# INTRODUCTION

iotechnological processes are relatively difficult objects for control. Their features have been discussed repeatedly. Among the most-widely used control models for Biotechnological Processes are the so called unstructured models, based on mass balance. In these models the biomass is accepted as homogeneous, without internal dynamic. Most widely used are models based on the description of the kinetic via the well known equation of Monod or some of its modifications (Neeleman, 2002;Galvanauskas, 1998;Staniskis, 1992Pirt, 1975;).

The use of the classical methods of the linear control theory is embarrassed, mainly due to the fact that the noise in the system is not of Gaussian or colored type. The changes of the values of the structural parameters of the Monod kinetics models also lead to bad estimates when using Kalman filtering (Diop, 2009). Another serious flaw is that using classical linearization and control solutions via the feeding rate, the linear system is not observable (Wang, 1987). In addition, the Monod kinetics models are characterized by another feature of the optimal control solutions. The dynamic optimization based on the Pontryagin maximum leads to singular optimal control problems (Alekseev, 1979;Krotov, 1973). The above problems have led to development of extended dynamical models in which the dynamic of the changes of the growth rate of the BTP is described by separate equation on the general differential equation. Such extended observable models based on Monod kinetics are the Monod-Wang and Wang-Yerusalimsky models used in the paper (Pavlov, 2008).

These characteristics of biotechnological processes and models have led to search for solutions via approaches and methods related to a wide range of contemporary mathematical areas (DeLisa, 2001;Levisauskas, 2003;Roeva, 2007;Bamieh, 2007;Diop, 2009). Such areas are adaptive systems, nonlinear systems, theory of variable structure systems and their main direction sliding mode control (SMC) (Emelyanov, 1993;Seli?teanu, 2007;Mohseni, 2009). In the last decade up to date methods and approaches in the areas of functional analysis, differential geometry and its modern applications in the areas of nonlinear control systems as reduction, equivalent diffeomorphic transformation to equivalent systems and optimal control have been used (Neeleman, 2002;Pavlov, 2001;Mahadevan, 2001;Bamieh, 2007;.

In the paper is proposed a new control solution based on a contemporary differential geometric approach for exact linearization of non-linear Monod type models. In our control solution are used observable model based on Monod kinetics, namely the Monod- 


# Global Journal of Computer Science and Technology Volume XI Issue XX Version I

One of the most important characteristics of biotechnological processes, which make the control design more difficult, is the change of cell population state. A serious obstacle is the existence of noise of non Gaussian type. This type of noise appears in the measurement process as well as in the process of the determination of the structure parameters of the model. But may be the most serious obstacle is provoked by the differences in the rate of changes of the elements of the state space vector of the control system. Combined with the strong nonlinearity of the control system of the Monod type this feature of the control system leads to numerical instability or to unsatisfactory performance of the control algorithms (Neeleman, 2002;Tzonkov, 2006;Roeva, 2007). these models a control design for optimal control and stabilization of the specific growth rate of fed-batch biotechnological processes is presented. The control is written based on information of the growth rate (Neeleman, 2002).


