simulation - cit-wulkow.de
TRANSCRIPT
Simulationunique numerical
algorithms
on-line information
direct interaction
open data formats
off-line capabilities
PDE's for spatial
profiles
3D-visualization of
time dependent
spatial problems
PRESTO is a sophisticated tool for the modeling of any kind of kinetic reactions.High-end mathematical techniques are combined with a state-of-the-art user-interface. Within this framework the definition of reactors s, coefficient or speciesis completely to the user. All data may be plotted to graphical outputopenwindows during simulation. The output be updated after each individualcanintegration time-step. Furthermore, during simulation information about thetheactual chemical and numerical problem status is provided. The full simulation orselected parts of itcan be recorded for later examination.
The ASCII format for in- and output files facilitates editing anyof information bysoftware of your convenience.
A sophisticated parameter estimator helps deriving the rate coefficients fromexperimental data. Standard algorithms, stochastic algorithms, andbox search procedures incombination with contourplots provide a detailedproblem analysis.
Experimental data can bevisualized together withthe simulation.
extensive numericalinformation
Spatial problems can be formulated using partialdifferential equations (PDE) considering diffusion,convection and spatial differentiated reactionrates. The highly efficient solver cares forautomatic, self-adaptive and non-equidistantdiscretization with respect to the spatialcoordinate and integration time axis. Thus thecomputational efforts are reduced to a minimum.
Results are showneither as the fullprofiles at each timestep as1, time slices ata certain spatialcoordinate2, timedependent integratedvalue of each profile ,3
or time dependentasevolution ofgraphsspatial speciesprofiles4.
Example:
Polymer production andspatial equilibration afterpoint-wise initiator feed.
(1)(2)
(3)
(4)
(4)
Declare components using names of your ownchoice and define their thermodynamic properties.
Implement your reaction systemfollowing intuitive steps.
Define a module with 3 clicks!
Interfaceeasy-to-use GUI
flat dialog
structure
intuitive
efficient
Reactors: An arbitrary number of reactors can be defined, each having individual characteristicssuch as: temperature(-profiles), pressure, volume, heat-balance, operation mode (batch,feed-batch, continuous flow, CSTR or tubular). A powerful phase concept enables thetreatment of multiple phases for .heterogeneous reactions
Coefficients: Rate coefficients and constants within a model may be defined as being temperatureand pressure dependent following an Arrhenius term.
Reactants: Species and species profiles defined considering their thermodynamicare byspatialproperties, initial values of concentration and profiles or feed-strategies.
Reaction steps: Most of the reaction systems may be built up using the general kinetic step shownabove. It provides any order of reaction with respect to the reactants or the directimplementation of the reaction rate. The stoichiometry may be defined totally free aswell as the overall rate. Furthermore, there exists a large variety of modules especiallydesigned for certain application fields. These include bio-kinetics, diffusion, convection,mass-transfer and several types of partial differential equations.
The creation of a model is supported by the comfortable Workshop. This Workshop is shared by all CiTtools to unify model implementation and to switch easily in-between the CiT tools without extendedphases of individual training.
Any component that is necessaryto picture a process can bedefined by templates. An appro-priate set of resources from the
Windows GUI (graphical user interface) expects the user's input. It is demonstrated here for thedefinition of a general reaction step.
The definition of recipes allows anefficient control of the reactor operationstrategy. The concept of tank recipescorresponds to strategies of batch andfeed-batch operation as well as ofcontinuous flow. Within each tank evenspecies mixtures can be defined. For thesefeeds it is possible to import mass versustime profiles directly as ASCII data fromyour balance. When modeling processesthat require multiple reactors, for eachreactor an individual recipe can bedefined.
The reactor module accounts forcontinuous and discontinuousdischarges from the reactor.Complete sampling programs can besimulated using this module.The volume control can be setmanually or automatically; full massbalances are calculated in any case.
Special file formats for files containing experimentaldata allow recipe control by measured data sets.
Controlrecipe concept
batch, feed-batch,
CSTR, tubular,
complex operation
modes
tracing experimental
data
Other special features are:
- By combining measured data files with recipes complex reactor operation programsare simulated that account for temperature and feed programs.
- Using this feature in combination with the parameter estimation tools, multiple ratecoefficients can be estimated from whole sets of experimental data referring todynamic reactor operation strategies simultaneously.
- Individual scaling or weight factors as well as penalty functions can be defined toaccount for differing precision of experimental data.
Data Editingfast manipulation
merging of different
data sets
quick display
As an extra feature, acustomized editor for thehandling of measured data filesis available. It simplifies editingdata and key words consideringall features of the parameterestimation tool. This tool is alsocapable of mergingexperimental data resultingfrom measurements ondifferent time bases.
A graphical displayis included for quickdata inspection.This is useful fortesting estimationprocedures by in- orexcluding groups ofmeasured data.
The reactor control ofmeasured data is superiorto the definition of recipes.
