1278-1287.

图/表 14

图1 LLE算法示意图

Fig.1 Locally linear embedding

图2 LLEGFK算法示意图

Fig.2 Locally linear embedding-geodesic flow kernel

表 1 三种工况数据

Table 1 Data of three working conditions

工况	G/H比率	产品等级/(kg/h)
1	50/50	14076
2	10/90	14077
3	90/10	11111

表 2 各工况下不同算法参数软测量均方根误差对比

Table 2 Comparison of RMSE of soft sensor of different algorithm parameters under different working conditions

Item	成分A				成分B				成分C
Item	PLSR	LLE	GFK	LLEGFK	PLSR	LLE	GFK	LLEGFK	PLSR	LLE	GFK	LLEGFK
1—2	1.739	0.989	0.890	0.460	1.192	1.089	0.956	0.719	0.620	0.543	0.537	0.425
1—3	2.635	2.201	1.464	0.839	5.219	4.864	4.793	4.179	0.901	0.859	0.744	0.476
2—1	1.670	1.577	0.808	0.437	1.851	1.618	1.135	0.625	0.711	0.646	0.599	0.510
2—3	2.151	1.920	1.597	0.983	4.845	4.251	4.570	3.963	0.937	0.875	0.776	0.613
3—1	1.629	1.397	0.356	0.328	4.682	4.472	4.314	4.095	0.746	0.725	0.695	0.501
3—2	1.484	1.057	0.704	0.655	3.873	4.902	3.924	3.783	0.661	0.621	0.635	0.550

图3 工况1迁移到工况2对成分A含量软测量结果

Fig.3 Predicted results of 1—2 component A

图4 工况1迁移到工况3对成分A含量软测量结果

Fig.4 Predicted results of 1—3 component A

图5 工况2迁移到工况1对成分A含量软测量结果

Fig.5 Predicted results of 2—1 component A

图6 工况2迁移到工况3对成分A含量软测量结果

Fig.6 Predicted results of 2—3 component A

图7 不同工况提取公共模式信息前后分布

Fig.7 Distribution of common feature before and after extraction under different working conditions

表3 各工况参数与实验次数

Table 3 MFR and number of experiments under different working conditions

工况	1	2	3	4	5
介质充填率	0.3	0.35	0.4	0.45	0.5
实验次数	139	103	88	95	102

表 4 各算法软测量均方根误差对比

Table 4 Comparison of RMSE of soft sensor of different algorithm parameters under different working conditions

Item	MBVR				PD				CVR
Item	PLSR	LLE	GFK	LLEGFK	PLSR	LLE	GFK	LLEGFK	PLSR	LLE	GFK	LLEGFK
1—2	0.534	0.208	0.302	0.074	0.056	0.130	0.062	0.016	0.087	0.132	0.131	0.102
1—3	0.746	0.406	0.356	0.133	0.196	0.175	0.056	0.041	0.296	0.207	0.182	0.134
1—4	1.835	0.491	0.397	0.102	0.596	0.235	0.075	0.049	0.397	0.115	0.356	0.109
1—5	2.151	1.920	0.541	0.242	1.326	0.185	0.157	0.061	0.822	0.430	0.746	0.296

