544 BIOCHEMICAL ENGINEERING FUNDAMENTALS Table 9.2 Relationships among substrate conversion 5 = (59 — s)/so and reactor design parameters for enzyme-catalyzed reactions in a PFTR* Enzymes in solution* or in immobilized pellets with negligible mass-transfer limita- tions ee ree eee ne Lae Lin eu Reaction rate expression for » PFTR design expression for Michaelis-Menten mee 595 ~ Ky In (1 — 6) Reversible Michaelis-Menten: nan — P/K) Kars K.+5+K,piK, (I-kE prado] K, 1 ~(ka ++ 5m) In (1 — 66) K+i where b Competitive product inhibition YaasS Kes ema —3y S¥KA1 + p/K) st i +n 9) K, ~(Ketae jin -a Substrate inhibition eS 5 Kin 8) 48 (5_% T+ K,/s +5/K, $09 — Ka In (1 — 8) + -3) " Adapted from R. A. Messing, Immobilized Enzymes for Industrial Reactors, p. 158, Academic Press, Inc, New York, 1975. ' The (1 ~ e)/e factor in these equations should be set equal to unity for en- zymes in solution. 9.2 REACTOR DYNAMICS . In this section we consider dynamic characteristics of bioreactors. Although dy- namics of CSTRs are the primary focus here, many of the concepts and principles introduced can be applied to other reactor configurations. We first develop the equations needed to describe unsteady-state reactor performance and thet examine use of those equations to characterize transient behavior of the bio~ reactor. We shall see that successful application of the,approaches presented her is often limited by lack of kinetic models which are accurate under transient! | operating conditions. | of ins jus Poi oft Che circ: an¢ arou CST char, nami oF fee neces for fi sttean applie reacto Ec all but equatic formati depend Will be Plify wrALs ugh dy- ‘inciples develop nd then the bio- ted here transient DESIGN AND ANALYSIS OF BIOLOGICAL REACTORS 545 9.2.1 Dynamic Models For dynamic studies of CSTRs the conservation equation (Eq, 7.4) must be modified to give the corresponding unsteady-state mass balance. d h (total amount in the reactor) = rate of addition to reactor dt rate of removal from reactor + rate of formation within reactor Thus, for a well-stirred vessel we have for component i qd. > P a Ye cd) = Flere — 6) + Vary, (9.26) Assuming that the feed stream and reactor contents have equal density, equality of inlet and outlet volumetric flow rates means the volume of the reactor contents Yq 18 constant, allowing rearrangement of Eq. (9.26) into the form de; dt Dleig = 6) +15, (9.27) This unsteady-state material balance is the starting point for characterization of reactor dynamics. Before introducing some general mathematical tools useful in dynamics analysis, we should comment on the new considerations required to Justify the use of a CSTR model to calculate dynamics of a bioreactor. At this Point we shall focus entirely on mixing phenomena, saving for later consideration of the required biological kinetics model for transient analysis. As mentioned in Chap. 8, one characteristic parameter of mixing in a vessel is the mixing or circulation time. This is an order of magnitude indication of the time Tequired for an element of fluid to return to a similar region of the reactor after circulating around the reactor according to the existing flow patterns. In order to apply a CSTR model, it is important that this circulation time be short relative to other characteristic times concerning the CSTR. In particular, a new characteristic time is introduced when we examine dy- namic behavior of a CSTR. Now, there is the possibility of time-varying feed rate or feed concentration. In order for the ideal mixing approximation to apply, it is necessary that the circulation time be much less than the characteristic time scale for fluctuations in the feed stream or, for that matter, in any other entering streams such as base additions for PH control. Exactly the same consideration applies to the use of the perfect mixing assumption for the case of fed-batch Teactors discussed earlier. Eq, (9.27) applies to each component considered in the bioreactor model. In all but the simplest case, then, the dynamic reactor model consists of a set of equations of the form of Eq. (9.27) which are usually coupled through the rate of formation terms ry, That is, in general the rate of formation of component i may depend on the concentrations of all of the other components in the reactor. Tt will be very convenient to introduce vector-matrix notation at this point to sim- plify writing large sets of equations,We shall adopt the notation convention that lowercase boldface letters like ¢ denote vectors and that uppercase boldface letters like A denote matrices. Then, the set of equations indicated by Eq. (9.27) ‘may be written in the form det) dt fle(©), p) (9.28) 346 BIOCHEMICAL ENGINEERING FUNDAMENTALS | i equal to the number of components considered in the model. Here p denotes a q-vector of model Parameters, such as feed concentrations, dilution rate, and kinetic parameters. The ith component of the vector-valued function f is equal to the right-hand side of Eq. (9.27). i Since the system described by Eq. (9.28) is in general nonlinear, we usually cannot go too far in our analysis without resorting to some approximations. Often we are interested in dynamic properties of the system near a particular steady state c,. In the notation of Eq, (9.28), the steady state concen- tration vector ¢, must satisfy where ¢ is a vector of concentrations with dimension m (with m elements or components; an m-vector) | f(c,,p) = 0 (9.29) ser em We can attempt to determine behavior near ¢, by expanding the right-hand side of (9.28) in a Taylor's series about ¢, and neglecting all terms of second order and higher in the deviations c(t) — ¢,,, since they are presumed