model-based design, scale-up, and operational optimization of · pdf filemodel-based design,...
Post on 07-Mar-2018
217 Views
Preview:
TRANSCRIPT
© 2013 Process Systems Enterprise Limited
Alejandro CanoHead of Consulting, Process Systems Enterprise
Model-based design, scale-up, andoperational optimization offixed bed catalytic reactors and processes
11th Topsøe Catalysis ForumMunkerupgaard, DenmarkAugust 29-30, 2013
© 2013 Process Systems Enterprise Limited
Private, independent companyincorporated in UK
1997
Company ‘spun out’Acquires technology
HISTORY: FROM RESEARCH TO INDUSTRY
Advanced Process Modelling< Software and services (60:40)< Major process industry focus – all sectors< Strong R&D reinvestment< ~100 people worldwide
1989 – 1997 2013
Royal Academy MacRobert Award for Engineering InnovationUK’s highest engineering award
100s of person-years ofR&D with industry
Simulation & modeling,optimization, numericalsolutions techniques,supply chain
London HQ Korea JapanGermanyUSA
© 2013 Process Systems Enterprise Limited
Advancedthermodynamics
Advanced modellibraries
PSE products & services
gPROMS modeling& solution platform
1b
1c
Safety
Solidsprocesses
Fuel cell &hydrogen
Engineering solutions
ModelBuilder
Crystalli-sation
1a
1d
General processmodeling tools 2 Consultancy & services
Processtechnology& services
4
GeneralCustom Model DevelopmentProcess Design/Optimization
Process-specificSafety studies for Oil & GasIndustry
© 2013 Process Systems Enterprise Limited
Modeling and simulation of heterogeneouscatalytic processes
Topic of this year’s forum:
© 2013 Process Systems Enterprise Limited
Fixed-Bed Catalytic Reactors (FBCRs)Key engineering problems
1. Scale-up- Laboratoryà Pilotà Commercial Plant- Maintenance of performance over scales- Cost efficiency in investment and operation
2. Thermal stability –management of hot spots- adjustment of catalytic bed properties – length,
activity, shape of particles, etc- design of cooling system
3. Catalyst lifetime- Management of catalyst de-activation over
operational cycle
© 2013 Process Systems Enterprise Limited
All models are wrong
…but some are useful
George E.P. Box, Empirical Model-Building and Response Surfaces, 1987
Modeling and simulation: Bad news and good news
© 2013 Process Systems Enterprise Limited
Characteristics of a useful model for process design
1. Model has the right level of detail needed to simulatekey system responses
2. Model parameters are fitted to experimental data
3. Model predictions are accurate enough to supportoptimization-based process design
4. Model is suitable for scale-up: predicts responses seenin large scale equipment
© 2013 Process Systems Enterprise Limited
The Advanced Process Modeling Approach: 4 steps
1 2 3 4Optimization-
BasedDesign
Model-targetedExperimentation
+ParameterEstimation
AdvancedProcess Modelwith all physics
relevant toproblem of
interest
Final Adjustmentsto Equipment
Design
© 2013 Process Systems Enterprise Limited
The Advanced Process Modeling Approach: Step 1
1 2 3 4Advanced
Process Modelwith all physics
relevant toproblem of
interest
© 2013 Process Systems Enterprise Limited
Coolant
Coolant
Key phenomena: Mass transport and reaction
Multicomponentsolid-fluid masstransfer
Axial/radialconvection intube-side fluid
Multicomponentintra-pore diffusion
Catalytic reactionon pore surface
Radial dispersion
Intra-poreconvection
Catalyst pellets/ inert
Tube wall
Thermowell
Catalystpellet
© 2013 Process Systems Enterprise Limited
Core model: transport and reaction in catalytic pellets
−1
=−
+
Implemented in spherical or cylindricalcoordinates:
• Different particle shapes• Solid or hollow (for cylinders)• With or without inert cores
=
© 2013 Process Systems Enterprise Limited
Coolant
Coolant
Key phenomena: Heat transfer
Heat transferbetween solidand fluid
Energy carried bysolid-fluid diffusivemass transfer
Intra-pellet heatconduction
Radial heatconduction inpacked bed
Heat transferbetweenpacked bedand wall
Axial convectiveheat transfer influid
Convectiveheat transfer incoolant
Radialconduction intube walls
Heat transferbetween tube andcoolant
© 2013 Process Systems Enterprise Limited
FBCR – Heat removal in a tubular fixed bed
Axial & radial variation ofbed temperature
Temperatures from center of bed to wall
Heat transfer coefficientat the bed-wall boundary
Wall-coolant heat transfer coeff,
Effective bed conductivity
Bed
Wall
Coolant
© 2013 Process Systems Enterprise Limited
Example: PSE’s Advanced Model Library forFixed Bed Catalytic Reactors
