Beruflich Dokumente
Kultur Dokumente
Department of Chemical Engineering, National Institute of Technology Rourkela, India e-mail: seshu.chemical@gmail.com
2*
Department of Chemical Engineering, National Institute of Technology Rourkela, India e-mail: mkundu@nitrkl.ac.in, Ph No: 91-0661-24622634, Fax No: 91-0661-2462999
Abstract
This paper proposes a PCA (Principal component analysis) based control strategy that maintains the entire operation in subspace. The desired process region has utilized in controller design and acted as reference to current measurements of the process. The proposed controller was implemented on yeast fermentation bioreactor to achieve steady state ethanol concentration in the presence of ever changing external influences. The results have proven the effectiveness of proposed controller in rejecting the disturbances.
purpose and later its application was extended to the batch processes [13, 14]. MVSPC methods have undergone many amendments and advances for real time and online monitoring. Few attempts were made towards the application of MVSPC methods in control aspects. Piovoso et al. (1992a, 1994b) formulated a controller that operates the process in subspace by computing manipulated actions using the generated subspace error, the difference between the score set-points and current score data [15, 16]. The same concept in conjunction with Model Predictive Controller was applied to industrial processes; Distillation column, Tennessee Eastman and Maize drying [17, 18]. This approach is very useful when some of quality variables are not available for measuring. Later Palma et al. (2010, 2011) incorporated a gain in the same controller for effective setpoint tracking [19, 20]. Shah et al. proposed similar kind of control structure which uses least squares technique to calculate control moves that minimize the sum of squares of errors between score set-points, obtained from data collected under nominal operating condition, and current scores of industrial tubular reactor process [21]. A canonical variate analysis algorithm was employed by Akamatsu et al. to identify a state space model in subspace from the data collected in alpha-olefins production plant for the design of feedback and feed forward controllers to fulfill the control objectives [22]. To implementing PCA based controller on any process, SPE of current process must be below its threshold. This indicates that current process region is similar to the nominal operating region at which controller gets designed. PCA has facilitated the analysis of finding significant input among set of inputs and output of real time data of gas mixing chamber in Copper Smelter Plant that led to determination of SISO model, describes the actual MISO system [23]. It is evident that the process model or data is essential for design of advanced and statistical model based controllers. If neither the process model nor the data is available, it looks impossible to design a controller. This paper presents a simple PCA based control approach. Process dynamics is reflected in PCA's loading vectors. It requires only the steady state values of the process variables. The main objective of bioreactor considered in this work is to hold ethanol concentration regardless of undesirable effects. This task is accomplished with considerable degree of success by employing the proposed PCA based control scheme. The remaining part of the paper is arranged in the following way. Section 2 provides the theoretical postulations of PCA. The controller formulation is presented in section 3. Section 4 presents the process description and study of its dynamic behavior. The implementation of the proposed controller and its performance evaluation is provided in section 5. Section 6 concludes the article.
R=
X' X n 1
(1) (2) 2
R = VV '
P =V ( 1 : a )
(3)
T = XP
(4)
Controller Formulation
Considering the benefit of PCA i.e. examine the current process status by comparing the new measurements with existed model, and based on PCA based control concept presented in Pivoso et al., a simple control approach is proposed here. Let the data matrix, X s , be made up of steady state values of process variables of interest. Using PCA theory presented in section 2, necessary information about process variables (desired state) is extracted by means of loading matrix, P . This is being the reference model (process operation region) upon which new measurement vector (on-line data) available at each time instant is projected to verify whether the process current state is analogous to the desired state. The online measurement is projected onto lower dimensions defined by PCA's loadings. The variation between the desired process region and current process region is subspace error given by equation (7). Using subspace error, a feedback controller is developed that results in proportional controller, shown in equation (8), to find control moves to minimize the magnitude of difference between the current process operation and required operation. The block diagram of control system is depicted in figure (1).
PCA Controller
Process
y(k)
Fig. 1
[ P ,Tdes ] = PCA( X s )
(5)
t new = x new P
(6)
e subsp = Tdes t new
(7)
u( t ) = u s + e subsp K C
(8)
carbon dioxide. The contents of reactor are biomass, substrate and product. A cooling jacket surrounds the reactor to minimize the heat liberated from exothermic reaction taken place in the reactor. Table 1 presents the initial values of the variables of the system. The values of parameters, used in this model, and required information regarding the effect of temperature and inorganic salts on equilibrium concentration of oxygen in liquid phase and temperature dependent kinetic parameters can be found in Zoltan (2007).