small, Then we obtain the following linear approximation for our system ax £0 ay (030) dt where the vector x(t) denotes the vector of deviations from the steady state ¢,: x(t) = elt) —¢, (931) The element a,, in the ith column of the matrix A is defined by amen 032) We should emphasize that A corresponds to some particular steady state. Some systems have more than one steady state for a given p and this usually implies that a different A matrix corresponds to each steady state. ‘The dynamic properties of the linearized system are relatively easy to determine since all solutions of Eq, (9.31) usually take the form x() = 5 ape" 0.33) 4 ‘The quantities B, and 2, are the corresponding pairs of eigenvectors and eigenvalues of A. Thus 4 satisfies the characteristic equation. det (A ~ 1) =0 034) (lis the identity matrix), and the B, satisfy DB =O i= 1...,m (9.35) a~ ‘The a; are constants to tx chosen to fulfill the specified initial conditions; consequently they satisY the linear algebraic equations x(0) 5 9.36) Ex where x(0) is a specified vector of initial deviations. DESI of ch consi refere behas These for ing v locally signitic depend P. This other ir form ~ this rela Accordi altogeth time rat engineer ments to dimensio 92.2 Sts Next we 5 the selecte be of grea retum to from the n will “run a ambiguity, We st that the ini the Euclide Then “ey su Slate ¢, is gh like negative & lead to. instability, tf deviations Local st (9.32) The st Parts *M. jeld, Analysis of Ei e8CI08." Cher,rats i that 27) (9.28) ctor) model vent of far in perties oncen- 9.29) aylor's 5 since, (9.30) 931) (9.32) ve more vonds to since all (9.33) (934) (9.35) vey satisfy (9.36) DESIGN AND ANALYSIS OF BIOLOGICAL REACTORS S47 This linearized dynamic model of the reactor provides a systematic framework for identification of characteristic response times of the system. Within the framework of local dynamics which are considered here, we can see from Eq. (9.33) that the time scales for decay of disturbances from the reference steady state ¢, are characterized by the eigenvalues 4, of the matrix A. Thus. in terms of local behavior, the system exhibits a spectrum of characteristic times indicated approximately by ™ 0.37, These values can be used to examine the relative may for input variations, for example. While the approach leading to the time-scale estimates given in Eq, (9.37) is systematic and locally rigorous, the time scale estimates so obtained are difficult to assign a particular physical significance. The eigenvalues 4, are in general functions of all entries of the matrix A and, as such, depend upon the entire vector of steady-state operating conditions ¢, and the entire parameter vector P. This situation does not allow convenient comparison of time scales for mixing. for reaction, and for other interactions in the system. In the case of a CSTR. the eigenvalues may be shown to have the form —D + (a value characteristic of the reaction network)’ Although providing some guidance. this relationship does not unravel the complex relationships between other parameters and the 2, Accordingly, some judgment, experience, and some art is needed in developing a reasonable yet not altogether mathematically rigorous approach for identification of different characteristic lengths and time ratios. Of course one systematic method for achieving this which is well known in chemical engineering is transformation of all system equations to dimensionless form. followed by reatrange- iments (o identify dimensionless groups which characterize the system's behavior. Frequently. such dimensionless groups take the form of ratios of characteristic length or time scales ignitudes of reactor time scales with time scales 9.2.2 Stability Next we shall examine how the dynamic characteristics of the reactor depend upon the function f and the selected parameter values p. For our purposes, the local stability of a particular steady state e, will be of greatest concern. If a steady state is locally asymptotic lly stable, the system concentrations will return to that steady state after a small disturbance has moved those concentrations slightly away from the reference steady state of interest. For an unstable steady state, the concentrations considered will “run away” from the steady-state values following certain small disturbances. To be sure to avoid ambiguity, we shall restate these definitions in more formal mathematical language. We shall say that the steady state ¢, is locally asymptotically stable if lim, et) +, provided that the initial state cy is sufficiently close to ‘¢, [Our mathematical measure of closeness for vectors is the Euclidean norm, defined by, Iel= ( 5 2) Then “eo sufficiently close to ¢; means that [ey ~ ¢,\ isa sufficiently small real number.] The steady State ¢, is globally asymptotically stable if lim, ..¢(t) = ¢, for any choice of ¢9 (except ridiculous ones like negative concentrations). Ife, is an unstable steady state, some initial states ¢, arbitrarily close to « lead to trajectories e(t) which do not approach or stay arbitrarily close to c,. Thus, in the case of instability, the magnitude of the deviation x(t) tends to increase from its initial value for some initial deviations. Tocal stability is determined in most cases by the eigenvalues J, of the matrix A defined in Eq (9.32). The steady state ¢, is locally asymptotically stable if all eigenvalues of A have negative real parts Rei) <0 i= ham (9.38) "'M, Field, ©. A. Asbjorsen, and K. J. strom, “Reaction Invariants and theit Importance in the Analysis of Eigenvectors, State Observability, and Controllability of the Continuous Stirred Tank Reactor.” Chem Eng. Sci.. 29:1917, 1974.