正在加载图片...
赵启东等:基于支持向量数据描述方法的生产过程监控、诊断与优化 ·1797· 0.10 in continuous casting.J Unir Sci Technol Beijing,2008,30(1): 0 (杨尚玉,王旭东,姚曼,等.改进的主元分析法在连铸结品 器过程监测中的应用.北京科技大学学报,2008,30(1): 80) B]Tax D M J,Duin R P W.Support vector domain description Pattern Recognit Lett,1999,20(11):1191 4]Tax DM J,Duin R P W.Support vector data description.Mach Leam,2004,54(1):45 一R控制限 [5]Tang M Z,Wang Y B,Yang C H.Modified support vector data 鲁受控样本点 description for fault diagnosis.Control Decis,2011,26(7):967 10 2030405060 (唐明珠,王岳斌,阳春华.一种改进的支持向量数据描述故 样本点序号 障诊断方法.控制与决策,2011,26(7):967) 图8优化后的监控图 [6] Khazai S,Safari A,Mojaradi B,et al.Improving the SVDD ap- Fig.8 Optimized monitoring chart proach to hyperspectral image classification.IEEE Geosci Remote Sens Lett,2012,9(4):594 通过表3可以发现,优化后各变量的值均在正常 ] Yousef A,Charkari N M.A novel method based on physicochemi- 范围内.通过上述优化方法可以及时将异常生产过程 cal properties of amino acids and one class classificationalgorithm 调整为受控状态,尤其是在新品种的开发过程中,可以 for disease gene identification.J Biomed Inf,2015,56:300 有效指导工艺控制参数与来料参数之间的正确匹配. SukchotratT,Kim S B,Tsung F.One-class classification-based control charts for multivariate process monitoring.IIE Trans, 3结论 2009,42(2):107 (1)提出基于虚警率为优化目标的模型参数优化 Na S G,Yang I B,Heo H.Abnormality detection via SVDD tech- 方法,可以根据数据内在的结构特点获得模型参数,无 nique of motor-generator system in HEV.Int J Automot Technol, 2014,15(4):637 需引入复杂的优化算法,可以实现监控模型的实时在 [10]Jiang Q C,Yan X F.Just-in-time reorganized PCA integrated 线调整 with SVDD for chemical process monitoring.AIChE,2014,60 (2)提出基于贡献图的异常点诊断和基于邻近点 (3):949 替换的工艺参数优化方法,可以实现对生产过程的监 01] Shen FF,Song Z H,Zhou L.Improved PCA-SVDD based mo- 控、诊断与优化,为实际生产过程提供更有效的指导. nitoring method for nonlinear process//201325th Chinese Con- (3)将SVDD方法应用于热轧薄板的产品质量监 trol and Decision Conference (CCDC).IEEE,2013:4330 [12]Chen M C.Hsu CC,Malhotra B,et al.An efficient ICA-DW- 控中,与传统的监控方法P2相比,SVDD具有更高的 SVDD fault detection and diagnosis method for non-Gaussian 检出率,能更加准确地监控生产过程的异常状态 processes.Int Prod Res,2016,54(17):1 (4)随着样本的实时在线更新,SVDD算法的计 [13]Jiang Q C,Yan X F,Lv Z M,et al.Independent component 算复杂度也会逐渐提高,难以满足工业现场实时高效 analysis-based non-Gaussian process monitoring with preselecting 的监控需求.因此,针对海量数据,如何提高SVDD算 optimal components and support vector data description.Int J 法的计算效率,以实现生产过程的监控、诊断与优化, Prod Res,2014,52(11):3273 n4] 是下一步需要深入研究的内容 Zhao F,Zhang J Y,Liu J.An optimizing kemel algorithm for improving the performance of support vector domain description 参考文献 Acta Autom Sin,2008,34(9):1122 (赵峰,张军英,刘敬.一种改善支撑向量域描述性能的核 Liang H.Tong C N.Online data-driven modeling for strip thick- 优化算法.自动化学报,2008,34(9):1122) ness based on subtractive clustering.J Unir Sci Technol Beijing, [15]Zhang J M,Xu X Z,Xie L,et al.Performance optimization of 2012,34(11):1338 SVDD and its application in non-Gaussian process monitoring. (梁辉,童朝南.基于减法聚类的带钢厚度数据驱动建模。北 Chem Ind Eng China,2010,61 (8):2072 京科技大学学报,2012,34(11):1338) (张建明,许仙珍,谢磊,等.支持向量数据描述性能优化及 Yang S Y,Wang X D.Yao M,et al.Application of improved 其在非高斯过程监控中的应用.化工学报,2010,61(8): principal component analysis method to mould process monitoring 2072赵启东等: 基于支持向量数据描述方法的生产过程监控、诊断与优化 图 8 优化后的监控图 Fig. 8 Optimized monitoring chart 通过表 3 可以发现,优化后各变量的值均在正常 范围内. 