基于双阈值AdaBoost算法的4-CBA含量软测量建模

doi:10.11949/j.issn.0438-1157.20161609

摘要/Abstract

摘要：

针对PX氧化过程中4-CBA含量无法在线测量的问题，提出了一种基于双阈值更新样本权重的AdaBoost算法，该算法以BP神经网络作为弱学习器，采用轮盘赌方法根据样本权重在训练样本集中选择部分样本训练弱学习器，采用上一轮弱学习器的训练相对误差绝对值来更新所有训练样本的权重，在此基础上，用双阈值对样本误差范围进行划分，然后用不同的权重因子与原来的样本权值相乘实现样本权值的二次更新。该过程降低了含有大误差的样本的权值，增加了较大误差的样本的权值，从而减小了在下一轮训练过程中选到异常样本的概率。分别采用5种不同的方法并用实测的工业数据建立了4-CBA含量软测量模型，仿真结果表明用提出的改进AdaBoost算法建立的4-CBA含量软测量模型，其预测误差小于其他方法建立的模型误差。

关键词: AdaBoost算法, 软测量, 双阈值, 异常样本, 4-CBA含量, 轮盘赌方法

Abstract:

A modified AdaBoost algorithm with updating sample weight by dual threshold technique was proposed to model a soft sensor for estimating 4-CBA concentration, which could not be measured on-line in PX oxidation process. In this method, weak learners of BP neural networks were trained by part of samples selected by their weights and roulette wheel mechanism. The absolute values of last round training relative errors in weak learners were adopted to update weights of all training samples. Then, a second round updating on sample weights were completed by the product of original sample value and its weighting factor, which was defined by ratio of error range over dual thresholds. In the second updating process, weights were decreased for samples with gross errors but were increased for those with medium error. Consequently, probability of selecting outliers was reduced in following iteration of the training process. Five different methods were applied to model soft sensor of 4-CBA concentration with industrial data. Simulation results showed that the modified AdaBoost algorithm can improve soft sensor performance of 4-CBA concentration with predicting error less than that of other models.

Key words: AdaBoost algorithm, soft sensor, dual threshold technique, outliers, 4-CBA concentration, roulette wheel mechanism

中图分类号:

TP274

刘瑞兰, 刘树云, 戎舟, 江兵, 庞宗强. 基于双阈值AdaBoost算法的4-CBA含量软测量建模[J]. 化工学报, 2017, 68(5): 2009-2015.

LIU Ruilan, LIU Shuyun, RONG Zhou, JIANG Bing, PANG Zongqiang. Modeling soft sensor of 4-CBA concentration by AdaBoost algorithm with dual threshold technique[J]. CIESC Journal, 2017, 68(5): 2009-2015.

