Dynamic Behaviour of Process Systems

Isothermal CSTR Modelling and Simulation

Alastair Wong and Iniobong Akpan 15/01/2012

DBPS project

Alastair Wong and Iniobong Akpan

15/01/2012

1. Dynamic Model Development

In this project, a perfectly mixed isothermal CSTR is used to carry out the following methanol synthesis reaction:

This reaction takes place in the presence of a solid catalyst implemented within a cage integrated with the stirring mechanism of the CSTR system, as shown in the figure below:

Figure 1: Schematic of the heterogeneous gas/solid catalytic reactor system.

For the reaction over this particular catalyst, the reaction at the surface is rate-limiting, as a result giving rise to a Langmuir-Hinshelwood-type rate expression [r is in kmol/kg-cat.h]: ( ( ) )

where the values of the various kinetic parameters are tabulated below:
Table 1: Kinetic parameter values. Parameter kr KCO KH2 KCH3OH Keq Value -4 1.434x10 -2 4.111x10 -2 1.227x10 -2 4.367x10 -2 1.958x10

Furthermore, the outlet flowrate is determined by:

where

DBPS project

Alastair Wong and Iniobong Akpan [ ]

15/01/2012

1.1. Dynamic model

The first part of this project requires a model capable of predicting time variation of composition in the batch reactor. For simplicity of the model, the species CO, H2, and CH3OH are denoted , respectively. Various assumptions had been made when developing the model. For instance, perfect mixing assumption had been employed, as it is usually a reasonable assumption for a CSTR model. On the other hand, ideal gas assumption had been employed due to the conditions the gas is under in this problem. Following employment of the ideal gas assumption, Daltons Law becomes valid and was used as one of the simplifying assumptions. Finally, as the CSTR of interest is in the form of a confined reactor, the available volume inside the reactor was assumed to be constant. Hence, the relevant equations constituting the model, along with their respective variables and assumptions involved, are tabulated in the table below:
Table 3: Dynamic model of the system of interest. Equations No. of equations 3 Variables No. of variables 8 Assumptions Perfect mixing

1 ( ( ) ) 3 [ ] 1 3 1 1 15 1 1

0 3

1 0 0 1 0 15

Daltons Law Constant volume Ideal gas law

Total

Since the number of equations equals the number of variables, there is no degrees of freedom, and the model is complete.

1.2. Notation Table

The notations used in this model along with their respective units and quantities in which they represent are tabulated in the table below.

DBPS project

Alastair Wong and Iniobong Akpan

15/01/2012

Table 4: Notation table for the dynamic model for the system of interest.

Variables

Units h kmol kmol/h kmol/h kg-cat

kmol/kg-cat.h 0.5 K kmol/kPa.h K kPa kPa kmol/kg-cat.h.(kPa)3 1/(kPa) 1/(kPa) 1/(kPa) 1/(kPa)2 kPa kmol/h kmol/m3 m3 J/mol.K

Quantities Time Molar hold up of specie i Inlet molar flowrate of specie i Outlet molar flowrate of gas Mole fraction of specie i in reactor Catalyst weight Stoichiometric constant of specie i Reaction rate Valve position Valve constant Temperature Pressure Atmospheric pressure Reaction rate constant CO kinetic constant H2 kinetic constant CH3OH kinetic constant Equilibrium constant Partial pressure of specie i Total molar hold up Molar density Total volume Molar gas constant

2. Determination of the valve constant cv

For this part of the project, the value of the valve constant which would result in a reactor pressure of 1300 kPa under steady-state conditions for inlet flowrates of 1.0x10-4 kmol/h of CO and 0.25x10-4 kmol/h of H2 was determined, given that the valve was half-open (i.e. ). In addition, the following design parameters were given:
Table 5: CSTR design parameters.

Reactor Property Catalyst Weight (w) Total Volume (V) T

Value 0.01 kg 0.005 m3 475.15 K

2.1. Method
First of all, several variable types were defined, namely Molar_fraction, Molar_flowrate, Moles, Pressure, Reaction_rate, Molar_density, and NoType, along with their respective lower and upper bounds. In terms of the model, the following parameters and variables were listed and were given their parameter and variable types, respectively, as follows: 3

DBPS project

Alastair Wong and Iniobong Akpan

15/01/2012

Table 6: Parameters used in the model along with their respective gPROMS identifier and parameter types.