1
© 2013 Process Systems Enterprise Limited
Library contents: Axial-flow catalytic bed reactors
< Catalyst Pellet Sections< Inert Sections
© 2013 Process Systems Enterprise Limited
Library contents: Shell-side models
< Fixed coolant< Cooling jacket< Multitubular cooling compartment< Boiling water cooling
© 2013 Process Systems Enterprise Limited
Additional building blocks
< Radial flow reactor sections
< Heat integrated annular sections
© 2013 Process Systems Enterprise Limited
Assembly of components into reactor model
Inert packing Inert packing
Catalyst/inert ratio #1Catalyst/inert ratio #2
BafflesBaffles
© 2013 Process Systems Enterprise Limited
The Advanced Process Modeling Approach: Step 2
1 2 3 4Model-targetedExperimentation
+ParameterEstimation
Advanced ProcessModel with all
physics relevantto problem of
interest
© 2013 Process Systems Enterprise Limited
Model ValidationEnsuring predictive accuracy throughmodel-targeted experimentation
2
© 2013 Process Systems Enterprise Limited
Experimentation in the APM approach
The purpose of the experiments is not to predict thebehaviour of the commercial-sized equipment
(that is the job of the validated model)
The objective of the experiments is to find the values ofunknown model parameters,
minimizing the uncertainty in these values
Model-targeted experimentation
© 2013 Process Systems Enterprise Limited
AML:FBCRModel parameters not derived from first principles
1. Kinetic parameters (due to variations in catalyst properties)- reaction pre-exponential constants and activation energies- reaction orders- adsorption constants and heats of adsorption
- for strongly adsorbing species
- deactivation model parameters2. Bed properties (due to deviations from ideal of perfectly spherical
particles of identical size)- coefficients in Ergun equation for pressure drop- coefficients in heat transfer parameter correlations
- bed effective radial conductivity (static and dynamic)- bed-wall heat transfer coefficient (static and dynamic)
3. Particle geometric properties- tortuosity
© 2013 Process Systems Enterprise Limited
Reactor model validationTwo-step experimental procedure (ideal)
Sampling tubes for internalgas composition
Fine inert Near-isothermaloperation
O(cm)
n-pellet experimentfor kinetics identification
Hollow cylinderpellets
Single-tube experimentfor heat transfer characterization
O(m)
Fixed bed(commercial pellets)
Multiplethermocouplesdirectly in bed
Jacketedtube
© 2013 Process Systems Enterprise Limited
Step 1 – Reaction kinetics
Processing ofexperimental datato evaluate themodel parameters
Parameter estimation tool
3
Lab reactor model2
Evaluatedparameters forthe reactionkineticsmodel
5
1
Laboratory data
Prediction of composition profiles in lab scale experimental reactor4
© 2013 Process Systems Enterprise Limited
Step 2 – Heat transfer parameters
Prediction of temperature profiles in pilot scale experimental reactor
Pilot reactor model
Processing theexperimental datato evaluate themodel parameters
Parameter estimation tool
Evaluatedparameters ofthe heattransfermodel
1
Pilot plant data
© 2013 Process Systems Enterprise Limited
Lessons from experience: Kinetic parameters (1/2)
< Search literature for kinetic expressions< Langmuir-Hinshelwood is a good starting point< Check for chemical equilibrium limitations< Break correlation between pre-exponential constant and
activation energy:
< Break correlation between rate constants and adsorptionconstants:
= − = , −1−
1
= 1 + + + =1 + + : + :
© 2013 Process Systems Enterprise Limited
Lessons from experience: Kinetic parameters (2/2)
< Vary temperature, pressure, feed composition< Include experiments at low conversion< Perform experiments with co-feed of products that participate
in secondary reactions
A B C
< Perform experiments with co-feed of strongly adsorbing by-products
< Measure temperature at several positions along the catalystbed
< Characterize carefully the experimental error in outletcomposition measurements
2O2O2
3O2
© 2013 Process Systems Enterprise Limited
Lessons from experience: Bed properties
< Vary gas velocity(to discriminate betweenstatic and dynamiccontributions)
< Use tubes of different size(to discriminate between bed-wall heat transfer and radialconductivity contributions)
< Cooling jackets preferable toclam shells or electric tape
< Coolant flow rate should behigh enough to yield turbulentflow
H.t. coeff. at thebed-wall boundary
Wall-coolant h.t. coeff.
