Use of DynoChem in Process Development. Wilfried Hoffmann.

Use of DynoChem in Process Development
by Wilfried Hoffmann
Old: Chemical R&D, Sandwich, UK
Worldwide Pharmaceutical Sciences

New: Scale-up Systems, Dublin, Ireland

DynoChem Scale-up Systems 1

The Fundamental Problem of Scale-up

The major objective of Process Development is the design of a sequence of
operations, which allow the safe and ecologically responsible manufacturing
of Active Pharmaceutical Ingredients at a scale demanded by market, in a
quality demanded by Regulatory Authorities, and at the lowest achievable
cost

This development is based on lab scale experiments

?


Traditional approach:

Lab Reaction Development Pre Scale-up Scale-up

Lab Robustness testing Risk of
Design Process Safety testing
Experiments Failure

This approach underestimates the effects of physical rates
on the overall performance

are functions of scale
- rate of heat transfer
and equipment and
- various rates of mass transfer
can compete with
- various rates of mixing
chemical rates


Process Development needs to consider scale and equipment

As large scale development experiments are prohibitive with respect to
cost, safety, and time but large scale performance information is required
the solution is:

Process Modelling

Process Modelling allows the prediction of the interactions of chemical and
physical rates as a function of operating conditions, scale, and equipment


Lab Design approach => Model based approach:

Process Understanding based
Design Scale-up
Model Generation
Model + Equipment data
Lab Data Predicted
Large Scale Process
Experiments (Model) Performance
Optimization

Experiments are performed to generate Process Understanding, not
necessarily to get good yields in the lab.
This Process Understanding is then captured by First Principles
Mechanistic Models

A software package used by Pfizer which supports the generation and
capture of this information is Scale-up Systems’ DynoChem

Process Understanding

What is Process Understanding?
In this context Process Understanding is the necessary required knowledge
to allow predictions on the process behaviour on scale
How can we access this knowledge?

The first action is an analysis of the different rate processes (elements)
in our process (for illustration I am using a chemical reaction, but the
same principles can be applied to other unit operations)
DynoChem uses a visualisation tool, which is very useful for the early part
of this modelling approach

In the following this tool will be demonstrated for a semibatch reaction in a
jacketed reactor with a solid phase present


FEED TANK BULK LIQUID Solvent, Tr0 Element 1:
The chemical rxn system
Chemistry
Including heat generation
Solvent, A,
T dos Element 2:
Heat transfer
Flow rate
B
UA Element 3:
(kLa)1 Heat Dosing mass transfer
B (s) out
H
SOLID
Element 4:
Solid/liquid system


Small Scale Large Scale

Analysis Translation Construction

Element 1 Element 1

Element 2 Element 2
Large Scale
Lab Process
reaction
Element 3 Element 3

Element 4 Element 4


First Principles Mechanistic Models are using Basic Rate Laws and
Thermodynamics combined with fundamental conservation of mass and
energy to present these elements
(in contrast to empirical or DoE type models)

Chemical Rate Laws:

Chemical Reactions are best described by a set of elementary reactions, i.e.
reactions on a molecular level. These reactions are either unimolecular (bond
scissions or rearrangements) or bimolecular (by collision of two species)
The advantage of this approach is that all unimolecular reactions are first order and all
bimolecular reactions are second order. The disadvantage is that a complex reaction
system will require a set of elementary reactions, each with a rate constant and an
Energy of Activation.
This approach may be very attractive to chemists, as rate models can be constructed
directly from their knowledge about mechanisms.


Physical Rate Laws:

In general are proportionate to a driving force
mass transfer rate: kLa ([A]∞ - [A]) (unit [conc/time])
Heat flow rate through jacket: AU (Tr-Tj) (unit [energy/time)

Thermodynamics:
Equilibrium and its temperature dependence is described by:
RT ln K  - H  TS
Conservation of mass and energy:
For example:
Mol balances in chemical reactions
Heat generation and heat removal control the degree of heat accumulation
(temperature change)


The conservation of mass sounds trivial, but for the description of chemical reactions
this appears to be one of the critical items in modelling

The reason for this is that most of the information of chemical reactions is generated
by LC based methods with UV-based detectors.

Raw data from these methods will only generate area% information of the detectable
species and no information about the mass balance

Before such data can be used for modelling they have to be converted to absolute
mol data. This can be done by using Relative Response Factors and reaction mol
balances of at least 95% accuracy

The consequences of not doing this homework will be shown by a simple example


The importance of the mol balance is demonstrated by a drastic example

Mass balance Analytical data

A+B→C

These data will not match

Either we have to change the mass balance (for example adding a rxn A → D),
or the analytical data are wrong and have to be corrected


The basis for modelling are time resolved profiles of experimental data

1) Analytical profiles
2) Heat generation rates
3) Additional online info (ReactIR, pH, gas generation, H 2 uptake, etc...)
4) Accurate temperature profiles

Experimental Data: Kinetic model:
moles

moles

time time

Example System

The following example system, which has been used in several DynoChem training
courses at Pfizer and which was related to real processes, will demonstrate the data
flow and the way of model generation for the scale-up of an exothermic semi-batch
reaction
Starting point is a simple reaction
k1
A + B P r1 = k1 [A] [B]
k2
A + P SP r2 = k2 [A] [P]

This reaction was run in the lab at 60oC and there were seen these 4 species
with a mass balance close to 100% . An analytical method was developed and
Relative Response Factors were measured. The reaction was followed against
an Internal Standard and so the HPLC data could be converted to absolute mol
data

Example System

This reaction system element in DynoChem is presented by a block of lines
Reactions in Bulk liquid

k> 1.00 E-03 L/mol s at 60 C Ea> 60 kJ/mol dHr 0 kJ/mol * A + B > P

k> 1.00 E-04 L/mol.s at 60 C Ea> 60 kJ/mol dHr 0 kJ/mol * A + P > SP

In this block there are estimated values for the rate constants and the Activation Energy and there
is no value of the exotherm (dHr = 0 kJ/mol) available, which will probably be the knowledge in an
early development stage.