Symbol

Description Units gPROMS identifier Number of no_components components Atmospheric kPa Patm pressure Molar gas J/mol.K ideal_gas_constant constant Reaction rate kmol/kgKr constant cat.h.(kPa)3 Equilibrium 1/(kPa)2 Keq constant CO kinetic 1/(kPa) KCO constant H2 kinetic 1/(kPa) KH2 constant CH3OH kinetic 1/(kPa) KCH3OH constant Catalyst kg catalyst_weight weight Total volume m3 total_volume Stoichiometric reaction_stoichiometry constant of specie i

Parameter type INTEGER REAL (AS REAL DEFAULT 101.325) REAL (AS REAL DEFAULT 8.314) REAL (AS REAL DEFAULT 1.434E-4) REAL (AS REAL DEFAULT 1.958E-2) REAL REAL REAL REAL REAL REAL (AS ARRAY (no_components) OF REAL)

Table 7: Variables used in the model along with their respective gPROMS identifier and variable types.

Symbol

Description Inlet molar flowrate of specie i Outlet molar flowrate of gas Mole fraction of specie i in reactor Molar hold up of specie i Total molar hold up Reaction rate

Units kmol/h

gPROMS identifier in_molar_flowrate

kmol/h -

out_molar_flowrate molar_fraction

Variable type Molar_flowrate(AS ARRAY(no_components) OF Molar_flowrate) Molar_flowrate Molar_fraction (AS ARRAY(no_components) OF Molar_fraction Moles (AS ARRAY(no_components) OF Moles) Moles Reaction_rate

kmol

holdUp_moles

kmol kmol/kgcat.h

total_holdup_moles molar_reaction_rate

DBPS project

Alastair Wong and Iniobong Akpan

15/01/2012

Partial pressure of specie i Pressure Valve position Temperature Density Valve constant

kPa

partial_pressure

kPa K mol/m3 K0.5kmol/k Pa.h

total_pressure valve_position temperature total_density valve_constant

Pressure (AS ARRAY(no_components) OF Pressure) Pressure NoType REAL Molar_density NoType

In addition to the above, the equations mentioned in part (1) of the project was implemented into the model. Afterwards, a process simulation entity was created, and the parameters and variables were set and assigned, respectively. Since this part required a steady state model, the initial conditions were set as . Subsequently, the process simulation was run, and the value of

was found to be 3.7317966x10-6.

For details of how the model and process simulation were set up, the reader is referred to the entities Models gas_phase_CSTR_question2 and Processes Simulate_CSTR_question2 in the file CSTR Model.

3. Estimation of steady-state time

For this part of the project, the system was started from steady state. The inlet flowrates of the two reactants, however, had changed at several specified times during operation according to the following table:
Table 8: Operation procedure.

0.0 1.0 10.0 1.0x10-4 0.1x10-4 1.2x10-4 0.25x10-4 2.75x10-4 0.25x10-4 0.0 0.0 0.0 Therefore, the problem in this part of the project required estimating the time it took for the reactor to reach its new steady-state.

Time [h]:

3.1. Method
In order to set up the model for this problem, the model used for the previous part of this project was used, with some modifications. First of all, as the valve constant had been computed in the previous problem, the value had been assigned in the process simulation entity for this problem in gPROMS. Second of all, the schedule function was used to assign the sequence for the inlet flowrates of the reactants as described in the table above. The assigned flowrates after 10h, as specified in the last column in the table above, had remained steady for time t>10h. As the system had started from steady-state, the initial conditions had remained = 1, 2, 3. for i

DBPS project

Alastair Wong and Iniobong Akpan

15/01/2012

The system was set so that it had operated until time t=110h. Hence, the following graph of variations of hold up moles of the various components over time was obtained. For such system without the interference of any fluctuations, steady-state is defined as the state in operation where various variables remain steady values over time. It can be seen from the graph below that this state was reached at t=100h. Hence, it can be concluded that it had taken 90h for the reactor to reach its new steady-state.

Figure 2: Graph of molar hold up vs. time for steady state-time determination.

For details of how the model and process simulation were set up, the reader is referred to the entities Models gas_phase_CSTR_question3 and Processes Simulate_CSTR_question3 in the file CSTR Model.

