scale-up of safety data using dynochem. tom vickery
TRANSCRIPT
Scale-up of Safety Data using Dynochem3rd Process Safety Forum
Wyeth-AyerstPearl River, NJ Oct 14th, 2008
T.P. Vickery, Merck & Co. Inc.
Dynochem Overview
• Process Modeling and Simulation Tool• Currently Excel-based• Can do fitting, simulation, optimization,
vessel characterization, physical properties
One key point!• In order to use this (or any) modeling tool, draw a
model of your process and list the key parameters that describe your model.
• For example for an ARC run, a typical model might be: with heat generation.
• Parameters would be Δ, amount of A, amount of solvent, φ, reaction start temperature, activation energy and pre-exponential factor
PA→
Overview
• Safety Investigation and scale-up risks• Potential gas generation on heat-up• Cold feed to hot batch at scale• Catalyzed destruction of a peracid
Why use Dynochem ?
• Integral Fit of Data – with Visualization
• Consistent Model for Scale-Up
• Modeling of What-if scenarios
Case 1: Dynochem modeling of an unstable cryogenic reaction
• Incident in small-scale prep lab believed related to decomposition of ArLi
• Aryllithium solutions are known to be unstable
• Possibly 2 ArXLi X-Ar-Ar-Li + LiX• 2 Exotherms – Heat of Addition (feed-limited)
and Heat of decomposition (T-dependent)
Approach
• Use the OmniCal Z-3 to obtain the heats of reaction– Heat of ArLi formation: -160.2 kJ/mole – Heat of ArLi Decomposition: -524 kJ/mole
• Use Dynochem to model the temperature- dependent portion of the reaction
Decomposition Data for Aryllithium (from Omnical Z-3)
0 100 200 300 400500
0
500
1000
1500
Heat1jmW
Baselinej( )
mW
Timejmin
80 60 40 20 0 20 40500
0
500
1000
1500
Heat1jmW
Baselinej( )
mW
TempjK
Experiment using 2.3 millimoles of aryl substrate
Heat Flow vs. Time Heat Flow vs. Temperature
Model Fit vs. Data
• Fit of a first order reaction (k, Ea) to the scanning Z-3 experiment
• Other models tried – no real improvement
Effect of BuLi addition Time – Generic 100 gallon vessel
Temperature and Decompostion vs Add Time
-60
-50
-40
-30
-20
-10
0
0 2 4 6 8 10 12 14 16 18
Feed Time (hr)
Tem
p(°C
)
0.00%
1.00%
2.00%
3.00%
4.00%
5.00%
6.00%Tmax
Impurity
Effect of BuLi addition Time – Generic 1000 gallon vessel
Temperature and Decompostion vs Add Time
-60
-50
-40
-30
-20
-10
0
0 2 4 6 8 10 12 14 16 18
Feed Time (hr)
Tem
p(°C
)
0.00%
1.00%
2.00%
3.00%
4.00%
5.00%
6.00%
TmaxImpurity
Feed Rate Control Case
To Control at For 50 gal in a 100 gal reactor
For 500 gal in a 1000 gal reactor
-50°C RateTime
0.093 L/min(16 hr charge)
0.45 L/min(32 hr charge)
-45°C RateTime
0.252 L/min(6 hr charge)
1.25 L/min(12 hr charge)
-40°C RateTime
0.402 L/min(3.75 hr charge)
2.11 L/min(7 hr charge)
Optimize Feed Time vs. Target Purity
Reactor Size
% Decomposition
100 gal 1000 gal
1% 87 min 232 min
0.2% 148 min 412 min
How Dynochem Helped
• Modeling the data, which had an imposed temperature
• Ability to simulate various run conditions to determine effect of parameters
• Ability to optimize to determine target addition time.
Case 2 - Gas Generating Reaction
• A malonate ester is heated to drive off CO2
• Gas data was collected off-line using a mass flow meter with totalizer during an RC-1 run
• Heat flow data was available from the RC- 1 experiment
The problem
• First analysis showed that heating to 80°C was too hot – not needed.
• What effect does heat rate have on gas generation?– Peak gas generation rate constrained by vent
piping (850L/min)• Fixed Jacket Rate – can it lead to a
dangerous runaway?
Approach
• Use Dynochem to fit a first-order reaction model to the combined heat / gas data.
• Use simulator to test the effect of reactor- temperature controlled heat rate
• Use simulator to raise the jacket temperature at a fixed rate.
