-
摘要: 為了提高產品質量的穩定性和可靠性，利用機器學習方法實現產品質量在線監控、在線優化和在線預設定，是鋼鐵企業目前亟待解決的關鍵技術。針對企業需求，提出基于軟超球體算法的產品質量異常在線識別和異常原因診斷方法、基于流形學習的工藝參數在線優化方法和基于多變量統計過程控制的工藝規范制定方法。通過將上述方法進行系統集成，并利用工業互聯網技術和大數據分析方法，研發了產品質量在線智能監控系統。目前該系統已在鋼鐵企業十余條生產線上推廣應用，質量在線判定的準確率達到99.2%，在線檢測時間不到0.1 s。Abstract: In recent years, Chinese iron and steel enterprises have mainly adopted the “sampling after the event” method to inspect the product quality before it leaves the factory. Due to the inability to achieve quality inspection for all products, customers often claim and return defective products, leading to major economic losses in steel enterprises. To improve the stability and reliability of product quality, the use of machine learning methods to realize the online monitoring, optimization, and preset of product quality is the key technology to be solved in iron and steel enterprises. Therefore, the online identification and diagnosis of abnormal product quality based on the soft hypersphere, online optimization of the process parameters based on manifold learning and process specification formulation based on the multivariate statistical process control were proposed. In this study, integrated methods of online monitoring, diagnosis, and optimization of product quality were proposed in which the abnormal point of the product quality by the soft hypersphere method, based on the support vector data description, was identified online, and the process parameters were diagnosed through the contribution chart. Optimizing in real time, abnormal process parameters via a local projective transformation of neighbor points was then achieved. The process parameter setting model based on manifold learning by multiclass neighborhoods to extract the manifold of process parameters was established. Meanwhile, the process specification model, based on the maximum inner rectangle of the soft hypersphere, was established to obtain an effective control interval of the process parameters. Through system integration with the proposed methods and using industrial internet technology and big data analysis methods, the system of intelligent online monitoring of product quality has been successfully developed. At present, the system has been applied to more than ten production lines in iron and steel enterprises. The accuracy rate of online quality determination is 99.2%, and the online detection time is less than 0.1 s.
-
圖 2 線性超橢球將異常點判為正常點
Figure 2. Abnormal samples misjudged as normal samples in the linear hypersphere
圖 3 樣本點從原始空間映射到特征空間
Figure 3. Samples mapped from the original space into the feature space
圖 9 時效溫度與快冷溫度的上、下限
Figure 9. Up and low limits of the aging temperature and fast-cooling temperature
圖 5 訓練集的控制限R2(a)和在線監測結果(b)
Figure 5. Control limit R2 (a) of the training set and online monitoring result (b)
圖 6 工藝參數(參數序號在表1中)的貢獻圖.(a)第25樣本;(b)第57樣本
Figure 6. Contribution chart of parameters: (a) sample No. 25; (b) sample No. 57 (serial numbers of the parameters are listed in Table 1)
圖 7 工藝參數與屈服強度的主流形。(a)碳含量流形;(b)錳含量流形;(c)精軋入口溫度流形;(d)均熱溫度流形
Figure 7. Main manifold between the process parameters and yield strength: (a) manifold of C; (b) manifold of Mn; (c) manifold of the entry temperature of finish rolling; (d) manifold of the soaking temperature
圖 8 熱處理工序的工藝參數上、下限.(a)均熱溫度與快冷溫度;(b)均熱溫設與時效溫度;(c)均熱溫度與緩冷溫度;(d)時效溫度與緩冷溫度
Figure 8. Up and low limits of the process parameters in the heat treatment: (a) soaking and fast-cooling temperature; (b) soaking and aging temperature; (c) soaking and slow-cooling temperature; (d) aging and slow-cooling temperature
表 1 關鍵工藝參數、質量指標及統計量
Table 1. Key process parameters, quality indexes, and statistics
