Co-reporter:Wenjian Zheng, Xuejin Gao, Yi Liu, Limei Wang, Jianguo Yang, Zengliang Gao
Chemometrics and Intelligent Laboratory Systems 2017 Volume 171(Volume 171) pp:
Publication Date(Web):15 December 2017
DOI:10.1016/j.chemolab.2017.10.009
•A novel semi-supervised nonlinear soft sensor for the Mooney viscosity prediction is developed.•It integrates extreme learning machine (ELM) and the graph Laplacian regularization into a unified modeling framework.•The useful information in unlabeled data are explored and utilized efficiently.•A bagging-based ensemble strategy is combined into semi-supervised ELM (SELM) to obtain more accurate predictions.•The industrial Mooney viscosity prediction results show its superiority.In industrial rubber mixing processes, the quality index (i.e., Mooney viscosity) cannot be online measured directly. Traditional data-driven empirical models for online prediction of the Mooney viscosity have not utilized the information hidden in lots of unlabeled data (e.g., process input variables during each mixing batch). A simple semi-supervised nonlinear soft sensor method for the Mooney viscosity prediction is developed. It integrates extreme learning machine (ELM) and the graph Laplacian regularization into a unified modeling framework. The useful information in unlabeled data can be explored and introduced into the prediction model. Furthermore, a bagging-based ensemble strategy is combined into semi-supervised ELM (SELM) to obtain more accurate predictions. The Mooney viscosity prediction in an industrial internal mixer exhibits its promising prediction performance of the proposed method by incorporating the information in unlabeled data efficiently.
Co-reporter:Hongying Deng, Yi Liu, Ping Li, Shengchang Zhang
Advances in Engineering Software 2017 Volume 114(Volume 114) pp:
Publication Date(Web):1 December 2017
DOI:10.1016/j.advengsoft.2017.07.007
•A novel empirical model is proposed to predict the multiple performance indices of the whole flow field using related impeller parameters of centrifugal pumps.•The complex nonlinearity relationship between multiple impeller parameters and performance indices can be described approximately.•It is demonstrated by the performance prediction of the whole flow field for the D82-19-2 centrifugal mine pump.•Compared with the computational fluid dynamics numerical simulation model, the higher prediction accuracy, more reliability prediction performance and less design time can be obtained.The relationship of multiple impeller parameters and performance indices is difficult to describe because of some unknown hydrodynamic phenomena. Modeling of performance indices of the whole flow field from impeller parameters often encounters some challenges, especially lower prediction accuracy in relatively small and large flow points, dependence on designers’ experience and time-consuming designing process. In this work, the least squares support vector regression (LSSVR) method is proposed to predict multiple pump performance indices of the whole flow field. To describe the performance more completely, the powder, the head, and the efficiency indices are chosen as the model outputs. Additionally, to improve the prediction accuracy and reduce the manufacture difficulty, nine impeller parameters and the flow rate are selected as the model inputs. With the LSSVR model, the complex nonlinearity relationship between multiple impeller parameters and performance indices can be described approximately. Moreover, the LSSVR model and the computational fluid dynamics numerical simulation model are applied to predict the powder, the head, and the efficiency of an actual centrifugal mine pump in the whole flow field. Compared with the performance test results, the superiority of the proposed method is demonstrated in terms of more accurate prediction performance and faster designing process.
Co-reporter:Yi Liu;Zengliang Gao
Journal of Applied Polymer Science 2015 Volume 132( Issue 22) pp:
Publication Date(Web):
DOI:10.1002/app.41958
ABSTRACT
Several data-driven soft sensors have been applied for online quality prediction in polymerization processes. However, industrial data samples often follow a non-Gaussian distribution and contain some outliers. Additionally, a single model is insufficient to capture all of the characteristics in multiple grades. In this study, the support vector clustering (SVC)-based outlier detection method was first used to better handle the nonlinearity and non-Gaussianity in data samples. Then, SVC was integrated into the just-in-time Gaussian process regression (JGPR) modeling method to enhance the prediction reliability. A similar data set with fewer outliers was constructed to build a more reliable local SVC–JGPR prediction model. Moreover, an ensemble strategy was proposed to combine several local SVC–JGPR models with the prediction uncertainty. Finally, the historical data set was updated repetitively in a reasonable way. The prediction results in the industrial polymerization process show the superiority of the proposed method in terms of prediction accuracy and reliability. © 2015 Wiley Periodicals, Inc. J. Appl. Polym. Sci. 2015, 132, 41958.
