Saturday , August 18 2018

Generalized Predictive Controller for Ball Mill Grinding Circuit
in the Presence of Feed-grindability Variations

Sivanandam VENKATESH1, Kannan RAMKUMAR1, Muralidharan GURUPRASATH2, Seshadhri SRINIVASAN3, Valentina E. BALAS4*

1 School of EEE, SASTRA University,
Thirumalaisamudram, Thanjavur, 613401, India
esvee@eie.sastra.edu, ramkumar@eie.sastra.edu
2 FLSmidth Pvt. Ltd, Cement and Minerals Projects India,
Chennai, 603 103, India
guruprasath.muralidharan@flsmidth.com

3 International Research Center, Kalasalingam University,
Srivilliputtur, India
seshucontrol@gmail.com
4 Aurel Vlaicu University of Arad, Romania

* Corresponding author

Abstract: Feed-grindability variations (clinker hardness) in cement grinding circuit affects the product quality and productivity of the cement plant. This investigation proposes a generalized predictive controller for cement grinding circuit that is more robust to feed-grindability variations. The main building block of the proposed controller is a new model for cement grinding circuit that directly relates product quality with elevator current and main drive load. The advantage of the model is that the effect of feed-grindability variations on product quality can be easily observed from the output. To develop such a model, this investigation adapts a data driven modelling approach. Experimental data obtained as measurements from cement grinding circuit in a cement mill located near Chennai, India is used to develop the transfer function model based on least squares approach. The model obtained from data driven modelling is used to design a generalized predictive controller whose objective is to optimize the product quality in the presence of feed-grindability variations without breaching physical and operational limits of the cement grinding circuit. The tuning parameters of the proposed generalized predictive controller is adjusted to meet performance metrics specific to cement industries. Our results show that the proposed controller provides better product quality in the presence of feed-grindability variations than other optimization based controllers such as the linear quadratic regulator.

Keywords: Generalized Predictive Controller (GPC), cement grinding circuit, feed-grindability.

>>Full text<<
CITE THIS PAPER AS:
Sivanandam VENKATESH, Kannan RAMKUMAR, Muralidharan GURUPRASATH, Seshadhri SRINIVASAN, Valentina E. BALAS, Generalized Predictive Controller for Ball Mill Grinding Circuit in the Presence of Feed-grindability Variations, Studies in Informatics and Control, ISSN 1220-1766, vol. 25(1), pp. 29-38, 2016.

  1. Introduction

Feed-grindability in cement grinding process denotes the ability of the clinker (the processed limestone from rotary kiln) to be broken down into smaller particles. Feed-grindability variations occur due to lack of homogeneity in clinkers procured from different vendors. Current practice in cement industries is to maintain the separator power constant and vary the separator speed for maintaining fineness [1]. This conventional way of controlling the fineness becomes ineffective beyond a particular operating range of separator when there are feed-grindability variations. Therefore, cement industries need control approaches that assure product quality in the presence of feed-grindability variations. The multi-variable nature and lack of models that relate feed-grindability variations with product quality (and production) make realizing such a controller challenging.

In literature, several control methods are studied for cement grinding circuit. The available methods can be broadly classified as: (i) classical [2]-[3], (ii) optimal [4]-[5], (iii) predictive [6,8] and (iv) model-based controllers [9]. Classical controllers such as PID [2-3], state-feedback [7], and cascaded controllers [3] are used in cement industries for controlling the grinding process. Though, classical controllers are simple and cheap, they suffer from performance limitations arising from multi-variable interactions, model uncertainties, and actuator constraints. To overcome these shortcomings, linear quadratic (LQ) controllers were studied for cement grinding process in [4] and [5]. The LQ controllers showed better performance than classical controllers. However, the physical and operating constraints of the process were not considered in their design, thereby making their adaptation in cement industries difficult. Moreover, in these results the material accumulated inside the cement grinding mill that causes plugging phenomenon [6] has not been studied.

The investigations in [6] and [9] proposed a predictive controller for cement grinding circuit for handling the multivariable interactions and operating constraints. Results of these investigations showed that combining predictions with optimization driven control decisions leads to good tracking and regulatory performance. In spite of such advantages, their performance is limited by the model accuracy and uncertainties in the process. Motivated by this, Guruprasath et al. [1] studied the use of model predictive controllers (MPCs) for cement grinding circuit. The proposed controller was not only optimal, but also handled model complexities, and constraints, inherently in its design. Results of the investigation illustrated that the MPC provided better product quality and process performance. The performance enhancement is primarily due to the ability of MPCs to work closer to constraint boundaries and handle multi-variable interactions effectively. However, the proposed MPC approach cannot handle feed-grindability variations.

A review of literature reveals that although, several approaches for controlling cement grinding circuit are available, these approaches are less robust to feed-grindability variations resulting in poor product quality and production rate. To overcome the shortcoming of existing literature, this investigation aims to design a predictive optimal controller that maintains product quality amidst feed-grindability variations. Furthermore, the controller should consider the physical and operating constraints in its design for being easily adapted in cement industries.