Parameters Max Min Mean Process parameters No.1 Mass fraction of C / % 0.0027 0.0011 0.0017 No.2 Mass fraction of Mn / % 0.160 0.100 0.126 No.3 Mass fraction of P / % 0.014 0.007 0.010 No.4 Mass fraction of S / % 0.0139 0.0024 0.0077 No.5 Exit temperature of heating furnace / °C 1277.3 1247.1 1263.04 No.6 Entry temperature of finish rolling / °C 1083.9 1014.0 1039.08 No.7 Exit temperature of finish rolling / °C 928.5 898.7 917.17 No.8 Coiling temperature / °C 755.4 654.5 711.70 No.9 Soaking temperature / °C 854.9 789.7 824.27 No.10 Fast-cooling exit temperature / °C 455.7 378.8 431.13 No.11 Aging exit temperature / °C 394.1 345.1 374.52 No.12 Slow-cooling exit temperature / °C 676.4 606.0 641.61 Quality indexes No.1 Tensile strength / MPa 308.0 276.0 290.1 No.2 Yield strength / MPa 125.0 160.0 139.4 No.3 Elongation / % 40.5 50.5 45.1 No.4 Plastic strain ratio 2.10 3.5 2.85 
下載: 導出CSV
表 2 屈服強度的質量設計
Table 2. Quality design of the yield strength
Process parameter Mass fraction
of C / %Mass fraction
of Mn / %Mass fraction
of P / %Mass fraction
of S / %Exit temperature of heating furnace /
°CEntry temperature of finish rolling /
°CExit temperature of finish rolling /
°CCoiling temperature /
°CSoaking temperature /
°CFast-cooling exit temperature /
°CAging exit temperature /
°CSlow-cooling exit temperature /
°CYield strength 130 MPa 0.0015–0.00185 0.105–0.125 0.008–0.012 0.006–0.0095 1261–1267 1030–1040 914–920 725–740 827–843 420–450 375–390 625–645 140 MPa 0.0016–0.0019 0.125–0.135 0.008–0.012 0.006–0.0095 1259–1266 1030–1040 914–920 725–740 817–835 420–450 370–385 630–650 150 MPa 0.0016–0.002 0.135–0.155 0.008–0.012 0.006–0.0095 1257–1265 1040–1050 914–920 650–660 810–827 420–450 365–380 635–655 160 MPa 0.0016–0.0021 0.155–0.170 0.008–0.012 0.006–0.0095 1256–1264 1040–1050 914–920 650–660 806–826 420–450 360–375 640–660
下載: 導出CSV
表 3 第25號樣本點的工藝參數調整值
Table 3. Adjustment of process parameters for sample No. 25
Process parameter Exit temperature of heating furnace / °C Entry temperature of finish rolling / °C Exit temperature of finish rolling / °C Coiling temperature / °C Soaking temperature / °C Fast-cooling exit temperature / °C Aging exit temperature / °C Slow-cooling exit temperature / °C Original value 1247.5 1036.9 917.2 740 840.1 445.1 389.4 664.3 Adjustment 11.2 6.3 3.3 ?36.0 ?12.4 ?6.3 ?5.3 ?19.7 Real value 1258.7 1043.2 920.5 704 827.7 438.8 384.1 644.6
下載: 導出CSV
表 4 熱處理工藝參數的相關系數
Table 4. Correlation coefficient of the process parameters in the heat treatment
Correlation coefficient Soaking temperature Fast-cooling temperature Aging temperature Slow-cooling temperature Soaking temperature 1.0 0.12 0.10 0.10 Fast-cooling temperature — 1.0 0.72 0.61 Aging temperature — — 1.0 0.43 Slow-cooling temperature — — — 1.0
下載: 導出CSV
中文字幕在线观看表 5 工藝參數的預設值
Table 5. Preinstalling values of process parameters
Process parameter Mass fraction of C / % Mass fraction of Mn / % Mass fraction of P / % Mass fraction of S / % Exit temperature of heating furnace / °C Entry temperature of finish rolling / °C Exit temperature of finish rolling / °C Coiling temperature / °C Soaking temperature / °C Fast-cooling exit temperature / °C Aging exit temperature / °C Slow-cooling exit temperature / °C Soft hypersphere 0.0024–
0.00180.150–
0.1000.0115–
0.00650.0118–
0.00481273–1255 1055–1030 908–924 740–685 839–814 451–398 392–357 662–620 Max–min 0.0027–
0.00110.160–
0.100.014–
0.0070.0139–
0.00241277–1247 1084–1014 928–899 755–654 855–790 456–379 394–345 674–606 6σ 0.0026–
0.00080.1759–
0.07990.0156–
0.00480.0133–
0.00251280–1246 1067–1013 926–907 834–582 857–793 489–376 407–345 675–609
下載: 導出CSV
-
參考文獻