通过上述优化方法可以及时将异常生产过程 调整为受控状态,尤其是在新品种的开发过程中,可以 有效指导工艺控制参数与来料参数之间的正确匹配. 3 结论 ( 1) 提出基于虚警率为优化目标的模型参数优化 方法,可以根据数据内在的结构特点获得模型参数,无 需引入复杂的优化算法,可以实现监控模型的实时在 线调整. ( 2) 提出基于贡献图的异常点诊断和基于邻近点 替换的工艺参数优化方法,可以实现对生产过程的监 控、诊断与优化,为实际生产过程提供更有效的指导. ( 3) 将 SVDD 方法应用于热轧薄板的产品质量监 控中,与传统的监控方法 T2 PCA 相比,SVDD 具有更高的 检出率,能更加准确地监控生产过程的异常状态. ( 4) 随着样本的实时在线更新,SVDD 算法的计 算复杂度也会逐渐提高,难以满足工业现场实时高效 的监控需求. 因此,针对海量数据,如何提高 SVDD 算 法的计算效率,以实现生产过程的监控、诊断与优化, 是下一步需要深入研究的内容. 参 考 文 献 [1] Liang H,Tong C N. Online data-driven modeling for strip thick￾ness based on subtractive clustering. J Univ Sci Technol Beijing, 2012,34( 11) : 1338 ( 梁辉,童朝南. 基于减法聚类的带钢厚度数据驱动建模. 北 京科技大学学报,2012,34( 11) : 1338) [2] Yang S Y,Wang X D,Yao M,et al. Application of improved principal component analysis method to mould process monitoring in continuous casting. J Univ Sci Technol Beijing,2008,30( 1) : 80 ( 杨尚玉,王旭东,姚曼,等. 改进的主元分析法在连铸结晶 器过程监测中的应用. 北京科技大学学报,2008,30 ( 1 ) : 80) [3] Tax D M J,Duin R P W. Support vector domain description. Pattern Recognit Lett,1999,20( 11) : 1191 [4] Tax D M J,Duin R P W. Support vector data description. Mach Learn,2004,54( 1) : 45 [5] Tang M Z,Wang Y B,Yang C H. Modified support vector data description for fault diagnosis. Control Decis,2011,26( 7) : 967 ( 唐明珠,王岳斌,阳春华. 一种改进的支持向量数据描述故 障诊断方法. 控制与决策,2011,26( 7) : 967) [6] Khazai S,Safari A,Mojaradi B,et al. Improving the SVDD ap￾proach to hyperspectral image classification. IEEE Geosci Remote Sens Lett,2012,9( 4) : 594 [7] Yousef A,Charkari N M. A novel method based on physicochemi￾cal properties of amino acids and one class classificationalgorithm for disease gene identification. J Biomed Inf,2015,56: 300 [8] Sukchotrat T,Kim S B,Tsung F. One-class classification-based control charts for multivariate process monitoring. IIE Trans, 2009,42( 2) : 107 [9] Na S G,Yang I B,Heo H. Abnormality detection via SVDD tech￾nique of motor-generator system in HEV. Int J Automot Technol, 2014,15( 4) : 637 [10] Jiang Q C,Yan X F. Just-in-time reorganized PCA integrated with SVDD for chemical process monitoring. AIChE J,2014,60 ( 3) : 949 [11] Shen F F,Song Z H,Zhou L. Improved PCA-SVDD based mo￾nitoring method for nonlinear process / / 2013 25th Chinese Con￾trol and Decision Conference ( CCDC) . IEEE,2013: 4330 [12] Chen M C,Hsu C C,Malhotra B,et al. An efficient ICA-DW￾SVDD fault detection and diagnosis method for non-Gaussian processes. Int J Prod Res,2016,54( 17) : 1 [13] Jiang Q C,Yan X F,Lv Z M,et al. Independent component analysis-based non-Gaussian process monitoring with preselecting optimal components and support vector data description. Int J Prod Res,2014,52( 11) : 3273 [14] Zhao F,Zhang J Y,Liu J. An optimizing kernel algorithm for improving the performance of support vector domain description. Acta Autom Sin,2008,34( 9) : 1122 ( 赵峰,张军英,刘敬. 一种改善支撑向量域描述性能的核 优化算法. 自动化学报,2008,34( 9) : 1122) [15] Zhang J M,Xu X Z,Xie L,et al. Performance optimization of SVDD and its application in non-Gaussian process monitoring. J Chem Ind Eng China,2010,61( 8) : 2072 ( 张建明,许仙珍,谢磊,等. 支持向量数据描述性能优化及 其在非高斯过程监控中的应用. 化工学报,2010,61 ( 8) : 2072 ·1797·
<<向上翻页
©2008-现在 cucdc.com 高等教育资讯网 版权所有