If not otherwise indicated (it can be done if required), DynoChem assumes that the reactions after the * are
elementary reactions, so the rate laws are strictly first order in each component
i.e. d[P]/dt = k [A] [B] and d[SP]/dt = k [A] [P]

To get real rate parameters (k1 , k2 ,Ea1 , Ea2 ), a set of 4 experiments were performed
with a different ratio of [A]o / [B]o and at 4 different temperatures (40 o C, 50o C, 60o C,
and 70o C)

Example System

After fitting all the experimental data can be reproduced with just four rate parameters
k1, k2, Ea1 , Ea2

Example System

With a calorimetric experiment the individual heat of reactions can be determined
as well:


k> 2.7 E-03 L/mol s at 60 C Ea> 60 kJ/mol dHr -150 kJ/mol * A + B > P

k> 5.0 E-04 L/mol.s at 60 C Ea> 90 kJ/mol dHr -80 kJ/mol * A + P > SP

Example System

Process Safety data were generated directly together with the calorimetric run, when
a sample of the reacted mixture was subjected to a thermal stability investigation with
an ARC (Accelerating Rate Calorimeter)

This revealed a dangerous decomposition reaction at a higher temperature. The
kinetics of this decomposition was evaluated from the ARC data with DynoChem
and the result could be included in the kinetic description:


k> 2.71E-03 L/mol.s Tref 60 C Ea> 59.997 kJ/mol dHr -149.86 kJ/mol * A + B > P
k> 5.02E-04 L/mol.s Tref 60 C Ea> 90.011 kJ/mol dHr -80.60 kJ/mol * A + P > SP
k> 5.00E-07 1/s Tref 60 C Ea> 140.000 kJ/mol dHr -420.00 kJ/mol * P > Dec

These data will not have a big impact on the reaction at 60o C, but are of major
importance for the safe scale-up:

Example System

These data allow now the prediction of the product composition for any ratio
of A and B (where A can be added by a dosing system over any given time)
at any given reasonable temperature as a function of time

For scale-up there is no given temperature, but the reaction temperature is the
result of the interplay between heat generation and heat removal.

Here we need to add a jacket to our model, and provide the parameters

As we want to predict temperature changes, we need to use reasonable good values
for the physical properties of the reaction mixture and the feed, for example cp

These data can be estimated or measured by the same calorimetric experiment where
the heat flows were obtained

Example System

The following lines describe the heat exchange between a reaction mass
and a jacket
Cool Bulk liquid with Jacket
UA 310.3 W/K
UA(v) 0.82 W/L.K
Temperature C
Cp 2.2 kJ/kgK
coolant 5.5 kg/s

Here the heat transfer UA is defined as a linear function of the liquid phase
volume with an intercept of 310.3 W/K and a slope of 0.82 W/L.K, so that
AU can be adjusted in case of a semi-batch reaction

These heat transfers can be measured or calculated by DynoChem with
a heat transfer tool

Example System

The following simulation shows the temperature profile of a 1000 L run with a simple
Tr-controller implemented with a feed time of 1 hr

Example System

With a feed time of 2 hrs and less excess of B the result looks like this

Example System

It appears that we are now in a position to design our process to get a combination
of the best temperature, the best feed time, the best ratio of A/B, and the best use
of reactor time as a function of scale and equipment
This is indeed possible and DynoChem has a built in functionality, which can
optimize any given process outcome or user provided functionality, for example
a whatever complex cost function.

This is tempting, however, we need to consider Process Safety as well

One of the standard scenarios in Process Safety is the question of the system
behaviour in case of a loss of cooling capacity in the worst possible moment.

This question can be answered by setting the cooling capacity to 0 and calculating
the temperature profile for this adiabatic system

Example System

A simulation run at 60 oC with a loss of cooling capacity at the end of the feed
(this is the stoichiometric point and the worst point in our system) will give a thermal
explosion (run-away) about 3 hrs later!!
This 3 hrs time is called Time to Maximum Rate (TMR) and can be used as a
quantitative measure of thermal risk

Once we agreed to an acceptable thermal risk (may be 8 hrs), we can then include
this in the optimization

At a first view this risk is likely to be a function of the reaction temperature, and we
might think that lowering the temperature will reduce the risk

This might be wrong! A simulation with a starting temperature of 20 oC will give the
result shown on the next slide

Example System

Keeping all other parameters constant, there is usually a temperature where TMR
Is a maximum, as shown below (for a 2 hrs feed time)

10
rxn time after end of feed for 99% conv [h]

15

8

6 10
TMR [h]

4

5

2

0 0
20 30 40 50 60 70
Tr set [C]

Summary

It is now possible to include the thermal risk into the optimization of the large scale
operation conditions
As a result we will get a process optimized with the consideration of scale and
equipment, i.e. a change of scale and equipment will change this optimum
This concept can be used to transfer a process from

 Lab to Kilo Lab
 Kilo Lab to Pilot Plant
 Pilot Plant to small scale Manufacturing
 Transfer within Manufacturing between different scales
 Transfer between different equipment types, i.e.
Batch / Semibatch to Plug Flow or CSTR !

This is a significant advantage over traditional Process Development

Use of DynoChem in Process Development. Wilfried Hoffmann.

More Related Content

What's hot (20)

Similar to Use of DynoChem in Process Development. Wilfried Hoffmann. (20)

More from Scale-up Systems (10)

Recently uploaded (20)

Use of DynoChem in Process Development. Wilfried Hoffmann.