[1] Communication Promoters Group of the Industry-Science. Recommendations for Implementing the Strategic Initiative Industrie 4.0, 2013,http://www.acatech.de/industrie4.0 [2] China Electronics Standardization Institute. Industrial Big Data White Paper (2017 version). http://www.cesi.cn/201703/2250.html中國電子技術標準研究院. 工業大數據白皮書(2017版). http://www.cesi.cn/201703/2250.html [3] China Electronics Standardization Institute. Cyber-physical System White Paper (2017 version). http://www.cesi.cn/201703/2251.html中國電子技術標準研究院. 信息物理系統白皮書(2017版). http://www.cesi.cn/201703/2251.html [4] Gehrmann C, Gunnarsson M. A digital twin based industrial automation and control system security architecture. IEEE Trans Ind Inform, 2020, 16(1): 669 doi: 10.1109/TII.2019.2938885 [5] Tao F, Cheng J F, Qi Q L. IIHub: an industrial Internet-of-things hub toward smart manufacturing based on cyber-physical system. IEEE Trans Ind Inform, 2018, 14(5): 2271 doi: 10.1109/TII.2017.2759178 [6] Morgan D, Jacobs R. Opportunities and challenges for machine learning in materials science. Annu Rev Mater Res, 2020, 50(1): 71 doi: 10.1146/annurev-matsci-070218-010015 [7] Pablo J J, Jackson N E, Webb M A, et al. New frontiers for the materials genome initiative. Npj Comput Mater, 2019, 5(1): 1 doi: 10.1038/s41524-018-0138-z [8] Kim K, Ward L, He J G, et al. Machine-learning-accelerated high-throughput materials screening: Discovery of novel quaternary Heusler compounds. Phys Rev Materials, 2018, 2(12): 123801 doi: 10.1103/PhysRevMaterials.2.123801 [9] Boyd P G, Lee Y, Smit B. Computational development of the nanoporous materials genome. Nat Rev Mater, 2017, 2: 17037 doi: 10.1038/natrevmats.2017.37 [10] Arróyave R, McDowell D L. Systems approaches to materials design: Past, present, and future. Annu Rev Mater Res, 2019, 49(1): 103 doi: 10.1146/annurev-matsci-070218-125955 [11] Zunger A. Inverse design in search of materials with target functionalities. Nat Rev Chem, 2018, 2: 121 doi: 10.1038/s41570-018-0121 [12] Sanchez-Lengeling B, Aspuru-Guzik A. Inverse molecular design using machine learning: Generative models for matter engineering. Science, 2018, 361(6400): 360 doi: 10.1126/science.aat2663 [13] Noé F, Tkatchenko A, Müller K R, et al. Machine learning for molecular simulation. Annu Rev Phys Chem, 2020, 71(1): 361 doi: 10.1146/annurev-physchem-042018-052331 [14] Wu S, Kondo Y, Kakimoto M A, et al. Machine-learning-assisted discovery of polymers with high thermal conductivity using a molecular design algorithm. Npj Comput Mater, 2019, 5: 66 doi: 10.1038/s41524-019-0203-2 [15] Butler K T, Davies D W, Cartwright H, et al. Machine learning for molecular and materials science. Nature, 2018, 559(7715): 547 doi: 10.1038/s41586-018-0337-2 [16] Gomberg J A, Medford A J, Kalidindi S R. Extracting knowledge from molecular mechanics simulations of grain boundaries using machine learning. Acta Mater, 2017, 133: 100 doi: 10.1016/j.actamat.2017.05.009 [17] Bartók A P, De S, Poelking C, et al. Machine learning unifies the modeling of materials and molecules. Sci Adv, 2017, 3(12): e1701816 doi: 10.1126/sciadv.1701816 [18] Kourti T. Multivariate analysis for process understanding, monitoring, control, and optimization of lyophilization processes // Quality by Design for Biopharmaceutical Drug Product Development. New York, 2015: 537 [19] Huang J, Kaul G, Cai C S, et al. Quality by design case study: An integrated multivariate approach to drug product and process development. Int J Pharm, 2009, 382(1-2): 23 doi: 10.1016/j.ijpharm.2009.07.031 [20] Cui Y, Song X L, Chuang K, et al. Variable selection in multivariate modeling of drug product formula and manufacturing process. J Pharm Sci, 2012, 101(12): 4597 doi: 10.1002/jps.23322 [21] Kabzan J, Hewing L, Liniger A, et al. Learning-based model predictive control for autonomous racing. IEEE Robotics Autom Lett, 2019, 4(4): 3363 doi: 10.1109/LRA.2019.2926677 [22] Xu G, Li M, Xu J W. Application research of on-line quality control method to metallurgical products // 2019 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM). Macao, 2019: 390 [23] Hertneck M, K?hler J, Trimpe S, et al. Learning an approximate model predictive controller with guarantees. IEEE Control Syst Lett, 2018, 2(3): 543 doi: 10.1109/LCSYS.2018.2843682 [24] Mesbah A. Stochastic model predictive control: An overview