图8 工况1迁移到工况2料球比软测量结果

Fig.8 Predicted results of 1—2 MBVR

图9 工况1迁移到工况3料球比软测量结果

Fig.9 Predicted results of 1—3 MBVR

图10 工况1迁移到工况4料球比软测量结果

Fig.10 Predicted results of 1—4 MBVR

参考文献 30

1	Wang J G,Xie Z,Yao Y,et al.Soft sensor development for improving economic efficiency of the coke dry quenching process[J].Journal of Process Control,2019,77:20-28.
2	曹鹏飞,罗雄麟.化工过程软测量建模方法研究进展[J].化工学报,2013,64(3):788-800.
	Cao P F,Luo X L.Modeling of soft sensor for chemical process[J].CIESC Journal,2013,64(3):788-800.
3	刘毅,王海清,李平.用于发酵过程在线建模的自适应局部最小二乘支持向量机回归方法[J].化工学报,2008,59(8):2052-2057.
	Liu Y,Wang H Q,Li P.Adaptive local learning based least squares support vector regression with application to online modeling for fermentation processes[J].Journal of Chemical Industry and Engineering(China),2008,59(8):2052-2057.
4	刘毅,王海清,李平.局部最小二乘支持向量机回归在线建模方法及其在间歇过程的应用[J].化工学报,2007,58(11):2846-2851.
	Liu Y,Wang H Q,Li P.Local least squares support vector regression with application to online modeling for batch processes[J].Journal of Chemical Industry and Engineering(China),2007,58(11):2846-2851.
5	Zheng W,Liu Y,Gao Z,et al.Just-in-time semi-supervised soft sensor for quality prediction in industrial rubber mixers[J].Chemometrics and Intelligent Laboratory Systems,2018,180:36-41.
6	Jin H,Chen X,Li W,et al.Adaptive soft sensor development based on online ensemble gaussian process regression for nonlinear time-varying batch processes[J].Industrial & Engineering Chemistry Research,2015,54(30):7320-7345.
7	Jin H,Chen X,Yang J,et al.Multi-model adaptive soft sensor modeling method using local learning and online support vector regression for nonlinear time-variant batch processes[J].Chemical Engineering Science,2015,131:282-303.
8	Liu Y,Yang C,Gao Z,et al.Ensemble deep kernel learning with application to quality prediction in industrial polymerization processes[J].Chemometrics and Intelligent Laboratory Systems,2018,174:15-21.
9	Liu J X,Tao L,Chen J H.Quality prediction for multi-grade processes by just-in-time latent variable modeling with integration of common and special features[J].Chemical Engineering Science,2018,191:31-41.
10	杜永贵,李思思,阎高伟,等.基于流形正则化域适应湿式球磨机负荷参数软测量[J].化工学报,2018,69(3):1244-1251.
	Du Y G,Li S S,Yan G W,et al.Soft sensor of wet ball mill load parameter based on domain adaptation with manifold regularization[J].CIESC Journal,2018,69(3):1244-1251.
11	Lv X B,Guan Y,Deng B Y.Transfer learning based clinical concept extraction on data from multiple sources[J].Journal of biomedical informatics,2014,52:55-64.
12	Grubinger T,Chasparis G C,Natschläger T.Generalized online transfer learning for climate control in residential buildings[J].Energy and Buildings,2017,139:63-71.
13	Sun S,Shi H,Wu Y.A survey of multi-source domain adaptation[J].Information Fusion,2015,24:84-92.
14	Liu Y,Yang C,Liu K,et al.Domain adaptation transfer learning soft sensor for product quality prediction[J].Chemometrics and Intelligent Laboratory Systems,2019,192:103813.
15	贺敏,汤健,郭旭琦,等.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[J].自动化学报,2019,45(2):398-406.
	He M,Tang J,Guo X Q,et al.Soft sensor of wet ball mill load parameter based on domain adaptation with manifold regularization[J].Acta Automatica Sinica,2019,45(2):398-406.
16	Li J,Liu W,Zhou Y,et al.Domain adaptation with few labeled source samples by graph regularization[J].Neural Processing Letters,2019.doi:10.1007/s11063-019-10075-z. DOI
17	Wang C.A Geometric Framework for Transfer Learning Using Manifold Alignment[D].University of Massachusetts Amherst,2010.
18	Shrivastava A,Shekhar S,Patel V M.Unsupervised domain adaptation using parallel transport on Grassmann manifold[C]//2014 IEEE Winter Conference on Applications of Computer Vision (WACV).Steamboat Springs, CO, USA:IEEE,2014:277-284.
19	王存睿,张庆灵,段晓东,等.基于流形结构的人脸民族特征研究[J].自动化学报,2018,44(1):140-159.
	Wang C R,Zhang Q L,Duan X D,et al.Research of face ethnic features from manifold structure[J].Acta Automatica Sinica,2018,44(1):140-159.
20	Gopalan R,Li R,Chellappa R.Domain adaptation for object recognition: an unsupervised approach[C]//International Conference on Computer Vision.Barcelona, Spain:IEEE,2011:999-1006.
21	Gong B,Shi Y,Sha F,et al.Geodesic flow kernel for unsupervised domain adaptation[C]//IEEE Conference on Computer Vision and Pattern Recognition.Providence, RI, USA:IEEE,2012:2066-2073.
22	Roweis S T,Saul L K.Nonlinear dimensionality reduction by locally linear embedding[J].Science,2000,290(5500):2323-2326.
23	Zhang Y,Wang C.Modeling and monitoring of multimodes process[C]//International Symposium on Neural Networks.Berlin, Germany:Springer,2012:159-168.
24	Samat A,Gamba P,Abuduwaili J,et al.Geodesic flow kernel support vector machine for hyperspectral image classification by unsupervised subspace feature transfer[J].Remote Sensing,2016,8(3):234.
25	Wang J,Feng W,Chen Y,et al.Visual domain adaptation with manifold embedded distribution alignment[C]//2018 ACM Multimedia Conference on Multimedia Conference.Seoul, South Korea:ACM,2018:402-410.
26	Zou W,Xia Y,Li H.Fault diagnosis of Tennessee-Eastman process using orthogonal incremental extreme learning machine based on driving amount[J].IEEE Transactions on Cybernetics,2018,48(12):3403-3410.
27	Hsu C C,Chen L S,Liu C H.A process monitoring scheme based on independent component analysis and adjusted outliers[J].International Journal of Production Research,2010,48(6):1727-1743.
28	Wang X,Ren J,Liu S.Distribution adaptation and manifold alignment for complex processes fault diagnosis[J].Knowledge-Based Systems,2018,156:100-112
29	阎高伟,贺敏,汤健,等.基于最大均值差异多源域迁移学习的湿式球磨机负荷参数软测量[J].控制与决策,2018,33(10):70-75.
	Yan G W,He M,Tang J,et al.Soft sensor of wet ball mill load based on maximum mean discrepancy multi-source domain transfer learning[J].Control and Decision,2018,33(10):70-75.
30	汤健.基于频谱数据驱动的旋转机械设备负荷软测量[M].北京:国防工业出版社,2015.
	Tang J.Soft Sensing of Rotating Machinery Equipment Load based on Spectrum Data Drive[M].Beijing:National Defense Industry Press,2015.