# II. MONOD AND MONOD-WANG BIOTECHNOLOGICAL MODELS

Most widely used are models based on the description of the kinetic via the well known equation of Monod or some of its modifications. The rates of cell growth, sugar consumption, concentration in a yeast fed-batch growth are commonly described as follows:
F V S S V F X S s K S k S X V F X S s K S X m m = ? + ? = ? + = ? + ? ? ) ( 0 µ µ (1)
This differential equation is often part of more general and complex dynamic models. Here X is the concentration of the biomass, S is the substrate concentration, V is the volume of the bioreactor. The maximal growth rate is denoted by ? m and K S is the coefficient of Michaelis-Menten. With k we denote a constant typical for the corresponding process. The feeding rate is denoted by F. If the process is continuous (F/V) is substituted by the control D, the dilution rate of the biotechnological process (BTP). The third equation is dropped off. Often used models are described in table 1 (Staniskis, 1992;Zelic, 2004;Galvanauskas, 1998;Tzonkov, 2006). The first model is the well known Monod type model (Roeva, 2004(Roeva, , 2007)), the second is the Yerusalimsky model (Galvanauskas, 1998) and the third model includes inhibition term in the denominator. The non-observability of the Monod model has led to the development of the widened dynamical models, in which the dynamics of the specific growth rate of the BTP is described via separate equation in the system of differential equations. The dynamics of the growth rate µ in the Monod-Wang model is modeled as a first order lag process with rate constant m, in response to the deviation in the growth rate. This model called also the model of Monod-Wang determines a linear observable system in the classical linearization (Wang, 1987;Pavlov, 2007).  
F V S s K S m S S V F X k S X x F X X m = ? + = ? ? = ? = ? ? + ? ? ) ( ) ( 0 4 µ µ µ µ µ (2)V F ?X k A F V S s K S ? m V F S So X k S X V F ?X X m ? == = ? + = ? + ? = ? = ? ? ? ? ? ) ( µ µ µ (3)( , ) ( , 3 A
Here X denotes the concentration of biomass, [g/l]; S -the concentration of substrate (glucose), [g/l]; V -bioreactor volume, [l]; F -substrate feed rate, [h -1 ]; S 0 -substrate concentration in the feed, [g/l]; µ maxmaximum specific growth rate, [h -1 ]; K S -saturation constant, [g/l]; k and k 3 -yield coefficients, [g/g], mrate coefficient [-];. The dynamics of µ in the Monod-Wang model is modeled as a first order lag process with rate constant m, in response to the deviation in µ. The last equation describes the production of acetate (A). This equation is dynamically equivalent to the first one after the implementation of a simple transformation (X = (1/k 3 ) A). That is why we replace A with X in Yerusalimsky model. The best description is given by the so called model of Wang-Yerusalimsky:
. , , ) ( ) ( , ) ( , 3 A V F ?X k A F V X k k S s K S ? m V F S So X k S X V F ?X X i i m ? == = ? + = ? + ? = ? = ? ? + ? ? ? ) ( µ µ µ (4)
In the formula X is the concentration of biomass, [g/l]; S-the concentration of substrate (glucose),
[g/l]; V-bioreactor volume, [l]; F-substrate feed rate (control input), [h -1 ]; S 0 -substrate concentration in the feed, [g/l]; µ max -maximum specific growth rate, [h -1 ]; K S -saturation constant, [g/l]; k, k 3 - constants, [g/g]; m-coefficient [-]; E-the concentration of ethanol, [g/l]; A-the concentration of acetate [g/l].
The system parameters are as follows:
µ m =0.59 [h -1 ], K S =0.045 [g/l], m=3 [-], S 0 =100 [g/l], k=1/Y S/X ,, k=2 [-], k 3 =1/Y A/X , k 3 =53 [-], k i =50 [-], F max = 0.19 [h -1 ], V max =1.5 [ ]
. These data desribed an E. Coli process (Cockshott,   Specific Growth Rate and Sliding Mode Stabilization of Fed-Batch Processes Figure2 : Specific growth rate using Monod model ( 1) and Wang-Monod model (2). 1999) and are chosen close to data in table 2 (Roeva, 2004(Roeva, , 2007)):
Parameter Model 1 Model 2 Model 3 max µ , [h -1 ] 0,55 0,52 0,54 S k , [gl -1 ] 0,039 0,027 0,029 i k , [gl -1 ] 51,3 53,6 50,8 / S X Y , [gg -1 ] 0,501 0,498 0,497 / A X Y , [gg -1 ] 0,015 0,015 0,015
The last equation describes the production of acetate (A). This equation is dynamically equivalent to the first one after the implementation of a simple transformation (X = (1/k 3 )A). The initial values of the state variables are: X(0)=0.99; S(0)=0.01; µ(0)=0.1; A(0) = 0,03; V(0)=0. 5.

The following mathematical condition (k E ??) determines the Wang-Monod model as a restricted form of the Wang-Yerusalimsky model ( 4). The Monod model is a singular form of Wang-Monod model obtained by omission of the third equation. That is why the Wang-Yerusalimsky model is a more general model form.