To reach the objectives, this investigation proposes a new model for cement grinding circuit that varies separator power by maintaining separator speed at its optimum efficiency. Next, the proposed model is used to design generalized predictive controller (GPC). The choice of GPC is dictated by its wide-spread acceptance in process industries, ability to deal with model complexities such as multi-variable interactions, inherent constraint handling, and optimal performance. (see, [10]-[13] and references therein). Such properties are required for cement grinding circuit controllers to assure product quality in the presence of feed-grindability variations.

The main contributions of this investigation are: (i) a new data-based model for cement grinding circuit that relates elevator current and main drive load with product quality, (ii) GPC controller that uses the proposed model to optimize the product quality in the presence of feed-grindability variations, and (iii) illustration of the proposed controller in a cement industry and compare its performance with a linear quadratic regulator (LQR).

This paper is organized as follows. Section II discusses the ball mill cement grinding process. The idea behind the control strategy is discussed in section III. The data-driven transfer function model for the cement grinding circuit is presented in Section IV. Section V presents the GPC design for the cement grinding process. The performance of the proposed GPC is illustrated in Section VI. Conclusions and future prospects of the investigation are presented in Section VII.

REFERENCES

  1. PRASATH, G., B. RECKE, M. CHIDAMBARAM, J. B. JØRGENSEN, Soft Constrained Based MPC for Robust Control of a Cement Grinding Circuit, Preprints of the 10-th IFAC International Symposium on Dynamics and Control of Process Systems, The International Federation of Automatic Control, Mumbai, India, Dec. 18-20, 2013.
  2. POMERLEAU, A., D. HODOUIN, A. DESBIENS, E. GAGNON, A Survey of Grinding Circuit Control Methods: From Decentralized PID Controllers to Multivariable Predictive Controllers, Powder Technology, Elsevier, vol. 108, no. 2-3, 2000, pp. 103-115.
  3. BOULVIN, M., A. VANDE WOUVER, Modeling and Control of Cement Grinding Processes, IEEE Transactions On Control Systems Technology, IEEE, vol. 11, no 5, 2003, pp. 715-725.
  4. VAN BREUSEGEM, V., L. C, G. BASTIN, V. WERTZ, V. WERBROUCK, DE PIERPONT., An Industrial Application of Multivariable Linear Quadratic Control to a Cement Mill Circuit, IEEE Transactions On Industry Applications, IEEE, vol. 32, no. 3, 1996, pp. 670 -677.
  5. BOULVIN, M., C. RENOTTE, A. Vande WOUVER, M. REMY, S. TARASIEWICZ, P. CESAR, Modeling, Simulation and Evaluation of Control Loops for a Cement Grinding Process, European Journal of Control, Elsevier, vol. 5, 1999, pp. 10–18.
  6. MAGNI, L., G. BASTIN, V. WERTZ, Multivariable Non linear Predictive Control of Cement Mills, IEEE Transactions on Control Systems Technology, vol. 7, 1999, pp. 502-508.
  7. GROGNARD, F., F. JADOT, L. MAGNI, G. BASTIN, R. SEPULCHRE, V. WERTZ, Robust Stabilization of a Nonlinear Cement Mill Model, IEEE Transactions On Automatic Control, IEEE, vol. 46, no. 4, 2001, pp. 618-623.
  8. MARTIN, G., S. McGAREL, Nonlinear Mill Control, ISA Transactions, Elsevier, vol. 40, no. 4, 2001, pp. 369-379.
  9. LIU, Z., YANG, W. YAN, Intelligent Modeling And Predictive Control of Pre-Grinding System, Advanced Materials Research, Trans Tech Publications, vol. 433-440, 2012, pp. 2120-2127.
  10. WEN, S., J. ZHU, X., LI, S. H., CHEN, A Double-loop Structure in the Adaptive Generalized Predictive Control Algorithm for Control of Robot end-Point Contact Force, ISA Transactions, Elsevier, vol. 53, no. 5, 2014, pp. 1603-1608.
  11. YANFEI, Y., W. BAILIN, S. MINGHENG, Z. YONGQI, Polymerizer Temperature Cascade Control System Based on Generalized Predictive Control, Physics Procedia, Elsevier, vol. 24, Part A, 2012, pp. 184-189.
  12. CLARKE, D. W., C. MOHTADI, P. S. TUFFS, Generalized Predictive Control – Part I. The Basic Algorithm, Automatica, Elsevier, vol. 23, no. 2, 1987, pp. 137-148.
  13. MUTHUKUMAR, N., SESHADHRI SRINIVASAN, K., RANKUMAR, P., KAVITHA, V. E., BALAS, Supervisory GPC and Evolutionary PI Controller for Web Transport Systems, Acta Polytechnical Hungarica, accepted for publication, 2015.
  14. NING, Y.-Y., Z. WANG, M.-Z. YUAN, Z.-H. LI, Intelligent Control Based on Case-based Reasoning for Fineness of Raw, ZHONGNAN DAXUE XUEBAO(ZIRAN KEXUE BAN)/Journal of Central South University (Science and Technology), Central South University of Technology, vol. 42 (SUPPL. 1), 2011, pp. 918-923.
  15. ROSSITER, J. A., Model-based Predictive Control: A Practical Approach, Control Series, CRC Press, 2005.
  16. CAMACHO, E. F., C. BORDONS, Model Predictive Control, Springer-Verlag London Limited, Great Britain.

https://doi.org/10.24846/v25i1y201604