Model based Analysis, Design, Optimization and Control of Complex (Bio)Chemical Conversion Processes Bioprocess Technology and Control - KULeuven.

Model based Analysis, Design, Optimization and Control of Complex (Bio)Chemical Conversion Processes

Bioprocess Technology and Control - KULeuven

Prelude …

Design, optimization and control

of (bio)chemical conversion processes

based on

Historical experience

• time consuming

• capital intensive

• operation/operator specific

• on-line measurements

• in silico design, optimization, and control studies

Mathematical model

Better and more

robust performance

practical implementationoptimization and control

manageabilityaccuracy

complex enough to cover main dynamics

Prelude: complexity trade-off

MODEL

accuracy

manageability

Primary model

Prelude: methodology

accuracy manageability

Model complexity reduction

Prelude: methodology

reaction transportaccumulation

Balance type equations

Complexity related to …

… # of states

… time & space dependency

… reaction kinetics

Complexity related to …

… # of states

Carbon and nitrogen removing activated sludge systems- biodegradation- sedimentation

Theme #1:

Fast & reliable simulations

Optimization & control

Objectives:

Complexity related to …

… # of states

Theme #1: Unit operationsASM1 model

Complexity: ASM1 model

(…)

input output

Complexity reduction

Aerated tank

Ss

Xbh

Xp

Sno

Snd

Xs

Xba

So

Snh

Xnd

time [day] time [day]

Theme #2: Filamentous bulking

Influent Effluent

Aeration tank Sedimentation tank

Activated sludge

ProcessControl

InfluentWastewater

Aeration TankEnvironment

Microbial Community

Selection

Effluent WaterQuality

Improvement

Long term objectives

Image AnalysisProcedure

Image AnalysisProcedure

Experimental set-up @ BioTeC

Influent

EffluentEFFLUENTEFFLUENT

TurbidityQuality

SLUDGESLUDGE

ConcentrationLoading

SettleabilityCharacteristics

Robustness test

Influence of microscope, camera and sludge type ?

ARX model

Theme #3: sWWTPS Rotating Biological Rotating Biological

ContactorContactor

Submerged Aerated Submerged Aerated FilterFilter

Milestones

Model complexity reduction for unit operationsModel complexity reduction for unit operations Linear Linear MultiMulti ( (or Fuzzyor Fuzzy)) MModel odel approach withapproach with

highhigh predictipredictiveve quality quality (input or state driven) (input or state driven) Significant Significant reduction inreduction in computation timecomputation time due to analytic due to analytic

solution of LTI state space modelsolution of LTI state space model (within 1 class)(within 1 class)

Simple Simple linear modellinear model for forrisk assessmentrisk assessment and and feedback feedback (MPC) (MPC) controlcontrol

Microbial dynamics: Microbial dynamics: exploiting image analysis information…exploiting image analysis information…

Application to (s)WWTPS…Application to (s)WWTPS…

Complexity related to …

… reaction kinetics

* Metabolism of bacterium Azospirillum brasilense* Quorum sensing of bacterium Salmonella typhimurium* Lag/growth/inactivation/survival …

Case studies:

Macroscopic/microscopic cell metabolism modeling

Objective:

High added value of specialty chemicalsHigh added value of specialty chemicals(food additives, vaccins, enzymes, …)(food additives, vaccins, enzymes, …)

Quantification of the influence of external signals onQuantification of the influence of external signals on cell metabolism (cell metabolism (A. brasilenseA. brasilense), and ), and quorum sensing (quorum sensing (S. typhimuriumS. typhimurium).).

Optimal experimental design of Optimal experimental design of bioreactor experimentsbioreactor experiments

Complexity

Primary modeling: identification of 14 parameters

EFT [h] EFT [h]

Co

[%]

Mal

ate

[g/L

]

OD

578

GU

S a

ctiv

ity

[M.U

.]D

[1/h]

Primary modeling: validation

EFT [h] EFT [h]

Co

[%]

Mal

ate

[g/L

]

OD

578

GU

S a

ctiv

ity

[M.U

.]D

[1/h]

Sensitivity function based model reduction

Sensitivity functionsSensitivity functions reflect the sensitivity of model predictions to (small) variations in model parameters with given inputs

time

0

5

-5

j

i

p

y

time

0

0.001

-0.001

j

i

p

y

Essential

Reduced model: identification experiment

EFT [h] EFT [h]