Interesting moment is that these models are dynamically equivalent to the following Brunovsky normal form (Pavlov, 2001(Pavlov, , 2004(Pavlov, , 2007)):
W Y Y Y Y Y = = = ? ? ? 3 3 2 2 1 (5)
Here by W is noted the control input. This model is linear. The non-linearity of model ( 1) is transformed and included in the input function W (Montseny, 2008;   Bamieh, 2007; Elkin, 1999; Gardner, 1992). The input function W depends from the space vector of model (1)   and that has to be underlined because this is a limitation of the application of the Pontryagin maximum principle (Alekseev, 1979;Krotov, 1973;Hsu, 1972). Different diffeomorfic transformations of the Monod, Wang and Yerusalimsky models are analyzed in details in the following papers (Pavlov, 2001(Pavlov, , 2004(Pavlov, , 2007(Pavlov, , 2008)). The Brunovsky form is a linear model and permits easy optimal control solutions with application of the Pontryagin's maximum principle.

The complexity of the biotechnological systems and their singularities make them difficult objects for control. They are difficult to control also because of the fact that it is difficult to determine their optimal technological parameters. These parameters can depend on very complicated technological, ecological or economical market factors. Their taking into account in one mathematical model directly is impossible for the time being. Because of this reason often in practice expert estimates are used. From outside the estimates are expressed only by the qualitative preferences of the Biotechnologist. The preferences themselves are in rank scale and bring the internal indetermination, the uncertainty of the qualitative expression, which is a general characteristic of human thinking. Because of this reason here the mathematical models from the Utility theory and stochastic programming can be used ( Kivinen, 2004;Fishburn, 1970;Keeney, 1993;Aizerman, 1970).

Thus the incomplete information usually is compensated with the participation of imprecise human estimations. Our experience is that the human estimation of the process parameters of a cultivation process contains uncertainty at the rate of [10,30] %.

Here is used a mathematical approach for elimination of the uncertainty in the DM's preferences based both on the Utility theory and on the Stochastic programming (Pavlov, 2010(Pavlov, , 2011)). The algorithmic approach permits exact mathematical evaluation of the optimal specific growth rate of the fed-batch cultivation process according to the DM point of view even though the expert thinking is qualitative and pierced by uncertainty. The assessed utility criteria are shown on the following figure 4: Figure 4 : Expert utility versus Growth rate Thus we achieve totally analytical mathematical description of the complex system "Technologistbiotechnological process " (Pavlov, 2010(Pavlov, , 2011)).
III.

# OPTIMAL CONTROL AND STABILIZATION OF THE GROWTH RATE

The presentation of the control design follows the presentations in papers (Pavlov, 2004(Pavlov, , 2007(Pavlov, , 2008)). We use the general model (Technologistbiotechnological process). We preserve the notation U(.)

Specific Growth Rate and Sliding Mode Stabilization of Fed-Batch Processes

Table 2 for the DM utility function (Pavlov, 2010;Fishburn, 1970;Keeney, 1993). The control design of the fed-batch process is based on the next subsidiary optimal control problem: 
Max(U(µ(T int ))),
), ( ,
3 2 1 2 1 1 3 2 2 3 1 2 2 1 1 2 3 3 2 1 1 3 2 1 1 ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? = ? ? ? ? ? ? ? ? ? ? ? ? + + + ? = ? = = µ µ µ µ µ S S S X S X u S X u S X u ku u u u s K u m u k ku u u Y ku u u Y u Y o m (8)
The derivative of the function Y 3 determines the interconnection between W-model ( 7) and D-model (6).

The control design is a design based on the Brunovsky normal form and application of the Pontrjagin's maximum principle step by step for sufficiently small time periods T. The optimal control law has the analytical form (Pavlov, 2007):
( ) . 0 , 0 ) ( , 0 , 1 ) ( : 1 2 ) 2 1 ( ) ( , max 1 6 1 ) 1 ( ? = > = ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? = ? = ? r r sign r r sign where D kY t T t T ic sign D i i i opt µ µ (9)
The optimal control law of Wang-Yerusalimsky model (4) has the same form (Pavlov, 2008).This type of control may be used only for cumulative criteria for which the Bellman principle is valid in the optimal control (Hsu, 1972). For example, such are the amount of biomass at the end of the process and the timeminimization optimal control. The sum is the derivative of the utility function U(µ).The time interval T can be the step of discretization of the differential equation solver.

