CHEM-E3205 BIOPROCESS OPTIMIZATION AND

SIMULATION

TERO EERIKÄINEN
ROOM D416d
tero.eerikainen@aalto.fi
 COURSE LECTURES AND EXERCISES
Week Day Date Time Place Lectures/Execises
37 Mo 12.9.2016 10:15-11:45 Ke3 Lecture: Bioprocess modeling
Tue 13.9.2016 10:15-11:45 Luokka1 Exercise: Matlab bioprocess kinetics simulation
38 Tue 20.9.2016 10:15-11:45 Luokka1 Exercise: Matlab parameter estimation
Wed 21.9.2016 14:15-15:45 Ke5 Lecture: Bioprocess measurement and control
39 Tue 27.9. 10:15-11:45 Luokka1 Exercise: Matlab PID-control
Wed 28.9. 14:15-15:45 Ke5 Lecture: Design of experiments
40 Tue 4.10. 10:15-11:45 Luokka1 Exercise: Modde design of experiments
Wed 5.10. 14:15-15:45 Ke5 Lecture: Multivariate modelling
41 Tue 11.10. 10:15-11:45 Luokka1 Exercise: Simca exercises
Wed 12.10. 14:15-15:45 Ke5 Lecture: Neural network and other modelling
41 Tue 18.10. 10:15-11:45 Luokka1 Exercise: Neural network
Wed 19.10. 14:15-15:45 Ke5 Lecture: Quality Control
Exam
42
week
 EXERCISES (MANDATORY)
• PC CLASSROOM EXERCISES:
• YOU CAN PERFORM EXERCISES BY BEING PRESENT IN THE PC
CLASSROOM. IF CARRIED OUT INDEPENDENTLY BY YOURSELF YOU
SHOULD RETURN THE ANSWERS IN A REPORT. EXERCISES SHOULD BE
RETURNED AND GET APPROVED BEFORE THE COURSE IS ACCEPTED AS
COMPLETED. SEPARATE EXAM IS ALSO HELD.
• LAB WORK (SMALL GROUPS):
• FIND OUT HOW THE BIOREACTOR WORKS.
• MAKE A PLAN TO DEFINE REACTOR kLa-MEASUREMENT.
• MAKE A PLAN TO DEFINE KINETIC PARAMETERS FOR AN AERATED
YEAST FERMENTATION.
• REALIZE PLANS USING BIOSTAT C (5L) BIOREACTOR.
• MAKE A SIMULATION MODEL FROM MEASUREMENT RESULTS
• MAKE A REPORT OF YOUR WORK
 MATERIALS

• FUNDAMENTAL
BIOENGINEERING
• ONLINE VERSION AVAILABLE
FREELY

• CHAPTERS 7,8,14,15
• EXTRA ARTICLES
• LECTURE SLIDES

• EXERCISES

http://libproxy.aalto.fi/login?url=http://onlinelibrary.wiley.com/book/10.1002/9783527697441
CHEM-E3205 BIOPROCESS OPTIMIZATION
AND SIMULATION:
MODELING

• MODELING CATEGORIES
• BALANCE MODELS
• KINETIC MODELS
• BIOREACTOR MODELS
• (MASS TRANSFER MODELS)*
• (OTHER METHODS)*

*NOT IN THIS LECTURE

 MODELING

• DESCRIPTION OF THE PROCESS BY MEANS OF MATHEMATICS

• REASONS FOR MODELING:
• PROCESS DESIGN AND OPTIMIZATION
• SIMULATION
• THE DESIGN OF EXPERIMENTS AND THE RATIONALIZATION
• DEFINING PROCESS STATES AND DYNAMICS
• DESIGN AND TUNING OF PROCESS CONTROLLERS
• AS PART OF THE CONTROL ALGORITHM (ESTIMATION)
• QUANTITATIVE MODELING OF BIOPROCESSES ARE BASED ON MATERIAL AND /
OR ENERGY BALANCES AND THE KINETICS
• (SUBSTRATE) RAW MATERIAL PRE-PROCESSING IS AN IMPORTANT VARIABLE
• BIOLOGICAL SYSTEMS ARE COMPLEX AND NON-LINEAR
• DYNAMIC CHANGES ARE STRONGLY DEPEND ON THE INITIAL VALUES
 MODELING
• MECHANISTIC MODELS
• BASED ON NATURAL PHENOMENA, LAWS OF PHYSICS (BEING
"UNIVERSAL“)
KINEMATICS, DYNAMICS, STATICS

