DOI: 10.3724/SP.J.1219.2013.00693

Information and Control (信息与控制) 2013/42:6 PP.693-699

Multiple Neural Network Model Based on Data Partition Using Feature Clustering

For the shortage of poor robustness in the description of nonlinear systems with single data model, a multiple neural network modeling method based on data partition using feature clustering is proposed. By structuring feature function, the data set is clustered and divided into multiple data subsets. The partitioned data subsets are used to build sub-models by RBF (radial bases function) neural network. Then the sub-models are adaptively weighted by minimizing mean square error in order to improve the model accuracy and model robustness. Finally, the proposed modeling method is applied to nonlinear system and moisture regain soft sensing in textile slashing process, and is compared with the single neural network model and the equivalently weighted multiple neural network model. The experimental results show that the root mean square error and the max absolute error of the proposed multiple neural network model are less than the single neural network model and the equivalently weighted multiple neural network model.

Key words:nonlinear system,multiple neural network,data partition,feature clustering,adaptive weight

ReleaseDate:2015-04-15 18:52:39

[1] Misra M, Yue H H, Qin S J, et al. Multivariate process monitoring and fault diagnosis by multi-scale PCA[J]. Computers and Chemical Engineering, 2002, 26(9): 1281-1293.

[2] Cherry G A, Qin S J. Monitoring non-normal data with principal component analysis and adaptive density estimation [C]// Proceedings of the 46th IEEE Conference on Decision and Control. 2007: 352-359.

[3] Chong I G., Albin S L, Jun C H. A data mining approach to process optimization without an explicit quality function[J]. IIE Transactions on Operations Engineering, 2007, 39: 795-804.

[4] Kano M, Nakagawa Y. Data-based process monitoring, process control, and quality improvement: Recent developments and applications in steel industry[J]. Computers and Chemical Engineering, 2008, 32(1/2): 12-24.

[5] 汤健,赵立杰,柴天佑,等. 基于振动频谱的磨机负荷在线软测量建模[J]. 信息与控制,2012,41(1):123-128. Tang J, Zhao L J, Chai T Y, et al. On-line soft-sensing modeling of mill load based on vibration spectrum[J]. Information and Control, 2012, 41(1): 123-128.

[6] 桂卫华,阳春华,李勇刚,等. 基于数据驱动的铜闪速熔炼过程操作模式优化及应用[J]. 自动化学报,2009,35(6):717-724. Gui W H, Yang C H, Li Y G, et al. Data-driven operational-pattern optimization for copper flash smelting process[J]. Acta Automatic Sinica, 2009, 35(6): 717-724.

[7] Zhang Y X, Liu M. An operating parameters setting using empirical data for slashing process[J]. International Journal of Innovative Computing, Information and Control, 2010, 6(5): 2013-2023.

[8] 杨维维,乔俊飞. 基于递归高阶神经网络的污水处理系统建模[J]. 信息与控制,2005,40(5):710-714. Yang W W, Qiao J F. Wastewater treatment system modeling based on high-order recurrent neural network[J]. Information and Control, 2005, 40(5): 710-714.

[9] Wang X D, Luo R F, Shao H H. Designing a soft sensor model for a distillation column with the fuzzy distributed radial basis function neural network [C]// Proceedings of the 35th IEEE Conference on Decision and Control. Piscataway, NJ, USA: IEEE, 1999: 1714-1719.

[10] Cho S Z, Cho Y J, Yoon S C. Reliable roll force prediction in cold mill using multiple neural networks neural networks[J]. IEEE Transactions on Neural Networks, 1997, 8(4): 874-882.

[11] 熊智华,王雄,徐用懋. 一种利用多神经网络结构建立非线性软测量模型的方法[J]. 控制与决策,2000,15(2):173-176. Xiong Z H, Wang X, Xu Y M. Nonlinear software sensor modeling using multiple neural network[J]. Control and Decision, 2000, 15(2): 173-176.

[12] 常玉清,王小刚,王福利. 基于多神经网络模型的软测量方法及应用[J]. 东北大学学报,2005,26(6):519-522. Chang Y Q, Wang X G, Wang F L. Multi neural network method for soft sensing and its application[J]. Journal of Northeastern University, 2005, 26(6): 519-522.

[13] Ahmad Z, Zhang J. Combination of multiple neural networks using data fusion techniques for enhanced nonlinear process modelling[J]. Computers and Chemical Engineering, 2005, 30(2): 295-308.

[14] Ahmad Z, Pick Ha T, Mat Noor R A, Improving nonlinear process modeling using multiple neural network combination through Bayesian model averaging (BMA)[J]. IIUM Engineering Journal, 2008, 9(1): 19-36.

[15] Piuleac C G, Poulios I, Leon F, et al. Modeling methodology based on stacked neural networks applied to the photocatalytic degradation of triclopyr[J]. Separation Science and Technology, 2010, 45: 1644-1650.

[16] Jain A K, Murty M N, Flynn P J. Data clustering: A review[J]. ACM Computing Surveys, 1999, 31(3): 264-323.

[17] Xu R, Wunsch D. Survey of clustering algorithms[J]. IEEE Transactions on Neural Networks, 2005, 16(3): 645-678.

[18] Xie J Y, Jiang S, Xie W, et al. An efficient global k-means clustering algorithm[J]. Journal of Computers, 2011, 6(2): 271-279.

[19] Anil K. Jain. Data clustering: 50 years beyond k-means[J]. Pattern Recognition Letters, 2010, 31(8): 651-666.

[20] 李双虎,张风海. 一个新的聚类有效性分析指标[J]. 计算机工程与设计,2007,28(8):1772-1774. Li S H, Zhang F H. New index for clustering validation[J]. Computer Engineering and Design, 2007, 28(8): 1772-1774.

[21] Moody J, Darken C J. Fast learning in networks of locally-tuned processing units[J]. Neural Computation, 1989, 1(2): 281-294.

[22] Kim S, Vachtsevanos G J. An intelligent approach to integration and control of textile processes[J]. Information Sciences, 2000, 123(3/4): 181-199.