Effective bed conductivity
Bed
Wall
Coolant
© 2013 Process Systems Enterprise Limited
Lessons from experience: particle properties
< Use reliable laboratory tocharacterize particle properties- Particle size distribution- Pore size distribution- Porosity
< Conduct experiments withparticles of different size to adjusttortuosity factor
© 2013 Process Systems Enterprise Limited
Example: Management of catalyst deactivationin Methanol production process
31
© 2013 Process Systems Enterprise Limited
Phase I: determining accurate kinetics
Applications
< Step 1: Run experiments- gradient-free recycle reactor
system (Berty reactor)- total of 70 steady-state
experiments
< Step 2: Validate- use model of lab-scale
reactor system to obtainaccurate kinetic parameters
- 4 different kinetic modelsanalysed for suitability
Berty reactor flowsheet
Typicalparity plotforvalidateddata
© 2013 Process Systems Enterprise Limited
Phase II: model the commercial reactor
Megamethanol® process:2 coupled reactors
Reactor 1: Catalyst inside tubes, cooled by boiling water
Reactor 2: Catalyst in shell side, cooled by process deed
© 2013 Process Systems Enterprise Limited
Results: catalyst deactivation predictions(dynamic model)
< Accurate prediction ofcatalyst de-activationwith accumulatedproduction for eachreactor
Cata
lyst
activ
ity
Accumulated production (over weeks)
Actual
Predicted
Catalyst activity at axial and radial bed positions (100 days )
© 2013 Process Systems Enterprise Limited
Conclusions
< Using validated model, catalyst manufacturer could assist theclient to maximise production for a given time period until theend of run (within the given framework of plant constraints)
< Catalyst manufacturer could use the models and methodologyto optimise their catalyst for this client’s requirements
Conclusion 1
Conclusion 2
© 2013 Process Systems Enterprise Limited
The Advanced Process Modeling Approach: Step 3
1 2 3 4Optimization-
baseddesign
Model-targetedExperimentation
+ParameterEstimation
Advanced ProcessModel with all
physics relevantto problem of
interest
© 2013 Process Systems Enterprise Limited
Multitubular reactor – key design variables
< Tube side- number, diameter, length
of tubes- number and length of layers- catalyst/inert ratio- pellet shape & size- . . . . . .
