Reactor simulation and modeling: methodology

First-principles modelling validated against experimental data provides unparalleled predictive accuracy

What PSE customers have said:

"The model we built with PSE's assistance gave us perfect insight into what was happening in our [multitubular] reactor"

Dr Sang Phil Han, LG Chem, Ltd

"PSE helped us to create a and validate a rigorous kinetic model of a complex industrial reactor system with less than one month of effort."

Dr Christoph Bäumler, Süd-Chemie AG

Step 1 – Experiment model

Construct a model of the experimental setup. This may be a Berty reactor, or a single-tube pilot plant.

Step 2(a) – estimate

Estimate parameters and analyse confidence information

Estimate kinetic parameters and analyse confidence information

Step 3 – build full model

Construct the full reactor model, linking the gPROMS reaction models to a CFD model if necessary

Step 4 – execute

Execute and generate results

Steps 2(a)–(c): model-centric experimentation

Model-centric experimentation is at the heart of Model-Based Innovation. Click to see the full picture.

Sometimes apprently good parameter values carry hidden risk. Click for further examples.

First-principles models validated against laboratory or pilot data

PSE's approach to reaction is to use rigorous first-principles models validated against laboratory or pilot plant data for all reactor modelling. This approach provides unparalleled predictive accuracy.

Typically we work closely with customers through a ModelCare^®agreement in order to:

build and implement fit-for-purpose models within a rapid time frame
maximise the capture and use of customer knowledge
identify any additional experimental requirements
ensure transfer of modelling know-how to the customer.

Within our Model-Based Innovation framework we have developed a well-proven methodology for creating a high-accuracy model of a complex reactor.

Some or all of the following steps are applied during a typical project.

Stage 1: Determine the definitive reaction set and kinetic parameters

Many chemical and pharmaceutical companies operate reaction processes without a clear definition of what is happening inside the reactor.

Until now, it has been extremely difficult to characterise a reaction set of more than a few tens of reactions and species, and provide accurate rate constants for each reaction.

However without a proper characterisation it is difficult to understand how to improve operation or what is going wrong when there are problems.

PSE provides a well-defined reaction characterisation service that applies model-based data analysis using:

your experimental data
a model of the laboratory or pilot plant apparatus from which the data were collected.

to determine kinetic multiple parameters to a high degree of accuracy.

The experimental data can be a combination of steady-state and dynamic experiments, and previous experiments can be taken into account. Instrument error can be estimated simultaneously with the kinetic parameters.

Usually we work closely with customers during this process to ensure that maximum knowledge is captured within the models, and modelling know-how is transferred to customer personnel.

Step 1. Construct first-principles models of the fundamental phenomena being studied

This involves creating a first-principles model of the system for which experimental data are to be fitted – for example, a single catalyst-packed tube in a pilot plant – rather than for the target equipment.

This is usually a straightforward model based on PSE's well-established kinetic models for homogeneous and heterogeneous reaction. Initial reaction kinetic parameter values, if not available from earlier customer models, are typically found by literature search.

The model is constructed in a modular form allowing the components to be separated so that they can be easily be utilised within the full reactor geometry subsequently.

Step2(a). Apply model-based data analysis to estimate parameters

The model created in Step 1 is then used to extract high-accuracy parameter information from experimental data using gPROMS' parameter estimation techniques.

In addition to parameter values, this process also yields estimates of the accuracy of these values in the form of confidence intervals.

The confidence information is used to determine whether the parameters are sufficiently accurate to use them for subsequent design and operational decisions. It also provides an indication of whether the reaction set is incomplete or inconsistent.

If the latter is the case, we help adjust the reaction set, identifying and adding any missing reactions and removing insignificant reactions, until results are consistent with observed data.

Stage 2: Proceed to the full reactor model

Once the parameter confidence analysis indicates an acceptable margin of error in the parameters, the model is considered to be validated. It is then possible to proceed to creation of the full model with confidence.

Step 3. Build the full reactor model

We can take care of a large part of this for you, by customising advanced reactor models from the PSE libraries. Alternatively, we can assist your personnel in doing this.

Because the models use Maxwell-Stefan relationships to handle detailed multicomponent mass and heat transfer accurately, they can be configured to take into account virtually any phenomena that affect the reaction process.

The validated reaction set developed in Steps 1 and 2 are implemented within the full reactor model. If Steps 1 and 2 were carried out correctly, the validated reaction set should be capable of predicting all scales of operation accurately.

If necessary, the full model can be validated against operating or test run data to determine, for example, overall heat transfer or flow coefficients. The kinetic parameters already fixed in Steps 1 and 2 should not be included in this estimation.

If complex geometry or strongly non-ideal mixing warrants it, we link the reaction model to a Computational Fluid Dynamics (CFD) model using the PSE multitubular or multizonal interfaces.

Step 4. Execute and optimise

Once the full validated model of the process is available, it is ready to be used in steady-state and dynamic simulation or optimisation studies.

We can either execute the required studies, generate and interpret results, and implement the model online if required, or alternatively assist your personnel to do this.

Stage 3: If the data doesn't fit:

Refining parameter information using model-centric experimentation

Sometimes the confidence intervals from the model-based data analysis in step 2(a) indicate that the parameters calculated from the initial experimental data are not within acceptable risk limits.

This means that further experimentation is required.

Rather than being aimed at providing, for example, sets of concentration data at certain temperatures, it is possible to direct these additional experiments solely towards generating information that increases the accuracy of model parameters to an acceptable level.

The process is known as model-centric experimentation, and it adds two further steps:

Step 2(b). Model-based experiment design

gPROMS provides comprehensive model-based techniques for the design of experiments.

In contrast to the usual statistically-based techniques (e.g. factorial design), model-based experiment design takes advantage of the information that is already available – in the form of the mathematical model – to design experiments which yield the maximum amount of information.

This minimises the uncertainty in any parameters estimated from the results of these experiments.

This optimisation-based technique is applicable to the design of both steady-state and dynamic experiments, and can take account of any experiments that have already been performed.

Typical decision variables determined by this technique include:

the optimal conditions under which the new experiment is to be conducted (e.g. the temperature profile to be followed over the duration of the experiment)
the optimal initial conditions (e.g. initial charges and temperature)
the optimal times at which measurements should be taken (e.g. sampling times for off-line analysis).

The overall effect is that the required accuracy in the estimated parameter values may be achieved using the minimum number of experiments.

Step 2(c). Model-targeted experimentation

The model-targeted experiments are then carried out, following the experiment procedure determined from the model-based experiment design in Step 2(b).

Following the experiment, model-based data analysis is applied to determine the parameter values and their accuracy. Steps 2(b), 2(c) and 2(a) are repeated until a satisfactory level of parameter confidence is obtained.

PSE ModelCare

All of the services described here are provided as part of PSE's standard ModelCare.

Request
a free consultation