Co

[%]

Mal

ate

[g/L

]

OD

578

GU

S a

ctiv

ity

[M.U

.]D

[1/h]

Reduced model: validation experiment

EFT [h] EFT [h]

Co

[%]

Mal

ate

[g/L

]

OD

578

GU

S a

ctiv

ity

[M.U

.]D

[1/h]

max

Nmax

Escherichia coli K12 (MG1655), Brain Heart infusion, 36.3ºC

Microbial growth @ constant temperature

Stationary phase

Exponential phase

Lag phase

Estimation of microbial growth kinetics as function of temperature

Tmin Topt Tmaxsub-optimal temperature range

)()( minmax TTbT

SQUARE ROOT MODEL [Ratkowsky et al., 1982]

b

b minT

Tmin

Constrained input optimisation

1st experiment: based on po

Constrained input optimisation

2nd experiment: based on p1

Constrained input optimisation

Global identification of experiment 1 & 2

Constrained input optimisation

Milestones

Macroscopic modelingMacroscopic modeling: Sensitivity function : Sensitivity function analysis as a powerful tool to reduce the complexity analysis as a powerful tool to reduce the complexity of a physiology based, first principles modelof a physiology based, first principles model

Microscopic modelingMicroscopic modeling: : IBM (Individual based Modeling) linking IBM (Individual based Modeling) linking

• bio-informatics,bio-informatics, with with • macroscopic mass balance type modelsmacroscopic mass balance type models

Optimal experimental designOptimal experimental design of computer of computer controlled bioreactor experimentscontrolled bioreactor experiments

Complexity related to …

… reaction kinetics

Fed-batch growth process with non-monotonic kinetics

Case study:

Feedback stabilization: keep Cs constant

Objective:

Case study

u

time

Two valued function!

Case study

u

time

Two valued function!

Case study

u

time

Two valued function!

Controller (on-line Cx measurements)

Feedforward (OC) Stabilizing feedback

observer

I-action

P-action

= +1 = -1 or

Stabilizing feedback controller for fed-batchStabilizing feedback controller for fed-batchnon-monotonic growth processesnon-monotonic growth processes

Only based on on-line biomass Only based on on-line biomass concentration measurementsconcentration measurements

Adaptive: no detailed kinetics information Adaptive: no detailed kinetics information needed (needed ( observer) observer)

Conclusions

Complexity related to …

…time & space dependency

Tubular chemical reactorsCase study:

Optimal jacket fluid temperature control of - classical reactors, and - novel type reactors

Objective:

Tubular chemical reactor

C = reactant concentration [mole/L]

T = reactor concentration [oK]

Tw = jacket fluid temperature [oK]

Model for tubular reactor: PDE/DPS

Combined terminal/integral objective

Conversion

Hot spots

Temperature run-away

Determine optimal jacket fluid temperature profile

( )2

Comparison with suboptimal profiles

maximum-singular-minimummaximum-singular-minimum profile profile optimal, but optimal, but

singular part difficult to implementsingular part difficult to implement maximum-minimummaximum-minimum profile profile

not optimal, but not optimal, but practically realizablepractically realizable

how much how much optimality optimality is lost?is lost?

0.3

Comparison with suboptimal profiles (I): Conversion

0.7

Comparison with suboptimal profiles (II): Hot Spots

Milestones: optimal control theory for …

… … optimal optimal analyticalanalytical jacket fluid temperature jacket fluid temperature profiles for profiles for classical classical chemical reactorschemical reactors steady statesteady state transienttransient

… … optimization of optimization of novel typenovel type reactors reactors cyclically operated reverse flow reactorscyclically operated reverse flow reactors circulation loop reactorscirculation loop reactors

… … optimal reactor optimal reactor designdesign

Postludium …

Dealing with complexityDealing with complexity during modeling for during modeling for optimization and control of optimization and control of (bio)chemical processes: (bio)chemical processes: a multimodal problem at the interface of a multimodal problem at the interface of various disciplinesvarious disciplines

We will pass several cases in review over the We will pass several cases in review over the years to come…years to come…

… emerging generic results

Development of widely applicable and Development of widely applicable and transferable quantitative tools for complex transferable quantitative tools for complex (bio)chemical processes(bio)chemical processes

WP3

WP1 WP2

WP4