• EMPIRICAL MODELS
• "BLACK-BOX" MODELS, DESCRIBING THE RELATIONS BETWEEN INPUT
AND OUTPUT VARIABLES FOR EXAMPLE USING REGRESSION MODELS
• QUALIFIED IN THE DOMAIN WHERE IDENTIFIED
 MODELING
• STATIC MODELS
• A DESCRIPTION OF THE TIME INDEPENDENT PHENOMENA
• FOR EXAMPLE PRODUCT QUALITY VS. RAW MATERIALS OR REACTIONS,
WHOSE RATE IS VERY HIGH

• DYNAMIC MODELS
• TIME DEPENDENT MODELS
• PROCESSES IN WHICH REACTION (SLOW) RATE AFFECT THE FINAL RESULT
• TIME SERIES MODELS
 MODELING

• CONTINUOUS PROCESSES
• PROCESS IS TRIED TO ADJUST TO A CERTAIN OPTIMUM -> STATIC MODEL

• BATCH PROCESS
• A BATCH PROCESS IS USUALLY TREATED AS A DYNAMIC PHENOMENON,
UNLESS ONLY THE FINAL PROCESS STATE MATTERS
 MODELING
• NONSEGREGATED MODEL
• THE CELLS JUST ONE AND THE
SAME BIOMASS Nonstructured
Ei-rakenteellinen Structured
Rakenteellinen
• SEGREGATED MODEL
• BIOMASS IS DIVIDED INTO
SUBPOPULATIONS, WHICH ARE Jakautunut
Segregated
TREATED AS SEPARATE nty
y
VARIABLES ää
lis
s
• NONSTRUCTURED MODEL suu
i
tka
• A DESCRIPTION OF THE nim
u
BIOMASS AT THE M
o
MACROSCOPIC LEVEL Jakautumaton
Nonsegregated
• STRUCTURED MODEL
• THE CELLULAR COMPONENTS
ARE HANDLED AS SEPARATE
VARIABLES
BALANCE MODELS

• MASS BALANCE
• ELEMENTAL BALANCE
• ENERGY BALANCE
• REDOX BALANCE
MASS BALANCE

MASS BLANCE FOR

SUBSTANCE A :

RA = AACCUM.RATE =

AIN - AOUT + AREACTION

 ELEMENTAL BALANCE
α*CaHbOc + β*O2 + χ*NH3 --->
CdHeOfNg + δ*CO2 + ε*H2O

• CARBON: a*α = d + δ
• HYDROGEN: b*α + 3*χ = e + 2*ε
• OXYGEN: c*α + 2*β = f + 2*δ + ε
• NITROGEN: χ=g
 ENERGY BALANCE

Energy Energy Energy

Energy = - +
convection convection from
accumul.
in out reaction

Energy Energy Energy

+ conduction - conduction + from
in out mixing

± Energy
radiation
 REDOX BALANCE

• ELECTRON BALANCE
• CO2, H2O AND NITROGEN COMPOUNDS
FORMED IN COMBUSTION REACTIONS
• VALENCES OF VARIOUS ELEMENTS:
• CARBON: 4
• HYDROGEN: 1
• OXYGEN: -2
• PHOSPHOROUS: 5
• SULPHUR: 6
• NITROGEN: -3 (NH3), 0 (N2), 5 (NO3-)
KINETIC MODELS

• dX/dt, dP/dt, dS/dt

• ENZYME KINETICS
• MICROBIAL GROWTH
KINETICS

• BIOCHEMICAL REACTIONS
IN LIVING ORGANISMS ARE
DEPENDENT INTERACTING
ENZYME REACTIONS
ENERGY REQUIREMENTS OF A
CHEMICAL REACTION

Figure 5.2

From Pearson Education, Inc.