[1]	康飞, 吕伟光, 巨锋, 孙峙. 废锂离子电池放电路径与评价研究[J]. 化工学报, 2023, 74(9): 3903-3911.
[2]	曹跃, 余冲, 李智, 杨明磊. 工业数据驱动的加氢裂化装置多工况切换过渡状态检测[J]. 化工学报, 2023, 74(9): 3841-3854.
[3]	诸程瑛, 王振雷. 基于改进深度强化学习的乙烯裂解炉操作优化[J]. 化工学报, 2023, 74(8): 3429-3437.
[4]	闫琳琦, 王振雷. 基于STA-BiLSTM-LightGBM组合模型的多步预测软测量建模[J]. 化工学报, 2023, 74(8): 3407-3418.
[5]	尹刚, 李伊惠, 何飞, 曹文琦, 王民, 颜非亚, 向禹, 卢剑, 罗斌, 卢润廷. 基于KPCA和SVM的铝电解槽漏槽事故预警方法[J]. 化工学报, 2023, 74(8): 3419-3428.
[6]	郭雨莹, 敬加强, 黄婉妮, 张平, 孙杰, 朱宇, 冯君炫, 陆洪江. 稠油管道水润滑减阻及压降预测模型修正[J]. 化工学报, 2023, 74(7): 2898-2907.
[7]	邵伟明, 韩文学, 宋伟, 杨勇, 陈灿, 赵东亚. 基于分布式贝叶斯隐马尔可夫回归的动态软测量建模方法[J]. 化工学报, 2023, 74(6): 2495-2502.
[8]	李艳辉, 丁邵明, 白周央, 张一楠, 于智红, 邢利梅, 高鹏飞, 王永贞. 非常规服役超临界锅炉的微纳尺度腐蚀动力学模型建立及应用[J]. 化工学报, 2023, 74(6): 2436-2446.
[9]	于源, 陈薇薇, 付俊杰, 刘家祥, 焦志伟. 几何相似涡流空气分级机环形区流场变化规律研究及预测[J]. 化工学报, 2023, 74(6): 2363-2373.
[10]	高学金, 姚玉卓, 韩华云, 齐咏生. 基于注意力动态卷积自编码器的发酵过程故障监测[J]. 化工学报, 2023, 74(6): 2503-2521.
[11]	周继鹏, 何文军, 李涛. 异形催化剂上乙烯催化氧化失活动力学反应工程计算[J]. 化工学报, 2023, 74(6): 2416-2426.
[12]	苏晓丹, 朱干宇, 李会泉, 郑光明, 孟子衡, 李防, 杨云瑞, 习本军, 崔玉. 湿法磷酸半水工艺考察与石膏结晶过程研究[J]. 化工学报, 2023, 74(4): 1805-1817.
[13]	郑书闽, 郭鹏程, 颜建国, 王帅, 李文博, 周淇. 微小通道内过冷流动沸腾阻力特性实验及预测研究[J]. 化工学报, 2023, 74(4): 1549-1560.
[14]	吴心远, 刘奇磊, 曹博渊, 张磊, 都健. Group2vec：基于无监督机器学习的基团向量表示及其物性预测应用[J]. 化工学报, 2023, 74(3): 1187-1194.
[15]	张生安, 刘桂莲. 高效太阳能电解水制氢系统及其性能的多目标优化[J]. 化工学报, 2023, 74(3): 1260-1274.