< Shell side- number & positioning of
cooling circuits- number & positioning of
baffles- coolant flowrate(s)
catalystpacking
catalystspec
Cata
lyst
ICa
taly
stII
Cata
lyst
IIITube andcatalystconfiguration
Iner
t
coolant flowrate
inner tube limit
baffle window size
baffle spanouter tube limit
coolant temp
reactor diameter
tube sizearrangement, pitch
© 2013 Process Systems Enterprise Limited
Multitubular reactor – objectives and constraints
< Minimize capital/operating costs< Maximize selectivity< Maximize catalyst life< Achieve production target< Keep pressure drop within limits< Prevent runaway – keep temperature within limits< Keep reactor dimensions within limits
- road transport considerations
< Keep shell-side velocities within limits- avoid erosion, vibration, fouling
< Prevent formation of undesirable phase
© 2013 Process Systems Enterprise Limited
Optimization-based design procedure (1/2)
1. Add equations that relate model variables to performanceindicators:
2. Select decision variables- specify initial guesses (e.g. current design)- specify allowable range of variation (continuous/discrete)
= × ×
= × × 3600 s/hr × Annual Operating Hours
= × − − ×
© 2013 Process Systems Enterprise Limited
Optimization-based design procedure (2/2)
3. Select constrained variables- any variable calculated by the model- specify upper and/or lower bound for constrained variable
4. Launch optimization
5. Inspect results.- pay attention to Lagrange Multipliers of decision or constrained
variables at bounds: estimates of improvement in objectivefunction that could be achieved by relaxing bounds.
6. Adjust bounds if appropriate, and launch optimization again.
© 2013 Process Systems Enterprise Limited
Example: Design of Reactor for Production ofCommodity Chemical at a U.S. Refinery
© 2013 Process Systems Enterprise Limited
Background
< Reaction NetworkA + Bà I + C (exothermic)I + Bó P + C (exothermic)B + Bó E + C (exothermic)BàW (high temperature)
< Several catalysts available< Non-isothermal bench-scale
experiments at high conversion< Challenge: design commercial
reactor with only 3 weeks of pilottesting
A : Refinery byproductB : Purchased raw materialI : Reaction intermediateC : Reaction byproductE : Reaction byproductP : ProductW: Waste
© 2013 Process Systems Enterprise Limited
PSE’s approach (1/2)
< Literature review- Basis for LH rate expressions- Identified species absorbed strongly on catalyst surface
< Re-evaluation of small scale experiments using AML:FBCR< Designed experiments to be carried out in pilot plant
- Low conversion experiments- By-product co-feed experiments
< Data collected for 17 days. Data collected in the first 15days was used to derive the model.
© 2013 Process Systems Enterprise Limited
PSE’s approach (2/2)
< Commercial design through optimization of validatedmodel, including recycling costs:- Number of tubes- Tube length- Tube pitch- Tube diameter- Number of baffles- Baffle cuts- Bed lengths- Choice of catalyst for each bed
© 2013 Process Systems Enterprise Limited
Results
< Commercial reactorsbuilt as designed
< Plant started up in2010.
< Production target met
© 2013 Process Systems Enterprise Limited
Example: HPPO process optimisationSimultaneous design of reactor and separation section
Advanced Modelling of
Fixed-BedCatalytic Reactors
© 2013 Process Systems Enterprise Limited
Hydrogen Peroxide route to Propylene Oxide (HPPO)
<Multitubular reactor- exothermic reaction
<Many distillation columns- two reactive distillations- one azeotropic distillation
< 25 components- reactants, solvent, product,
byproducts, impurities
< 2 large recycle streams
© 2013 Process Systems Enterprise Limited
HPPO process design
49 continuous and discretedecisions
< Too many decisionsfor trial-and-error simulation!