 BASICS OF ENZYME KINETICS

• QUANTITATIVE EVALUATION OF ALL FACTORS THAT CONDITION ENZYME ACTIVITY

• MOST IMPORTANT FACTORS:
• CONCENTRATIONS OF ACTIVE ENZYME, SUBSTRATES AND INHIBITORS
• pH AND TEMPERATURE

• KINETICS NEEDED FOR

• UNDERSTAND THE MOLECULAR MECHANISMS OF ENZYME ACTION
• DESIGN OF ENZYME REACTORS AND FOR
• PERFORMANCE EVALUATION

• DETERMINE INITIAL RATES OF REACTION

• DETERMINE THE QUANTITATIVE EFFECT OF THE IMPORTANT FACTORS
 ENZYME KINETICS
Activity proportional to the concentration of active enzyme

k1 k Henri kinetics
S+E ES P+E
k2 K1 dissoc. const ES
K2 dissoc. const EP

dS dP P=0 or insignificant
v=− = = k ⋅[ ES ] = k cat ⋅[ ES ]
dt dt

kE0 S Vmax S Rapid equilibrium

v= = Michaelis - Menten K= equilibrium constant
+S K +S
k2
The binding step in equilibrium and
k1 much faster than the conversion step

kE0 S V S Steady state hypothesis

v= = max Briggs - Haldane KD = dissociation constant
k2 + k
+S KD +S
k1
 BRIGGS-HALDANE APPARENT STEADY STATE
After a very short transient state the enzyme–substrate complex reaches steady state, so that its
concentration remains constant throughout the reaction

v = k [ES ] [ES ]= ([E0 ]−[ES ])[S ]

KD
ES formation = k1 [E ][S ]
[ES ]K M + [ES ][S ]= [E0 ][S ]
ES dissociation = (k 2 + k )[ES ]
[ES ]= [E0 ][S ]
In steady state: K D + [S ]
ES formation = ES dissociation ⇒
d [ ES ]
=0 v=k 0
[E ][S ] =V [S ]
K D + [S ] K D + [S ]
max
dt
k1 [E ][S ]= (k 2 + k )[ES ] when S >> K D →
[ES ]= [E ][S ] =
[E ][S ] v =Vmax = k [E0 ] (0 order reaction kinetics)
(k 2 + k ) k1 KD when S << K D →
[E ]= [E0 ]−[ES ] [S ]
v =Vmax (1st order reaction kinetics)
KD
KD is the dissociation constant of the ES complex into E and S
 KM and Vmax
Whatever the hypothesis, the rate equation is expressed in terms of two parameters:
KM is “Michaelis constant,” (not being dependent on either enzyme or substrate concentration)
Vmax is a lumped parameter containing the enzyme concentration (e or [E0]) and the catalytic rate constant (k or kcat).
(dimension of k or kcat will be determined by the enzyme concentration dimension)

When defining kinetic parameter, the value of the determined parameter KM should be in the midpoint of that
range. KM corresponds the substrate concentration in which reaction rate in half of the maximum.

18,0 Irreversible
16,0
reaction
14,0

12,0
Km =10
v (g/l s)

10,0

8,0
v = (Vmax*s)/(Km+s)
6,0

4,0 Vmax = 20
2,0 Km

0,0
0,0 20,0 40,0 60,0 With small substrate concentration 1st order kinetics
s (g/l) With large substrate concentration zero order kinetics
 LINEARIZATION METHODS

[S ] = [S ] +
Km
Langmuir
v v max v max

v
v = v max − Km Eadie − Hofstee
[S ]

1 1 Km 1
= + Lineweaver − Burk
v v max v max [S ]
LACTOSE HYDROLYSIS WITH β-GALACTOSIDASE: EMPIRICAL Km AND vmax VALUES:

experime Logaritmic
nt y = 0.4032Ln(x) + 0.4648
Lactose reaction
rate Best result with non- 1,8

mmol l-1 mmol l-1 min- linear parameter 1,6

1 estimation (not 1,4
S v shown here) 1,2
0,50 0,106 1

v
1,00 0,376 0,8
2,00 0,764 0,6
4,00 1,152 0,4
6,00 1,386 Km vmax
0,2
8,00 1,388 Logaritmic 5,6 2,3
0
14,00 1,500 Lineweaver-Burk -38 -8,9 0 10 20 30
Langmuir 3,8 1,8
20,00 1,438 S

Lineweaver-Burk 1/v Langmuir-plot

s/v=Km/vmax + s/vmax
1/v=Km/vmax *1/s + 1/vmax
y = 4.3184x - 0.1129 y = 0.5429x + 2.0834
10.0 16
9.0 14
8.0 12
7.0
10
6.0
S/v
1/V