Co-reporter:Yi Liu;Zengliang Gao
Journal of Applied Polymer Science 2015 Volume 132( Issue 6) pp:
Publication Date(Web):
DOI:10.1002/app.41432
ABSTRACT
In internal rubber-mixing processes, data-driven soft sensors have become increasingly important for providing online measurements for the Mooney viscosity information. Nevertheless, the prediction uncertainty of the model has rarely been explored. Additionally, traditional viscosity prediction models are based on single models and, thus, may not be appropriate for complex processes with multiple recipes and shifting operating conditions. To address both problems simultaneously, we propose a new ensemble Gaussian process regression (EGPR)-based modeling method. First, several local Gaussian process regression (GPR) models were built with the training samples in each subclass. Then, the prediction uncertainty was adopted to evaluate the probabilistic relationship between the new test sample and several local GPR models. Moreover, the prediction value and the prediction variance was generated automatically with Bayesian inference. The prediction results in an industrial rubber-mixing process show the superiority of EGPR in terms of prediction accuracy and reliability. © 2014 Wiley Periodicals, Inc. J. Appl. Polym. Sci. 2015, 132, 41432.
Co-reporter:Yi Liu, Zengliang Gao, Ping Li, and Haiqing Wang
Industrial & Engineering Chemistry Research 2012 Volume 51(Issue 11) pp:4313-4327
Publication Date(Web):February 15, 2012
DOI:10.1021/ie201650u
An efficient nonlinear just-in-time learning (JITL) soft sensor method for online modeling of batch processes with uneven operating durations is proposed. A recursive least-squares support vector regression (RLSSVR) approach is combined with the JITL manner to model the nonlinearity of batch processes. The similarity between the query sample and the most relevant samples, including the weight of similarity and the size of the relevant set, can be chosen using a presented cumulative similarity factor. Then, the kernel parameters of the developed JITL-RLSSVR model structure can be determined adaptively using an efficient cross-validation strategy with low computational load. The soft sensor implement algorithm for batch processes is also developed. Both the batch-to-batch similarity and variation characteristics are taken into consideration to make the modeling procedure more practical. The superiority of the proposed soft sensor approach is demonstrated by predicting the concentrations of the active biomass and recombinant protein in the streptokinase fed-batch fermentation process, compared with other existing JITL-based and global soft sensors.
Co-reporter:Yi Liu, Junghui Chen
IFAC Proceedings Volumes (December 2013) Volume 46(Issue 32) pp:361-366
Publication Date(Web):1 December 2013
DOI:10.3182/20131218-3-IN-2045.00025
One significant challenge in nonlinear system identification development for industrial processes is that the modeling samples often contain outliers and unknown noise. In this paper, a novel Correntropy-based Kernel Learning (CKL) method is proposed for identification of nonlinear systems with such uncertainty. Without resort to unnecessary efforts, the CKL identification method can reduce the effects of outliers by the use of a robust nonlinear estimator that maximizes correntropy. The superiority of the proposed CKL method is demonstrated through identification of an industrial process in Taiwan. The benefit of its more accurate and reliable performance indicates that CKL is promising in practice for identification of nonlinear systems with unknown noise.
Co-reporter:Kun Chen, Yi Liu
IFAC Proceedings Volumes (2013) Volume 46(Issue 13) pp:359-364
Publication Date(Web):1 January 2013
DOI:10.3182/20130708-3-CN-2036.00041
Many bioprocesses are difficult to control due to their highly nonlinear and time-varying characteristics. To design simple and suitable controllers for these processes, a nonlinear predictive controller using sparse kernel learning with a polynomial kernel form is designed. First, the nonlinear time-varying processes can be identified using the recursive kernel learning method. The online kernel identification model can be efficiently updated, with nodes increasing and decreasing, via recursive learning algorithms. Consequently, the proposed polynomial kernel learning-based controller can restrict its complexity, and trace the time-varying characteristics of a nonlinear process adaptively to achieve better performance. The obtained results on a continuous bioreactor with time-varying parameters show that the proposed controller is superior to the traditional proportional-integral-derivative (PID) controller and other kernel controllers with an offline model without online updating.