It is clear that the "time-minimization" control is determined from the sign of the utility derivative. Thus, the control input is D=D max or D=0. The solution is a "time-minimization" control (if the time period T is sufficiently small) (Pavlov, 2004). The control brings the system back to the set point for minimal time in any case of specific growth rate deviations. The demonstration is shown in (Pavlov, 2007). The previous solution permits easy determination of the control stabilization of the fed-batch process Max(U(µ(T int ))), where the variable µ is the specific growth rate, (µ?[0, µ max ], F?[0, F max ]). Here U(µ) is the utility function in figure (3) and F is the control input (the substrate feed rate):

. The control law is based on the solution of the next optimization problem: The Brunovsky normal form of Wang-Yerusalimsky model ( 4) is the same (Pavlov, 2008). Here Fig. 6 : Optimal profile After the stabilization of the system in equivalent sliding mode control position the system can be maintained around the optimal parameters with sliding mode control (Fig. 5, 6). Possible solution in sliding mode is alternation of ? m (as a function of the temperature and the acidity in the bioreactor) or alternation of F (Pirt, 1975; Pavlov, 2007). The iterative utility function design and the iterative corrections in the DM preferences permit adjustment of the control law and of the optimal control final results in agreement with the changes in the opinion of biotechnologist. The procedure could be interprets as learning procedure in the two opposite directions, in direction to biotechnologist or in direction to the final optimal solution.
F V S s K S m V F S So X k S X V F X X F F T t T U m = ? + = ? + ? = ? = ? ? ? ? ? ? ? ) ) ( ( ) ( ] , 0 [ ], , 0 [ ], ,0

# IV. MATHEMATICAL PROBLEMS ARISING FROM THE SLIDING MODE CONTROL

Control solution results are shown in control systems based on the solutions of variable structure systems (Seli?teanu, 2007;Mohseni, 2009). Here a different problem arises. Good control via sliding mode control is possible when the system is led to the initial position in a point from the area of the "equivalent control solution" (Pavlov, 2007(Pavlov, , 2008)). This is a specific task for fed-batch control via information about the growth rate of the biomass.

A common manifestation in sliding mode control is some overregulations of the biotechnological process. Such overregulations are shown in figures (7 and 8). In case of start in sliding mode in system conditions different from the area of "equivalent sliding mode control" then the process arrives in some over regulations (Utkin, 1981). We conclude that the control law (9) bring the system to the optimal point (optimal growth rate) with a" time minimization" control, starting from any deviation point of the specific growth rate (Fig. 5). Thus, we design the next control law:

1. At the interval [0, t 1 ] the control is "timeminimization" control (9), where ?(t 1 )=(x 30 -?), ?>0, x 30 =max(U(µ)). D is replaced with F=?F max , 1??>0, when D=D max . The choice of ? depends on the step of the equation solver and is not a part of the optimization (here ?=0.123); 2. At the interval [t 1 , t 2 ] the control is F=0 ( ?(t 1 )=(x 30 -?), ?(t 2 )=x 30 -to be escaped an overregulation); 3. After this moment the control is the control (9) with F=?F max , when D =D max (chattering control with 1??>0).

The deviation of the fed-batch process with this control is shown on figures (5,6). 


# BIOMASS CONCENTRATION X [g/l]

Specific Growth Rate and Sliding Mode Stabilization of Fed-Batch Processes contemporary mathematical areas. Such areas are differential geometry and its modern applications in the areas of nonlinear control systems as diffeomorphic transformation to equivalent systems. Solutions that fix the system in "equivalent control" position, staring from any initial positions are showed in (Fig. 9, 10). Detailed descriptions of such controls are discussed in (Pavlov, 2007(Pavlov, , 2008)). In the last decade up to date methods and approaches in the areas of functional analysis, differential geometry and its modern applications in the areas of nonlinear control systems as These characteristics of biotechnological processes and models have led to search for solutions via approaches and methods related to a wide range of These characteristics of biotechnological processes table new mathematical control problems for discussion and resolution. The overregulations in SMC are provoked by the differences in the rate of changes of the elements of the state space vector of the control system. It is needed control solution that fixes the fixation of the biomass 


# SLIDING MODE CONTROL AND STABILIZATION OF THE FED-BATCH PROCESS

The sliding mode control is a good solution for stabilization under varying conditions (parameters systems have been used for surmount the discussed difficulties (Pavlov, 2005;Bamieh, 2007;Diop, 2009). deviations, noises etc.) (Emelyanov, 1993(Emelyanov, , 1996;;Utkin 1987;Seli?teanu, 2007). In the paper is demonstrated a sliding mode control for stabilization of the specific growth rate in "the best" growth rate (Pavlov, 2007).