Fago
Cs, CX, CP, Tr
Fagi
Fe, Cs, CX, CP, Tr Fig. 2 The nonlinear dynamic model comprises of differential equations from (8) to (14) represent the actual process. There are five input variables; substrate feed concentration, feed flow rate, feed temperature, coolant flow rate and coolant input temperature, and six states; concentration of biomass, substrate, product (ethanol) and dissolved oxygen, reactor temperature and cooling jacket temperature. Table 2 provides the list of states together with their steady state values. It is desired to maintain ethanol concentration at steady state by controlling the reactor temperature as the concentration is not directly measurable. Among five inputs the flow rate of cooling medium can serve as manipulated and substrate feed temperature & substrate feed concentration can be considered as disturbances and the rest are maintained at their respective steady states. The nonlinear model was simulated under two different conditions for studying the dynamic behavior of bioreactor. A unit step change is made in substrate feed temperature that causes the significant increase in ethanol concentration, C P , and reactor temperature, Tr , as shown by figs 3(a) & 3(b). The change in substrate feed concentration does not have a considerable impact on product concentration, C P , and reactor temperature, Tr , depicted in figs 4(a) & 4(b), compared to substrate feed temperature. Therefore substrate feed temperature is considered as major disturbance input.
dV = Fi Fe dt
(8)
CS dC X = X CX K +C dt S S
(9)
K PCP Fe CX e V
CS dC P = PC X dt K S1 + C S
(10)
CS dC S 1 = X CX K +C dt RSX S S
(11)
dC O2 dt
* = ( K 1a ) C O CO2 rO2 2
Fe CO2 V
(12)
dTag dt
Fag Vj
(14)
15 o C 18 Lh 1
Table 2
1 4.5939 gL
28.7188 o C 26.3472 o C
1 3 1 2 .8
C p
(a ) 1 2 .6 1 2 .4 1 2 .2 1 2 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0 1 0 0
3 0 .5 3 0
T r
(b ) 2 9 .5 2 9 2 8 .5 0
1 0
2 0
3 0
4 0
5 0
6 0
7 0
8 0
9 0
1 0 0
tim e
Fig. 3
1 2 .1 8 1 2 .1 7 8
C p
1 2 .1 7 6 1 2 .1 7 4 1 2 .1 7 2 1 2 .1 7 0 1 0 2 0 3 0 4 0 5 0 6 0 7 0 8 0 9 0
(a )
1 0 0
2 8 .7 1 8 8 (b ) 2 8 .7 1 8 8
T r
2 8 .7 1 8 8 2 8 .7 1 8 8 2 8 .7 1 8 8 0
1 0
2 0
3 0
4 0
5 0
6 0
7 0
8 0
9 0
1 0 0
tim e
Fig. 4
12.18 (d) 12.175 (a) 12.17 4.6 (b) 4.595 Co2 4.59 (e) 4.585 Cp
Cx
0.854 0.8535
50
100
150
200
250
300
50
100
150
200
250
300
33.43 33.425
Cs
33.42 33.415 0 28.76 (c) 28.74 28.72 28.7 50 100 150 200 250 300
50
100
150
200
250
300
27.5 (f) 27
Tag
0 50 100 150 time 200 250 300
Tr
26.5 26
50
100
150 time
200
250
300
Fig. 5
Conclusion
A PCA model was determined using the process variables in the desired region of yeast fermentation operation. Subspace error involving the new measurement vector at each instant were determined to check whether the current process region was similar to the desired region. If there was a variation, accordingly control moves were computed to force the current process region nearer to desired region. Bioreactor was simulated to acquire consistent production even in the presence of disturbances using the proposed controller. Though the reference model was linear and static, the proposed controller reasonably dealt with nonlinearities of bioreactor. Due to simplicity in 7
structure and ease of implementation, the proposed controller might be applicable to higher order and non-minimum phase systems.