and perspectives for future research. IEEE Control Syst Mag, 2016, 36(6): 30 doi: 10.1109/MCS.2016.2602087 [25] Balta E C, Lin Y K, Barton K, et al. Production as a service: A digital manufacturing framework for optimizing utilization. IEEE Trans Autom Sci Eng, 2018, 15(4): 1483 doi: 10.1109/TASE.2018.2842690 [26] DeSmit Z, Elhabashy A E, Wells L J, et al. An approach to cyber-physical vulnerability assessment for intelligent manufacturing systems. J Manuf Syst, 2017, 43: 339 doi: 10.1016/j.jmsy.2017.03.004 [27] Michniewicz J, Reinhart G. Cyber-physical-robotics-modelling of modular robot cells for automated planning and execution of assembly tasks. Mechatronics, 2016, 34: 170 doi: 10.1016/j.mechatronics.2015.04.012 [28] Pirvu B C, Zamfirescu C B, Gorecky D. Engineering insights from an anthropocentric cyber-physical system: A case study for an assembly station. Mechatronics, 2016, 34: 147 doi: 10.1016/j.mechatronics.2015.08.010 [29] Jiang P Y, Ding K, Leng J W. Towards a cyber-physical-social-connected and service-oriented manufacturing paradigm: Social Manufacturing. Manuf Lett, 2016, 7: 15 doi: 10.1016/j.mfglet.2015.12.002 [30] Zawadzki P, ?ywicki K. Smart product design and production control for effective mass customization in the industry 4.0 concept. Manag Prod Eng Rev, 2016, 7(3): 105 [31] Yao X F, Lin Y Z. Emerging manufacturing paradigm shifts for the incoming industrial revolution. Int J Adv Manuf Technol, 2016, 85(5-8): 1665 doi: 10.1007/s00170-015-8076-0 [32] Xu G. Application Research of CPS in On-Line Quality Control to Metallurgical Products [Dissertation]. Beijing: University of Science and Technology Beijing, 2019徐鋼. CPS在冶金產品質量在線管控中應用研究[學位論文]. 北京: 北京科技大學, 2019 [33] Xu J W. Quality Control Theory and Method of Metallurgical Production Process. Beijing: Metallurgical Industry Press, 2015徐金梧. 冶金生產過程質量監控理論與方法. 北京: 冶金工業出版社, 2015 [34] Jolliffe I T, Cadima J. Principal component analysis: A review and recent developments. Phil Trans R Soc A, 2016, 374(2065): 20150202 doi: 10.1098/rsta.2015.0202 [35] Tabora J E, Domagalski N. Multivariate analysis and statistics in pharmaceutical process research and development. Annu Rev Chem Biomol Eng, 2017, 8: 403 doi: 10.1146/annurev-chembioeng-060816-101418 [36] Johnson R A, Wichern D W. Applied Multivariate Statistical Analysis, 5th edition. Prentice Hall, 2002 [37] Tax D M J, Duin R P W. Support vector data description. Mach Learn, 2004, 54(1): 45 doi: 10.1023/B:MACH.0000008084.60811.49 [38] Zimek A, Schubert E, Kriegel H P. A survey on unsupervised outlier detection in high-dimensional numerical data. Stat Anal Data Min:ASA Data Sci J, 2012, 5(5): 363 doi: 10.1002/sam.11161 [39] Gang X. Online monitoring and control method of product quality based on embedded cyber-physical system models. J Mech Eng, 2017, 53(12): 94 doi: 10.3901/JME.2017.12.094 [40] Shawe-Taylor J, Cristianini N. Kernel Methods for Pattern Analysis. Cambridge: Cambridge University Press, 2004 [41] Volpi M, Tuia D, Camps-Valls G, et al. Unsupervised change detection with kernels. IEEE Geosci Remote Sens Lett, 2012, 9(6): 1026 doi: 10.1109/LGRS.2012.2189092 [42] Li M, Xu J W, Yang J H, et al. Multiple manifolds analysis and its application to fault diagnosis. Mech Syst Signal Process, 2009, 23(8): 2500 doi: 10.1016/j.ymssp.2009.05.001 [43] Belkin M, Niyogi P, Sindhwani V. Manifold regularization: A geometric framework for learning from labeled and unlabeled examples. J Mach Learn Res, 2006, 7: 2399 [44] Zhang Z, Zha H. Principal manifolds and nonlinear dimensionality reduction via tangent space alignment. SIAM J Sci Comput, 2004, 26(1): 313 doi: 10.1137/S1064827502419154 [45] Saul L K, Roweis S T. Think globally, fit locally: unsupervised learning of low dimensional manifolds. J Mach Learn Res, 2003, 4: 119 -
點擊查看大圖
圖(10) / 表(5)
計量
- 文章訪問數: 1084
- HTML全文瀏覽量: 441
- PDF下載量: 78
- 被引次數: 0