The feeding rate F(t) is derived from the substrate concentration:
F(t)=(kX(t)µ(t)V(t)/(So-S e )
, where X(.) is the quantity of biomass in the bioreactor. Deviations of the system in sliding mode are showed in Fig. (11,12).

Specific Growth Rate and Sliding Mode Stabilization of Fed-Batch Processes The value ? m 2 is a sufficiently small supplementary value. This SM control law eliminates the deviations of the parameters, noises and structure modifications. This solution overcomes successfully some of the difficulties mentioned above.  (Pirt, 1975) . This control gives us the possibility to use the temperature (t o ) and the acidity (pH) as input control values. More classical SM solutions with substrate concentration S as control value could be seen in the literature (Seli?teanu, 2007;Mohseni, 2009).

The sliding affine subspace is defined by the equation For some types of BTP it is possible to choose the control via the temperature and/or via changing the acidity in the bioreactor (Pirt, 1975). What we mean is indirect influence on the biomass maximum growth rate, which, according to the contemporary researches, is one of the main factors, determining the quality of the cultivation process (Neeleman, 2002). Here the problem about the observability of the control system arises again, because this directly concerns the possibility for satisfactory identification of the state space vector of the control system and more specifically for the determination of the specific growth rate of the BTP. This group of questions is still a topical branch in the theory of the biotechnological control systems (Diop, 2009). The devolving from the "discontinuity" to "continuity" in sliding mode control not only raise the control quality but permits to solve control problems with smaller quantity of priory information (Emelyanov, 1993(Emelyanov, , 1996)).

By now we have reached a full mathematical description of the complex system "Biotechnologist-fed-  Here is used the so-called "contraction" algorithm [5]. After Emelyanov the second order SM control input in the "contraction" algorithm becomes:  (15) It is known that this algorithm ends for finite time (Emelyanov, 1993(Emelyanov, , 1996)). The input in second order SM is smoother but the control becomes is more imprecise. The performances of the system with this SM control are showed in Fig. (13,14)  (Pavlov, 2007(Pavlov, , 2008)). 
)) 59 . 0 )( ) ( ? ((3

# Se

The Russian scientists Emelyanov, Korovin and Levant evolve high-order sliding mode methods in control systems (Emelyanov, 1993(Emelyanov, , 1996)). We propose in our investigation a second order sliding mode control following Emelyanov and Korovin. The control algorithms of second order are used so that the system deviations become cooler but a little more imprecise. Out of this approach the second order SM manifold becomes: batch process". We have overcome the restrictions we have overcome the obstacles with the singularities of the optimal control via finding and using of the Brunovsky normal form of the differential equation. The system was led to the working point from the "equivalent connected with the observability of the Monod kinetics; control in sliding mode" smoothly and stabilized in the optimal specific growth rate position (Pavlov, 2007). The solution and the determination of the optimal profile was The inclusion of a value model as objective function as part of a dynamical system reached a full mathematical description of the complex system "Biotechnologist-fed-batch process". This is done with the use of the expected utility theory and the stochastic programming. Such a good objective function would allow the user to vary iteratively his value judgments and to correct iteratively the control law in agreement with his value judgments.