References
1. Reddy GP, Chidambaram M (1994) Nonlinear control of bioreactors with input multiplicities. Bioprocess Eng 11: 97-100 Jyothi SN, Chidambaram M (2001) Nonlinear feedforward control of bioreactors with input multiplicities. Bioprocess Biosyst Eng 24: 123-129 Radhakrishnan TK, Sundaram S, Chidambaram M (1999) Non-linear control of continuous bioreactors. Bioprocess Eng 20: 173-178 Galluzzo M, Cosenza B, Matharu A (2008) Control of a nonlinear continuous bioreactor with bifurcation by a type-2 fuzzy logic controller. Compu Chem Eng 32 (12): 2986-2993 Constantino D, Pierre D, Claude F, Philippe L, Marc S (1995) Adaptive predictive control of dissolved oxygen concentration in a laboratory-scale bioreactor. J Biotechnol 43:21-32 Tham HJ, Ramachandran KB, Hussain MA (2003) Sliding Mode Control for a Continuous Bioreactor. Chem. Biochem. Eng. Q. 17 (4): 267275 Ramaswamy S, Cutright TJ, Qammar HK (2005) Control of a continuous bioreactor using model predictive control. Process Biochem 40: 27632770 Zhu G-Y, Zamamiri A, Henson MA, Hjortso MA (2000) Model predictive control of continuous yeast bioreactors using cell population balance models. Chem Eng Sci 55: 6155-6167 Zoltan KN (2007) Model based control of a yeast fermentation bioreactor using optimally designed artificial neural networks. Chem Eng J 127:95109
2.
3.
4.
5.
6.
7.
8.
9.
10. Jackson JE (2003) A Users Guide to Principal Components. Wiley, New York 11. Takros R, Gerigk M, Paschold H, Wandrey C (2001) Principal Component Analysis for Microbial LPhenylalanine Production. Bioproc Biosyst Eng 24: 93-99 12. Jackso JE, Mudholkar G (1979) Control procedures for residuals associated with principal component analysis. Technometrics 21: 341-349. 13. Kresta J, MacGregor JF, Marlin TE (1991) Multivariate statistical monitoring of process operating performance. Can J Chem 69 (1): 35-47 14. Nomikos P, MacGregor JF (1994) Monitoring Batch Processes using Multiway Principal Component Analysis. AIChE J 40(8): 13611375 15. Piovoso MJ, Kosanovich KA, Pearson PK (1991) Monitoring process performance in real-time. In Proc. of 1991 ACC Boston, Massachusetts: 721-724. 16. Piovoso MJ, Kosanovich KA (1994) Applications of Multivariate Statistical Methods to Process Monitoring and Controller Design. Int J Control 59 (3): 743-765.
17. Chen G, McAvoyy TJ, Piovoso MJ (1998) A multivariate statistical controller for on-line quality improvement. J Proc Cont 8 (2): 139-149 18. Liu X, Chen X, Wu W, Zhang Y (2006) Process control based on principal component analysis for maize drying. Food Control 17: 894899 19. Palma LB, Coito FV, Gil PS, Neves-Silva R (2010) Process Control based on PCA Models. 15th IEEE Int. Conf. on Emerging Technologies and Factory Automation, Univ. of the Basque Country, Bilbao, Spain. 20. Palma LB, Coito FV, Gil PS, Neves-Silva R (2011) Design of Adaptive PCA Controllers for SISO Systems. 18th IFAC World Congress Milano, Italy 21. Shah SL, Miller R, Takada H, Morinaga K, Satou T (1998) Modelling and control of a tubular reactor: A PCA based approach. Fifth IFAC symposium on dynamics and control of process systems, Corfu, Greece: 1722. 22. Akamatsu K, Lakshminarayanan S, Manako H, Takada H, Satou T, Shah S (2000) Data-based control of an industrial tubular reactor. Control Eng. Pract 8: 783-790. 23. Escano JM, Dorado F (2009) PCA based Pressure control of a Gas Mixing Chamber. IEEE Conference on Emerging Technologies & Factory Automation, Mallorca, Spain.
Figure Captions Fig. 1: PCA based control system Fig. 2: Bioreactor Fig. 3 Open loop response of bioreactor to unit step change in substrate feed temperature. Fig. 4 Open loop response of bioreactor to unit step change in substrate concentration Fig. 5 Disturbance rejection performance of PCA based and PID controllers.
10