5.0 8
4.0 6
3.0
4
2.0
2
1.0
0.0 0
0 5 10 15 20
0.000 1.000 2.000 3.000
1/S S
 MICROBIAL GROWTH KINETICS
• BIOCHEMICAL REACTIONS IN LIVING ORGANISMS ARE OFTEN
ENZYMATIC REACTIONS DEPENDENT ON EACH OTHER.
• CELL GROWTH IS AN AUTOCATALYTIC REACTION: THE GROWTH
RATE IS DIRECTLY PROPORTIONAL TO THE AMOUNT OF PREFORMED
GROWTH OF CELLS.
• THE SPECIFIC GROWTH RATE MAY BE DETERMINED WHEN THE CELL
CONCENTRATION CHANGE RATE dX/dt IS DIVIDED WITH CELL
CONCENTRATION

𝑑𝑑𝑑𝑑 1
𝜇𝜇 = �
𝑑𝑑𝑑𝑑 𝑋𝑋
 MICROBIAL GROWTH
14
• IN BATCH CULTIVATION:
stationary phase
12 • LAG-PHASE
10 • ACCELERATING PHASE
X (kg/m3)

8
Exponential grow th
• EXPONENTIAL GROWTH
6
• DECELERATING GROWTH
4

2
• STATIONARY PHASE
0
Lag-phase
• CELL DEATH
0 1 2 3 4
• 1ST ORDER KINETICS
time (d)
 MONOD EQUATION
• SUITABLE FROM EXPONENTIAL
TO STATIONARY PHASE
• SPECIFIC GROWTH RATE µ IS µ max S
µ=
GIVEN AS A FUNCTION OF Ks + S
SUBSTRATE CONCENTRATION.
PARAMETERS ARE MAXIMUM
SPECIFIC GROWTH RATE µMAX µ max S
AND MONOD CONSTANT KS: S << K s → µ ≈
Ks
S >> K s → µ ≈ µ max
µ max
S = Ks → µ ≈
2
 MONOD CONSTANT

• “IF AN ANALYSIS OF FERMENTATION

DATA BY THE MONOD MODEL
GIVES A KM VALUE SUBSTANTIALLY
DIFFERENT FROM THE LITERATURE
VALUES, THERE IS REASON TO
BELIEVE THAT THE WRONG
STRUCTURE OF THE KINETIC
EXPRESSION HAS BEEN CHOSEN. “
• “THUS, IF KM IS FOUND TO BE 1
g/L, THIS IS A SIGN THAT THE
WRONG LIMITING SUBSTRATE HAS
BEEN CHOSEN OR THE REACTION
SUFFERS FROM PRODUCT
INHIBITION.”
 EXPANDED MONOD KINETICS

• “THE SIMPLE MONOD MODEL CAN ALSO BE EXPANDED TO INCLUDE BOTH

SUBSTRATE INHIBITION (7.21) AND PRODUCT INHIBITION (7.22) OR (7.23)”
• “ALL THREE EQUATIONS ARE EMPIRICAL IN NATURE, AND THEIR FORM JUST
MIMICS ORAL MODELS FOR, RESPECTIVELY, SUBSTRATE AND PRODUCT
INHIBITION OF CELL REACTIONS”
• “IN WINE FERMENTATION, EQ. (7.22) GIVES A GOOD REPRESENTATION OF
FERMENTATION DATA WHEN p pmax ≈120–130 g ethanol/L FOR MOST
WINE YEASTS.”
 EFFECT OF SUBSTRATE INHIBITION TO SPECIFIC GROWTH RATE WITH
VARIOUS KS AND KI VALUES, WHEN µMAX=1.0 H-1:
A) KS=1; KI=10 B) KS=0.1; KI=10
C) KS=1; KI=20 D) KS=0.1; KI=20
UNITS FOR S, KS, KI : kg m-3

Non-competitive substrate
inhibition :

µ max
µ=

1 +
K S 
1 +
[S]

 [S ]  K I 
IN A MORE COMPLEX KINETIC MODELS ONE
MAY TAKE INTO CONSIDERATION:

• CELL DEATH
• SUBSTRATE AND PRODUCT INHIBITION
• MAXIMUM CELL DENSITY (POPULATION)
• PREFERENCE OF CARBON SOURCE
• TEISSIER, CONTOIS, MOSER, LOGISTIC
• EFFECT OF MANY LIMITING CARBON SOURCES:

µ = µ max
[S1 ]
⋅
[S2 ]
⋅
K I1
⋅
KI2
K S 1 + [S1 ] K S 2 + [S 2 ] K I 1 + [I1 ] K I 2 + [I 2 ]
 YIELD COEFFICIENTS

• YX/S , YX/O2 YX/ATP

• YIELD COEFFICIENTS DEFINE THE RELATIONSHIP OF FORMATION AND
CONSUMPTION OF DIFFERENT PRODUCTS AND SUBSTRATES
(INCLUDING ENERGY)

dC X
dt ∆C X
YX / S =− ≈−
dC S ∆C S
dt
EXAMPLE: KINETIC MODEL FOR BIER
FERMENTATION
Yeast suspension consists of three components in a segregated model

From Lag-phase to active yeast, reaction (1)

Dead yeast sedementation rate µSD, half of the inoculation amount is dead, reaction (2)

Active yeast grows, a part will die, a part is in a lag-phase (1,3,4)

 Specific sedimentation rate

Specific growth rate Effect of temperature

Specific substrate consumption rate

Ethanol inhibition effect and specific product formation rate

 EXPERIMENTAL
ARRANGEMENTS
 PILOT-SCALE CULTIVATION ALONG THE
TEMPERATURE PROFILE AND THE ESTIMATES FROM
THE KINETIC MODELS
 BIOREACTOR MODELING

• BATCH CULTIVATION

• CONTINUOUS CULTIVATION

• FED-BATCH CULTIVATION

• PLUG-FLOW REACTOR
 BATCH CULTIVATION

Total volume:
dV
Exhaust gas
=0
dt
Substrate:
dS − µX
V =( − mX )V
dt YX / S
Aeration
Batch cultivation Biomass:
dX
V = µXV
dt
CONTINUOUS CULTIVATION

Total volume:
dV
= Fin − Fout = 0
Substrate Exhaust gas
dt
Substrate:
dS − µX
V = F ( Sin − S ) + ( − mX )V
dt YX / S
Biomass:
dX
Aeration Product V = − FX + µXV
Chemostat dt
Dilution rate:
F
D= =µ
V
Chemostat variables as a function of dilution rate
 FED-BATCH CULTIVATION

Total volume:
dV
Substrate Exhaust gas
= Fin
dt
Substrate:
dS − µX
V = F ( Sin − S ) + ( − mX )V
dt YX / S
Biomass:
Aeration
Fed-batch
cultivation dX
V = − Fin X + µXV
dt
 STATE-SPACE
x = Ax + Bu
REPRESENTATION y = Cx
• FIRST ORDER DIFFERENTIAL EQUATIONS
missä
CAN BE REPRESENTED BY STATE-SPACE  X(t ) 
EQUATIONS x= 
 S ( t ) 
• THE SYSTEM STATE x, CONTROLS u u = [S in (t )]
AND OUTPUT y VALUES ARE
REPRESENTED IN A MATRIX FORM  µ −D 0 
A =  − µ − m − D
• MATRIX REPRESENTATION ENABLES THE  
STABILITY ESTIMATION AND HELPS TO  YX / S 
CALCULATE TRANSFER FUNCTIONS 0
B= 
• HERE IS DESCRIBED CONTINUOUS  D
CULTIVATION THE STATE EQUATION :
C = [1 0]
 PLUG-FLOW REACTOR

• CONTINUOUS PROCESS
• CAN BE MODELLED WITH
STATIC OR DYNAMIC
MODELS
Tuote

• STATIC MODEL RESEMBLES

BATCH CULTIVATION THE
CONCENTRATION IS
FUNCTION OF THE
POSITION
Substraatti • DYNAMIC MODEL IS
CREATED DIVIDING THE
REACTOR LENGTH/HEIGHT
TO FINITE ELEMENTS
THE PLUG FLOW REACTOR CONCENTRATION CHANGES AS A
FUNCTION OF THE POSITION Z IN AN EQUILIBRIUM:
A) NO REACTION B) WITH THE REACTION, AND C) DYNAMIC FINITE ELEMENT MODEL

a)

b)

c)
OVERVIEW OF DIFFERENT TYPES OF DATA FOR
ODE-BASED KINETIC MODELS OF METABOLISM