# VI. CONCLUSION

A methodology for specific growth rate optimal control is developed. This approach aims utilization of control based only on specific growth rate measurement. In the paper are investigated the possibilities of the second order sliding mode. We have overcome the restrictions connected with the observability of the Monod kinetics through Monod-Wang and Wang-Yerusalimsky models; we have overcome the obstacles with the singularities of the optimal control via finding and using of the Brunovsky normal form of the differential equation. The system was led to the working point from the "equivalent control in sliding mode" smoothly and stabilized in the optimal specific growth rate position. This solution permits to win through difficulties arising from the biotechnological peculiarities in order to be obtained good control solutions.

done via synchronized usage of several mathematical approaches for modeling, reduction of nonlinear system, application of the Pontryagin maximum principle.
![biotechnological process. Obviously model (1) is a singular form of model (2). The comparison of both models is shown in Figures (1, 2, 3):One general description of the fed-batch biotechnological process looks like(Pavlov, 2007).](image-2.png "")
3![Figure 3 : Feeding rate using Monod model (1) and Wang-Monod model (2).](image-3.png "Figure 3 :")
![where the variable µ is the specific growth rate, (µ?[0, µ max ], D?[0, D max ]). Here U(µ) is an aggregation objective function (the utility function -fig.4) and D is the control input (the dilution rate): © 2011 Global Journals Inc. (US) Global Journal of Computer Science and Technology Volume XI Issue XX Version I 22 2011 December Specific Growth Rate and Sliding Mode Stabilization of Fed-Batch Processes . ) ,](image-4.png "")
![Global Journal of Computer Science and Technology Volume XI Issue XX Version I T int is sufficiently small the optimization is in fact "time minimization". The differential equation in(6) describes a continuous fermentation process. The model permits exact linearization to the next Brunovsky normal form (Goursat, as regard to the differential forms)(Gardner, 1992;Elkin, 1999;Pavlov 2001):](image-5.png "")
7![Fig.7 : Continuous process -overregulation of the biomass in SMC](image-6.png "Fig. 7 :")
519![Fig. 5 : Stabilization of the fed-batch process1](image-7.png "Fig. 5 : 1 Fig. 9 :")
10![Fig.10 : Fed batch process, Chattering control: growth rate (1); feed rate control (2)](image-8.png "Fig. 10 :")
5![2011 Global Journals Inc. (US) Global Journal of Computer Science and Technology Volume XI Issue XX Version I 25 2011 December reduction, equivalent transformation to equivalent system in "equivalent control" position, staring from any initial positions.](image-9.png "5 ©")
8![Fig.8 : Fed batch process-overregulation of the growth rate in SMC](image-10.png "Fig. 8 :")






1?1max µ(kS s +S)2max µ(k sS +S)(k ik i +A)3max µ(ks+SS +S2/ki)
0.450.40.350.30.250.20.150.10.050024681012141618Time [h]10.90.80.70.60.50.40.30.20.10024681012141618Time [h]

			2011 DecemberSummary -
			December
			© 2011 Global Journals Inc. (US) Global Journal of Computer Science and Technology Volume XI Issue XX Version I 21 2011 December
			© 2011 Global Journals Inc. (US) Global Journal of Computer Science and Technology Volume XI Issue XX Version I 28 2011 December
			© 2011 Global Journals Inc. (US) Global Journal of Computer Science and Technology Volume XI Issue XX Version I 29 2011 December
		
		
			

				


* 
	
		Potential Function Method in the Theory of Machine Learning
		
			MAizerman
		
		
			EBraverman
		
		
			LRozonoer
		
		
			1970. 1970
			Moscow, Nauka
		
	




* 
	
		Optimal Control
		
			VAlekseev
		
		
			VTihomirov
		
		
			SFomin
		
		
			1979. 1979
			Moscow, Nauka
		
	




* 
	
		On Discovering Low Order Models in Biochemical Reaction Kinetics
		
			LBamieh
		
		
			2007. 2007
		
	




* 
	
		
	
	
		American Control Conference
		
			July 11-13. 2007
		
	




* 
	
		Modelling the Effects of Glucose Feeding on a Recombinant E. coli Fermentation
		
			ARCockshott
		
		
			ID LBogle
		
	
	
		Bioprocess Engineering
		
			20
			
			1999
		
	




* 
	
		Generic Model Control of Induced Protein Expression in High Cell Density Cultivation of Escherichia coli Using On-line GFPfusion Monitoring
		
			DelisaM
		
		
			HChae
		
		
			WWeigand
		
		
			JValdes
		
		
			GRao
		
		
			WBentley
		
	
	