Omit this slide
• The totalized gas flow was normalized to the theoretical: 0.11 moles total
Model before any fittingBulk liquid.Temperature (Imp) (C)Bulk liquid.Product (Exp) (mol)Bulk liquid.Qr (Exp) (W)Bulk liquid.Product (mol)Bulk liquid.Reagent (mol/L)Bulk liquid.Substrate (mol)Jacket.Temperature (C)Bulk liquid.Temperature (C)Bulk liquid.Volume (L)Bulk liquid.Qr (W)GasGen (mol/min)GasFlow (L/min)
Fit k> and dHr to data for Expt 1 (80 mL)
Time (min)
Proc
ess
prof
ile (s
ee le
gend
)
0.0 19.2 38.4 57.6 76.8 96.0-8.5E-4
0.0392
0.0792
0.1192
0.1592
0.1992
After fitting k and ΔHr
Bulk liquid.Temperature (Imp) (C)Bulk liquid.Product (Exp) (mol)Bulk liquid.Qr (Exp) (W)Bulk liquid.Product (mol)Bulk liquid.Reagent (mol/L)Bulk liquid.Substrate (mol)Jacket.Temperature (C)Bulk liquid.Temperature (C)Bulk liquid.Volume (L)Bulk liquid.Qr (W)GasGen (mol/min)GasFlow (L/min)
Fit k> and dHr to data for Expt 1 (80 mL)
Time (min)
Proc
ess
prof
ile (s
ee le
gend
)
0.0 19.2 38.4 57.6 76.8 96.0-0.065
3.935
7.935
11.935
15.935
19.935
After the Ea Fit
Bulk liquid.Temperature (Imp) (C)Bulk liquid.Product (Exp) (mol)Bulk liquid.Qr (Exp) (W)Bulk liquid.Product (mol)Bulk liquid.Reagent (mol/L)Bulk liquid.Substrate (mol)Jacket.Temperature (C)Bulk liquid.Temperature (C)Bulk liquid.Volume (L)Bulk liquid.Qr (W)GasGen (mol/min)GasFlow (L/min)
Fit k> and dHr to data for Expt 1 (80 mL)
Time (min)
Proc
ess
prof
ile (s
ee le
gend
)
0.0 19.2 38.4 57.6 76.8 96.00.0
0.04
0.08
0.12
0.16
0.2
After the Ea Fit
Bulk liquid.Temperature (Imp) (C)Bulk liquid.Product (Exp) (mol)Bulk liquid.Qr (Exp) (W)Bulk liquid.Product (mol)Bulk liquid.Reagent (mol/L)Bulk liquid.Substrate (mol)Jacket.Temperature (C)Bulk liquid.Temperature (C)Bulk liquid.Volume (L)Bulk liquid.Qr (W)GasGen (mol/min)GasFlow (L/min)
Fit k> and dHr to data for Expt 1 (80 mL)
Time (min)
Proc
ess
prof
ile (s
ee le
gend
)
0.0 19.2 38.4 57.6 76.8 96.00.0
4.0
8.0
12.0
16.0
20.0
Comparison of ΔHr
• RC-1 – Integration of Heat Flow:• -157.2 kJ/mole
– Automatically a “good fit” as it is just a numerical integration of the heat flow
• Dynochem – Model fitting• -140.2 kJ/mole
– The good fit and the good agreement between the two values give confidence that the model is reasonable, and that the integration is working
Gas flow from a ramp in a 100 gal- reactor
Rate of Temperature
Increase(K/m)
Peak Gas Flow (L/min)Tj ramp
Peak Gas Flow (L/min)Tr ramp
0.5 49 44
1 87 84
1.5 101 124
2 105 164
3 108 240
How Dynochem Helped
• Fitting to two different data sets
• Graphical representation of fits
• Use of data in actual reactor model
• Able to demonstrate reasonable heat-up profiles could not generate excessive gas flow
N-Oxide Formation
• Heat of Reaction and 1st-order rate constant at 52°C available from a CRC experiment
• Charge of cold reagent to warm batch• Avoid overcooling (reaction stalling) and
overheating (potential gas generation)
Approach
• Use vessel estimation tools to calculate heat transfer parameters– 1000L vessel UA=(1.04*V(liter)+159) W/K
• Estimate the activation energy as 125 kJ/mole (30 kcal/mole)
• Set up a “Universal” model in Dynochem– Allows for specification of a wide variety of
parameters in Excel
Items Bulk Liquid Bulk LiquidBulk
Liq uid
Bulk Liquid Bulk Liquid
Variables Volume Temperature Substrate Reagent Solvent
Units L C kg kg kg
Imposed Jacket Scale-up 450 55 30 0 420
Imposed Tr data 2 450 55 30 0 420
Batch Mode data 3 660 25 30 30 420
Adiabatic Batch data 4 660 25 30 30 420
Feed tank Feed tank Feed tank Feed tank Dosing Jacket Jacket Jacket Jacket
Volume Temperature Reagent Solvent Qv UA UA(v) coolant Temperature
L C kg kg L/min W/K W/L K kg/s C
210 20 30 180 3.5 154.19 1.04 2 55
210 20 30 180 3.5 154.19 1.04 2 55
210 20 30 180 0 154.19 1.04 2 55
210 20 30 180 0 0 0 2 55
Comparison of the effect of addition time with a fixed jacket temperature
30 min
1 hr
Comparison of the effect of addition time with a fixed batch temperature
30 min
1 hr
Temperature profile if run in batch mode – 55°C Jacket
Batch Mode – Stepwise heating
How Dynochem helped
• Incorporate reaction data from other sources
• Run multiple studies in one simulation• Visual comparisons of scenarios• Easy varying of parameters (feed rate,
jacket temperature)
A case study - peracid
• Highly exothermic decomposition
• Currently treated with sulfite (3 tanks)
• Thermal degradation (one tank)?
A case study – peracid
• Dynochem for Data Regression
A case study – peracid
• Dynochem for Destruction Profile
A case study – Peracid
• Vessel modeling and safe jacket temperatures (Dynochem and MathCAD)
10 20 30 40 500
50
100
150Heat GenerationHeat Transfer
Figure 2 - Semenov Plot for 25°C Jacket Temp
Reactor Temperature °C
Hea
t (kW
)
0.309
1.576
-1
0
1
2
3
4
5
6
-3 -2 -1 0 1 2 3
Hei
ght (
m)
Conclusions
• Dynochem is very useful for generating a kinetic fit from temperature-scanning data
• Dynochem provides direct visual feedback during the fitting in addition to the fitting statistics
• The kinetic model parameters from Dynochem can be plugged directly into the real equipment model
• The data needed for Dynochem is obtained as part of Merck’s standard testing.