AdvancedFeatures
bio-kinetics
heat transferthermodynamics
multiple phase
concept
internal script
language
library
A flexible reactor concept providesstand-alone reactors or cascades. Foreach reactor the heat-balance andmultiple phases can be defined. Specialmodules allow the species exchange in-between the phases.
A b - sio kinetic pattern1 assist inmodeling this significant class ofprocesses.
Partial differential equations3 (PDE-modules) facilitate the implementationof spatial profiles, particle distributionsor reactions in heterogeneous reactionmixtures.
Using the integrated functioninterpreter (no external compilation)arbitrary functionalities may beimplemented for heat-balance , phase-4
exchange, particle-formation,-agglomeration or -degradation.
For special demands ordinarydifferential equations (ODE -modules) 2
provide direct access and control e.g. tospecies, species distributions or balancevariables.
There are several options to control feedand exit streams directly, using the:recipe or by ODEs.module
The library in combination with the function interpreter enables the calculation of self-defined quantitiesor implementation of control concepts, generating individual on-line output.
The syntax is intuitive and comparable to usualprogramming languages. Procedures may be usedfor structural implementation avoidingunnecessary and error-prone redoubling of code.
This is ideal for adopting a model to sensors thatare measured experimentally, e.g. in the casethese are not related directly to single species.
(1)
(2)
(3)
(4)
ParameterEstimation
rate parameter
Arrhenius
estimation
model parameter
statistical
information
no sheet or data
limitations
value scaling
value weigh ingt
free penalty
functions
Rate coefficients and anyother model parameterscan be fitted by theparameter estimationtool. As sensors anyspecies and systemvariable may act. Alsoquantities that aredefined by the user viathe function interpretercan be used.
An unlimited number ofdata sheets can beprocessed, eachcorresponding to anindividual experimenthaving its own recipe,operation mode,temperature and feedprofile.
Quick sensitivity tests canbe performed withoutchanging the generalsetup of the estimationarrangement via activatingor deactivating individualconstants or data sheets.
The configurationsummarizeses the datacollection, model and ratecoefficients as well asscaling and weightingpreferences together withpenalty functions.
Sophisticated statistical information isdisplayed on-line during the fittingprocedure. Results providing criteria for thegoodness of fit, correlation of parametersand individual sensitivity are available.
There are no principle limitations withrespect to the number of data sheets orparameters to be estimated.
Concerning rate coefficients, either theirabsolute values may be fitted and evaluatedas Arrhenius term further on, or theArrhenius parameters may be estimateddirectly. During estimation limiting boundsmay be set.
ASCII log-files provide a history of the fittingprocedure for later inspection of the efficiencyof the estimation process. All statisticalinformation together with the course ofparameter estimates are collected. Evendetailed information about individualcontributions of single compounts to theresidual within each experiment is available.
The box-search is the idealtool for taking a first glanceat a fitting problem andgetting an idea aboutparameter sensitivities orcorrelations, leading toefficient initial values.
For the fitting of more thantwo parameters, thesensitivity of any of theircombinations may beinspected using the internalcontour plot function.
ASCII output files enable theprocessing and thevisualization of data usingother programs.
In order to avoid laborious screening, a simulatedannealing algorithm is implemented for a statisticalsearch. The combination of gradient and randomwalk techniques assures finding identical solutionsfrom varying initial conditions while concentratingcomputational efforts on the sensitivity area ofinterest.
The data are logged into ASCII files for laterinspection and graphical representation by othersophisticated programs.
k1
k2
residual
residual
k1
k3
The integrated 3D OpenGL visualization helps to inspectsensitivity matrices and correlations of estimated parametersefficiently. The surface may be zoomed, shifted and rotated byany axis. It may be displayed as colored or shaded surface aswell as transparent grid.
This feature is also available asstand-alone tool for thevisualization of matrices.
ParameterEstimation
box-search
simulated
annealing
3D-visualization
Connectibilityopen data formats
for import / export
batch automation
OLE capability
Web-functionality
The OLE automationserver enables widescreening of experimentalconditions and sensitivitystudies, controlled e.g. byVisual Basic or anyalternative OLE client.Large data quantities maybe processed followingautomated schemes.
HTML functionalitiessimplify documentation.The ouput files in HTML-format include the modeldescription, an archivethat may be run directlywith PRESTO andsimulation results. Thesefiles can be used as astand-alonedocumentation or asinput files for other textprocessors. Furthermore,they can be published onthe inter- or intranet.
k4
k5
PRESTO is not only a very efficient tool for kineticsystems described by ordinary and algebraicdifferential equations, but includes a leading solverfor systems of partial-differential equations.