		Bioprocess and Biosystems Engineering
		
			24
			
			2001
		
	




* 
	
		On the Biomass Specific Growth Rates Estimation for Anaerobic Digestion using Differential Algebraic Techniques
		
			SDiop
		
		
			ISimeonov
		
	
	
		International Journal Bioautomation. V
		
			13
			
			2009
		
	




* 
	
		Reduction of Non-linear Control Systems: A Diferential Geometric Approach
		
			VElkin
		
	
	
		Mathematics and its Applications
		
			472
			1999
		
	




* 
	
		High-order sliding modes in control systems
		
			SEmelyanov
		
		
			SKorovin
		
		
			ALevant
		
	
	
		Computational mathematics and modelling
		
			7
			3
			
			1996. 1996
		
	




* 
	
		Higherorder Sliding Modes in Control Systems
		
			SEmelyanov
		
		
			SKorovin
		
		
			ALevant
		
		
	
	
		Diferential Equations
		
			29
			11
			
			1993. 1993
		
	




* 
	
		Utility Theory for Decision-Making
		
			PFishburn
		
	
	
		Proceedings
		
			1970
		
	




* 
	
		Model Based Design of a Biochemical Cultivation Process
		
			VGalvanauskas
		
		
			RSimutis
		
		
			NVolk
		
		
			ALubbert
		
	
	
		Bioprocess Engineering
		
			18
			
			1998
		
	




* 
	
		The GS algorithm for exact linearization to Brunovsky normal form
		
			RGardner
		
		
			WShadwick
		
	
	
		IEEE Trans. Autom. Contro
		
			37
			2
			
			1992
		
	




* 
	
		
			JHsu
		
		
			AMeyer
		
		Modern. Control Principles and Applications
				New York
		
			McGRAW-HILL
			1972
		
	




* 
	
		Decision with Multiple Objectives: Preferences and Value Tradeofs
		
			RKeeney
		
		
			HRaiffa
		
		
			1993. 1976. 1993
			Cambridge University Press
			Cambridge & New York
		
	




* 
	
		Online Learning with Kernels
		
			JKivinen
		
		
			ASmola
		
		
			RWilliamson
		
	
	
		IEEE Transactions on Signal Processing
		
			52
			8
			2004. August 2004
		
	




* 
	
		Methods and Problems in the Optimal Control
		
			VKrotov
		
		
			VGurman
		
		
			1973. 1973
			Moscow, Nauka
		
	




* 
	
		Model-based Optimization of Viral Capsid Protein Production in Fed-batch Culture of recombinant Escherichia coli
		
			DLevisauskas
		
		
			VGalvanauskas
		
		
			SHenrich
		
		
			KWilhelm
		
		
			NVolk
		
		
			ALubbert
		
	
	
		Bioprocess and Biosystems Engineering
		
			25
			
			2003
		
	




* 
	
		Differential flatness based nonlinear predictive control of fedbatch bioreactors
		
			RMahadevan
		
		
			SAgrawal
		
	
	
		Control Engineering Practice
		
			9
			8
			
			2001. August 2001
		
	




* 
	
		Design of sliding mode controller for the optimal control of fed-batch cultivation of recombinant E. Coli
		
			SMohseni
		
		
			VBabaeipour
		
		
			AVali
		
	
	
		Chemical Engineering Science. V
		
			64
			21
			
			2009
			Elsevier
		
	




* 
	
		Operatorial Parametrizing of Controlled Dynamic Systems-Application to the Fed-Batch Bioreactor Control Problem
		
			EMontseny
		
		
			ADoncescu
		
		
			2008
			17th World Congress The International Federation of Automatic Control
			
			Seoul, Korea
		
	




* 
	
		Sliding mode and adaptive sliding mode control of a class of nonlinear bioprocesses
		
			DSeli?teanu
		
		
			EPetre
		
		
			VR?svan
		
		
	
	
		Int. J. Adaptive Control and Signal Processing. V
		
			21
			
			2007
			Wiley InterScience
		
	




* 
	
		Automation of Biotechnological Process: measurements, optimization and control
		
			J.-KStaniskis
		
		
			DLevisauskas
		
		
			USmutis
		
		
			MViesturs
		
		
			Kristapsons
		
		
			1992. 1992
			Zinatne, Riga
		
	