参考文献 30

[1]	冒永生, 杨开香, 王丽军. PTA装置氧化过程实时优化与先进控制[J]. 化工自动化及仪表, 2012, 39(9): 1128-1132. MAO Y S, YANG K X, WANG L J. Real-time optimization and advanced control for pta oxidation process[J]. Control & Instruments in Chemical Industry, 2012, 39(9): 1128-1132.
[2]	王丽军. PX氧化动力学研究及氧化反应器模拟[D]. 杭州: 浙江大学, 2001. WANG L J. Studies on the kinetics of p-xylene oxidation oxidation and the reactor simulation[D]. Hangzhou: Zhejiang University, 2001.
[3]	牟盛静. 石化工业过程建模与优化若干问题研究[D]. 杭州: 浙江大学, 2004. MOU S J. The studies of process modeling and optimization in petrochemical industry[D]. Hangzhou: Zhejiang University, 2004.
[4]	刘瑞兰, 牟盛静, 苏宏业, 等. 基于支持向量机和粒子群算法的软测量建模[J]. 控制理论与应用, 2006, 23(6): 895-900. LIU R L, MOU S J, SU H Y, et al. Modeling soft sensor based on support vector machine and particle swarm optimization algorithms[J]. Control Theory and Applications, 2006, 23(6): 895-900
[5]	颜学峰, 余娟, 钱锋. 基于自适应偏最小二乘回归的初顶石脑油干点软测量[J]. 化工学报, 2005, 56(8): 1151-1156. YANG X F, YU J, QIAN F. An evolution algorithm with select-best and prepotency operator and parameter estimation of 4-CBA model[J]. Journal of Chemical Industry and Engineering(China), 2005, 56(8): 1151-1156.
[6]	胡永有, 古勇, 苏宏业, 等. 基于BPANN的4-CBA软测量模型研究[J]. 仪器仪表学报, 2003, 24(3): 226-230. HU Y Y, GU Y, SU H Y, et al. The research of 4-CBA soft- sensor model based on BPANN[J]. Chinese Journal of Scientific Instruments, 2003, 24(3): 226-230.
[7]	陈渭泉, 刘瑞兰, 牟盛静, 等. 基于贝叶斯方法的4-CBA含量的软测量研究[J]. 化工自动化及仪表, 2003, 30(5): 49-51. CHEN W Q, LIU R L, MOU S J, et al. The research on soft sensor of the concentration of 4-CBA based on Bayesian approach[J]. Control & Instruments in Chemical Industry, 2003, 30(5): 49-51.
[8]	LIU R L, SU H Y, MOU S J, et al. Fuzzy neural network model of 4-CBA concentration for industrial PTA oxidation process[J]. Chinese Journal of Chemical Engineering, 2004, 12(2): 234-239.
[9]	DU W L, QIAN F, LIU M D, et al. 4-CBA soft sensor based on fuzzy CMAC neural networks[J]. Chinese Journal of Chemical Engineering, 2005, 13(3): 437-440.
[10]	郑小霞, 钱锋. 基于证据框架的最小二乘支持向量机在精对苯二甲酸生产中的应用[J]. 化工学报, 2006, 57(7): 1612-1616. ZENG X X, QIAN F. Application of least squares support vector machine within evidence framework in PTA process[J]. Journal of Chemical Industry and Engineering(China), 2006, 57(7): 1612- 1616.
[11]	刘瑞兰, 徐艳, 戎舟. 基于稀疏最小二乘支持向量机的软测量建模[J]. 化工学报, 2015, 66(4): 1402-1407. LIU R L, XU Y, RONG Z. Modeling soft sensor based on sparse least square support vector machine[J]. CIESC Journal, 2015, 66(4): 1402-1407.
[12]	ZANG Y, SU H Y, LIU R L. Fuzzy support vector regression model of 4-CBA concentration for industrial PTA oxidation process[J]. Chinese Journal of Chemical Engineering, 2005, 13(5): 642-648.
[13]	颜学峰. 基于MLFN-PLSR的PX氧化反应组合建模方法[J]. 化工学报, 2007, 58(1): 149-154. YAN X F. Delvelop p-xylene oxidation reaction model based on MLFN_PLSR[J]. Journal of Chemical Industry and Engineering(China), 2007, 58(1): 149-154 .
[14]	FREUND Y, SCHAPIRE R E. A decision-theoretic generation of online learning and an application to Boosting[J]. Journal of Computer and System Science, 1997, 55(1): 119-139.
[15]	DRUCKER H. Improving regressor using Boosting techniques[C]//Proc. of the 13th Annual Conf. on Computational Learning Theory. San Francisco, 1997: 208-219.
[16]	SOLOMATINE D P, SHRESTHA D L. AdaBoost.RT: a boosting algorithm for regression problem[C]//Proc. of the Int. Joint Conf. on Networks. Budapes, 2004: 1163-1168.
[17]	BAUMANN F, ERNST K, EHLERS A, et al. Symmetry enhanced AdaBoost[C]//International Conference on Advances in Visual Computing. Springer-Verlag, 2010: 286-295.
[18]	袁双, 吕赐兴. 基于PCA改进的快速AdaBoost算法研究[J]. 科学技术与工程, 2015, 15(29): 62-67. YUAN S, LÜ C X. Fast AdaBoost algorithm based on improved PCA[J]. Science Technology and Engineerring, 2015, 15 (29): 62-67.
[19]	查翔, 倪世宏, 张鹏. 关于AdaBoost.RT集成算法时间序列预测研究[J]. 计算机仿真, 2015, 32(9): 391-395. ZHA X, NI S H, ZHANG P. AdaBoost.RT intergrating prediction of time series based on adaptive and dynamic threshold[J]. Computer Simulation, 2015, 32(9): 391-395.
[20]	ZHANG P B, YANG Z X. A novel AdaBoost framework with robust threshold and structural optimization[J]. IEEE Trans Cybem., 2016, (99): 1-13.
[21]	胡国胜. 基于加权支持向量机与AdaBoost集成的预测模型研究[J]. 计算机应用与软件, 2012, 29(12): 280-281. HU G S. Study on forecasting model based on WSVR and AdaBoost[J]. Computer Applications and Software, 2012, 29(12): 280-281.
[22]	刘庆华, 丁文涛, 涂娟娟, 等. 优化BP_AdaBoost算法及其交通事件检测[J]. 同济大学学报(自然科学版), 2015, 43(12): 1829-1833. LIU Q H, DING W T, TU J J, et al. Improved BP_AdaBoost algorithm and its application in traffic incident detection[J]. Journal of Tongji University(Natural Science), 2015, 43(12): 1829-1833.
[23]	HU W M, HU W, MAYBNK S. AdaBoost-based algorithm for network intrusion detection[J]. Systems, Man, and Cybernetics, Part B: Cybernetics, IEEE Transactions on, 2008, 238(2): 577-583.
[24]	SHRESTHA D L, SOLOMATINE D P. Experiments with AdaBoost. RT, an improved boosting scheme for regression[J]. Neural Computation, 2006, 18(7): 1678-1710.
[25]	DEVON K B, SVEN F C. A comparison of AdaBoost algorithms for time series forecast combination[J]. International Journal of Forecasting, 2016, 32(4): 1103-1119.
[26]	HOLLAND J H. Adaptation in Natural and Aritificial Systems[M]. Ann. Arbor: University of Michigan Press, 1992: 126-137.
[27]	李希, 谢刚, 华卫琦. PTA技术国产化的主要化学工程问题及其研究思路[J]. 聚酯工业, 2001, 14(1): 1-7. LI X, XIE G, HUA W Q. Key problems and research program for PTA process domestic development[J]. Polyester Industry, 2001, 14(1): 1-7.
[28]	尹云华, 向阳, 刁磊, 等. PTA生产工艺及技术的研究进展[J]. 化工工业与工程技术, 2011, 32(5): 33-39. YIN Y H, XIANG Y, DIAO L, et al. Research progress of PTA production technology[J]. Journal of Chemical Industry &Engineering, 2011, 32(5): 33-39.
[29]	黄浩. 中国PTA行业的发展和现状[J]. 聚酯工业, 2016, 29(1): 1-2. HUANG H. Development and current situation of PTA industry in China[J]. Polyester Industry, 2016, 29(1): 1-2.
[30]	刘瑞兰, 戎舟. 工业PX氧化过程4-CBA含量的软测量[J]. 信息与控制, 2014, 43(3): 339-343. LIU R L, RONG Z. A soft sensor for 4-CBA soncentration in industrial PX oxidation processes[J]. Information and Control, 2014, 43(3): 339-343.