< Apply model-basedoptimisation technique:“MIXED INTEGER NONLINEAROPTIMISATION”
© 2013 Process Systems Enterprise Limited
HPPO process design – key results
1. Process design meeting all constraints- confidence arising from detailed models & formal model validation- good basis for Detailed Engineering & Operational Support
2. Large reductions in annualised cost- tens of millions of €/year compared to Base Design- two columns eliminated entirely
- €5m/year saved from one column alone
3. Significant operating cost savings from heat integration- attractive ROI: payback < 4 months
© 2013 Process Systems Enterprise Limited
The Advanced Process Modeling Approach: Step 4
1 2 3 4Optimization-
BasedDesign
Model-targetedExperimentation
+ParameterEstimation
AdvancedProcess Modelwith all physics
relevant toproblem of
interest
FinalAdjustments to
EquipmentDesign
© 2013 Process Systems Enterprise Limited
Hybrid CFD / gPROMS - Overview
CFD (e.g. ANSYS Fluent®)< Complex geometries< Modeling mixing and
turbulence key capability< Spatial domain discretized in
105 – 106 cells< 10’s of variables per cell
- Practical limit on number ofreactions and species
gPROMS< Relatively simple geometries< Model processes of
arbitrary complexity- Population balances- Large reaction networks- Intraparticle resistances
< 102 -104 variables perdiscretization element
Combination provides predictive capability of performance ofindustrial-scale unit operations with practical use of computerresources
© 2013 Process Systems Enterprise Limited
Comprehensive performance assessmentfor commercial-scale multitubular reactors
à Highest-accuracy predictive model on both tube-and shell sides
Wall heattransfercoefficient;coolanttemperaturethroughoutshell
CFD modelof shell
side
Heat fluxdistributionthroughoutshell
via interpolation ofrepresentativetubes
Heat fluxat surface ofrepresentativetubes
gPROMS modelof tube side
Wall heattransfercoefficient;coolanttemperatureat surface ofrepresentativetubesbaffle
gPROMS/CFDs/w interface
© 2013 Process Systems Enterprise Limited
Flow characteristics
Velocity magnitude [m/s]Particle tracks
coolant inlet
coolant outlet
© 2013 Process Systems Enterprise Limited
Example of multitubular reactor modellingTemperature profiles across pilot reactor
x = 0plane
z = 1.05mplane
Tube/coolantheat transfercoefficient [W/m2.K]
Tube surfacetemperature [oC]
ΔT = 16
Tube centretemperature [oC]
ΔT = 60
Coolanttemperature [K]
ΔT = 7
© 2013 Process Systems Enterprise Limited
Example:Debottlenecking of existing partial oxidation reactor
© 2013 Process Systems Enterprise Limited
Geometric configuration
Actual reactor drawings used to build the CFD model
© 2013 Process Systems Enterprise Limited
Tube Center Temperature
• Maximum horizontal temperaturedifference: 40℃
• This difference is highlydependent on the coolanttemperature difference.
• High temperature regions :around the coolant outlet, aroundthe reactor wall.
E B D
A
B
D
E
C
© 2013 Process Systems Enterprise Limited
520
540
560
580
600
620
640
660
680
700
720
740
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4R
RR
.cen
ter_
tem
pera
ture
RRR.Distance
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
RR
R.organic_m
ol_fraction
RRR.center_temperature(0,) (left) RRR.center_temperature(0,) (left)RRR.center_temperature(0,) (left) RRR.center_temperature(0,) (left)RRR.center_temperature(0,) (left) RRR.center_temperature(0,) (left)RRR.center_temperature(0,) (left) RRR.center_temperature(0,) (left)RRR.organic_mol_fraction(0,1,) (right) RRR.organic_mol_fraction(0,1,) (right)RRR.organic_mol_fraction(0,1,) (right) RRR.organic_mol_fraction(0,1,) (right)RRR.organic_mol_fraction(0,1,) (right) RRR.organic_mol_fraction(0,1,) (right)RRR.organic_mol_fraction(0,1,) (right) RRR.organic_mol_fraction(0,1,) (right)
Performance Difference between TubesTube Center Temperature and Reactant Mass Fraction
• Different Tube Center Temperatureprofile between tubes
• Different Feed Consumptionbehavior in the middle of the tubes
• But similar conversion of feed atthe end of tubes
4
1940
43
61
80
2 84
2194380
44061842
194380
4406184
Tmax
© 2013 Process Systems Enterprise Limited
Debottlenecking example – key results
1. Combined CFD/gPROMS model identified tubes proneto high temperatures- Improved placement of thermocouples
2. Combined model used to study relationship between feedcomposition/rate and maximum temperature variationsamong tubes
3. Combined model allowed to find operating point thatincreased production while keeping all tube temperatureswithin safety limits
© 2013 Process Systems Enterprise Limited
1. Model-based design, scale-up, and operational optimization offixed bed catalytic reactors and processes is being applied inindustry today
2. Applications include: catalyst pellet design, reactor design,whole process design, optimization of operation throughoutcatalyst deactivation cycle
3. Necessary models and modeling platforms are available ascommercial, “off the shelf” software
top related