Numerous modules for a general description ofconvection, transport, diffusion, dispersion,boundary conditions, agglomeration and arbitrarykinetic reactions allow an easy transfer of kineticschemes to distributed variables (e.g. axial or radialco-ordinates in a non-stationary tubular reactor,mixing problems, general PDE's). Parameter andcoefficients of these processes can be formulated asspatial dependent.The Galerkin h-p-algorithm works full self-adaptively, all discretizations are error-controlled. Inconnection with a parameter estimation procedure,the algorithm is even more efficient. The drawbacks
of other solvers (poor error control, missing adaptivity, large number of approximating variables,restriction to local operators, method-of-lines approach), are avoided and overcome in PRESTO. A seriesof challenging problems could be solved with this tool in a unique way. Even free boundary problems(e.g. single particle models) can easily be solved in PRESTO usingthese powerful features. The handling of large PDE systems is assimple as for standard kinetics.
PartialDifferentialEquations
spatial gradients
species profiles
dynamic
computation
By this, in the same manner as forstandard models complex simulationmodels may be defined by linkingelementary basic objects. Recycleoperators provide the facility tomodel loop reactors in an intuitiveway. For results of thepolymerization model shown heresee page 2.
The profile species may bedisplayed as timedependent integral, timeslices at a certaincoordinate using theoutput functions of thelibrary or as full timedependent 3D profilesusing OpenGL techniques(see also page 2).
The example graphs showthe behavior of a tracerpulse within a loopreactor where thebackmixing characteristicis modeled by diffusionand convection operators.
ExtendedProjectManagement
multiple projects
for a single model
quick process
change
direct access to
measured data
PRESTO is available in a secondversion (PRESTO-Kinetics II), thatsupports extended projectmanagement and shares theoutstanding mathematicaltechniques.
It is developed to accompany multi-purpose plants. Plant specific datasuch as reactor configuration andspecified substances are storedsuperior to individual projecttasks.
The project tasks contain processspecific items such as the kineticmodel, rate coefficients, recipesand parameter estimationconfiguration.
Any of the project tasks has access to a commondatabase of experimental data for comparison andparameter estimation.
By changing tasks models may be tested veryefficiently without changing the gobal project. Anydevelopment of model or process adaption issummarized within the same gobal project to keepthings together.
Moreover, an interface is available to link species todatabases for calculation of thermodynamicproperties such as activity coefficients.
Behind this advanced interface for multipleproject management there exist the same user-friendly dialogues as in PRESTO-Kinetics.
Therefore there is no need for retraining whenswitching to PRESTO-Kinetics II and back.
Project tasks are fully compatible with modelsof PRESTO-Kinetics II and vice versa.
Data Hologramfree label and
group definition
object oriented
data classification
application
bounded links
for files
graphical
representation of
data
Initial situation:
Unstructured collectionconsisting of individualdata files withinhomogeneouscontent ...
... collect labels anddata ranges togroups of your ownchoice and parsethem hierarchicallyto find the data setsof your interest ...
multi-dimensional overview of extended file collections for experimental data
hierarchical structurization follows flexible differing aspects of interest
labels for structurization are based on the individual files content
user-defined labels that describe numerical data, abstract meanings, ranges of subjects
user defined filters
summarizing labels or numerical data in self-defined groups
(easy) automatic scanning of labels
Organization using the hologram:
User defined labels in arbitrary orderdescribe the content of the data fileand build the basic structure thathas to be followed ...
Dr. Michael Wulkow Computing in Technology GmbHOldenburger Straße 200D - 26180 Rastede
Germany
Tel.: +49 (0) 44 02 / 84 2 48Fax: +49 (0) 44 02 / 939 [email protected]://www.wst.shuttle.de/cit/
References
Further Tools
The products of CiT are used world-wide in academia and industry.
On your request it will be a pleasure for us providing a comprehensive list of reference customers and
projects.
Version 3.03.06
Potential applications of PRESTO range from simulating general kinetic schemes up to treating
2-dimensional problems or bio-kinetics. And there are more products of CiT that share all the same
outstanding numerical technology:
Predici: rigorous treatment of complete molecular weight distributions in arbitrarypolymerization processes
reaction step patterns for propagation, termination, transfer (monomer, agent,polymer), depolymerization and polymer degradation, description of oligomers, chain-length dependent reactions
copolymerization with any number of monomer species
treatment of composition and branching distributions
particle size distributions in heterogeneous systems
parameter estimation (multiple experiments)
temperature- and pressure profiles as well as recipes for reactors or cascades of reactors
manipulation of kinetic rate coefficients and reactants by use of an interpreter mechanism
analysis of polymer properties
input of measured concentrations as well as molecular weight distributions
Parsival: rigorous treatment of particle size distributions in heterogeneous process units (e.g.crystallization)
reaction step patterns for particle growth, nucleation, agglomeration, breakage, fine'sdissolution, convection, diffusion, boundary conditions, kinetic reactions
flow sheets with classified flow and exit
parameter estimation
Rioval: management tool for water quality modeling in streams
interface for the description of rivers and input from treatment plants
simulation based on a comprehensive and well-tested biofilm model