* 
	
		Functional State Approach to Fermentation Process Modelling, Prof. Marin Drinov Academic Poublishing House
		
			Tzonkov
		
		
			BSt
		
		
			Hitzmann
		
		
			2006. 2006
			Sofia, Bulgaria
		
	




* 
	
		Sliding Mode and Applications in Variable Structure Control
		
			VUtkin
		
	
	
		Nauka
		
			1981. 1981
		
	
	in Russian




* 
	
		Application of a Robust Multivariable Controller to Non-linear Bacterial Growth Systems
		
			TWang
		
		
			CMoore
		
		
			DBirdwell
		
		
	
	
		Proc. of the 10-th IFAC Congress on Automatic Control
				of the 10-th IFAC Congress on Automatic ControlMunich
		
			1987
			
		
	




* 
	
		Modeling of the Pyruvate Production with Escherichia coli in a Fed-batch Bioreactor
		
			BZelic
		
		
			DVasic-Racki
		
		
			CWandrey
		
		
			RTakors
		
	
	
		Bioprocess and Biosystems Engineering
		
			26
			
			2004
		
	




* 
	
		Reduction of Complexity via Operatorial Parametric Formulations for Some Nonlinear Dynamic Problems of Biology
		
			EMontseny
		
		
			ADoncescu
		
		
	
	
		22nd International Conference on Advanced Information Networking and Applications -Workshops (aina workshops 2008)
				Netherlands
		
			Neeleman R
			2008. 2002. 2002
			231
			22
		
		
			Thesis, Wageningen University
		
	
	Biomass Performance: Monitoring and Control in Biopharmaceutical production




* 
	
		Exact Linearization of a Non Linear Biotechnological Model /Brunovsky Model/, Comptes Rendus de l'Academie Bulgares des Sciences
		
			YPavlov
		
		
	
	
		Proceedings of Bulgarian Academy of Sciences), N10
				Bulgarian Academy of Sciences), N10
		
			2001. 2001
			
		
	




* 
	
		Brunovsky Normal Form of Monod Kinetics and Growth Rate Control of a Fed-batch Cultivation Process
		
			YPavlov
		
		
	
	
		Online journal Bioautomation
		
			
			2007. 2007
		
	




* 
	
		Equivalent Forms of Wang-Yerusalimsky Kinetic Model and Optimal Growth Rate Control of Fed-batch Cultivation Processes
		
			YPavlov
		
	
	
		Bioautomation
		
			11
			
			2008. 2008
		
	
	Suppl




* 
	
		Preferences, Utility Function and Control Design of Complex process
		
			YPavlov
		
		
	
	
		Proceedings in Manufacturing Systems
				in Manufacturing SystemsBucharest, Romania
		
			2010
			4
			
		
	




* 
	
		Equivalent Models and Exact Linearization by the Optimal Control of Monod Kinetics Models
		
			YPavlov
		
		
			KLjakova
		
	
	
		BIOautomation
		
			1
			
			2004. 2004
		
	




* 
	
		Preferences based Stochastic Value and Utility Function Evaluation
		
			YPavlov
		
		
	
	
		Proceedings of "InSITE' 2011
				"InSITE' 2011Novi Sad, Serbia
		
			2011. June 18-23
		
	
	InSITE Conferences




* 
	
		Principles of Microbe and Cell Cultivation
		
			JPirt
		
	
	
		Blackwell Scientific publications
		
			1975
		
	




* 
	
		A Genetic Algorithms Based Approach for Identification of Escherichia coli Fed-batch Fermentation
		
			ORoeva
		
		
			TPencheva
		
		
			BHitzmann
		
		
			St
		
		
			Tzonkov
		
	
	
		Bioautomation
		
			1
			
			2004
		
	




* 
	
		Multiple model approach to modelling of Escherichia coli fed-batch cultivation extracellular production of bacterial phytase
		
			ORoeva
		
	
	
		Electronic Journal of Biotechnology
		
			10
			4
			
			2007
		
		
			Pontificia Universidad Católica de Valparaíso -Chile