4. Proportional-integral controller tuning

Starting from the steady-state computed in part (2), a proportional-integral (PI)controller was introduced to manipulate the outlet valve position to effect a step change in the reactor pressure from 1300 kPa to 1200 kPa: ( )

where , with being the pressure set-point and the error of the actual pressure from the pressure set-point. Furthermore, is the steady-state valve position. In order to assess the controller performance, the integral square error (ISE) was determined:

DBPS project

Alastair Wong and Iniobong Akpan

15/01/2012

The constants and are the controller parameters of the PI controller, and the determination and tuning of these parameters to minimise the ISE constitutes the problem in this part of the project.

4.1. Method
First of all, in order to set up the model for this problem, the above equations and variables are added to the model which was used in part (2) of this project. In addition, the steady-state valve position and the pressure set-point were added to the list of parameters. Furthermore, since the constants and were to be determined, they were placed in the variables section in this problem and were then determined later through optimization. In terms of the process simulation entity, the value of the valve constant was once again set as . In addition, the pressure set-point was set as 1200 and the steady-state valve position as 0.5. Since the constants and were to be optimized later, initial guesses were assigned as -0.01 and 0.1, respectively. Note that the value of is negative, since a positive change in the manipulated variable (the opening of the valve) causes a negative change in the controlled variable (the reactor pressure). Furthermore, the initial conditions were as follows:
Table 9: Initial conditions used for the process simulation.

Variables

ISE

Values 0.0014240555 kmol/h 4.179494x10-5 kmol/h 1300 kPa 0 0

Notice that the initial values of the molar hold up of species 1 and 2 were obtained from results in part (2). After setting up the optimizer to minimize the ISE value, setting the time horizon to 30 and control type as time-invariant and allowable values as continuous, optimum values of the constants and were found to be -3.0586x10-2 and 4.3334x101, respectively. For details of how the model, process simulation, and the PI controller were set up, the reader is referred to the entities Models gas_phase_CSTR_question4, Processes gas_phase_CSTR_question4, and PI_controller_question4.

5. Estimation of Kr
In order to validate the kinetic model, an experiment was performed corresponding to the conditions described in part (3) of the project. The partial pressures of carbon monoxide and methanol were measured at intervals of approximately 1h over a period of about 50h. The measurements obtained (not shown) were used in re-estimating the value of , in which the current available value was found to be particularly unreliable and could, in fact, be wrong by up to an order of magnitude. 7

DBPS project

Alastair Wong and Iniobong Akpan

15/01/2012

5.1. Method
First of all, the model developed for part (3) of this project was modified for this parameter estimation; the reaction kinetic was moved to the variable section. Subsequently, it was assigned -4 an initial guess of 1.434x10 in the process simulation entity. On the other hand, in the parameter estimation entity, the control variables for the experiments were the inlet flow rates, which were given piece-wise constant control. On top of that, the upper and lower bounds of were set at -3 -5 1.434x10 and 1.434x10 , as the error may be of an order of magnitude. It was assumed that the standard deviation of the experiment was 5% and so the variance model was set as constant relative variance with a value of 0.05.

5.2. Results analysis

The table below shows the results of the model parameter estimation. It must be noted that: Probability of parameter lying between (Final Value -% Confidence Interval) and (Final Value +% Confidence Interval) = % The t-value shows the percentage accuracy of the estimated parameters, with respect to the 95% confidence intervals.
Table 10: Results of the model parameter estimation of part (5) of project.

Model Parameter Reactor. Kr

Final Value

Initial Guess

Lower Upper Bound Bound

Confidence Interval 90% 95% 99% 6.47210005

95% tvalue 7.701

Standard Deviation 2.4510005

0.00037584 0.0001434

4.07910- 4.8810005 005

Reference t-value (95%): 1.6646 Hence, the final value of is 0.00037584. In comparison to the estimated parameter value, the

confidence intervals and standard deviation are smaller and the 95% t-value is larger. Hence, the final value generated from this estimation seems to be more accurate than the formerly given value. However, the difference between the 95% t-value and the 95% reference t-value is not large, so the generated value is not sufficiently accurate enough.

5.3. Lack of fit test

The table below shows the results of a lack of fit test carried out on the experimental data and results:

DBPS project

Alastair Wong and Iniobong Akpan

15/01/2012

Table 11: Results of the lack of fit test.

Weighted Residual X2-Value (95%) 98.658 *

Comment

98.484 Model may be inadequate representation of physical system

From the table, it can be seen that the weighted residual is larger than the X2-value, indicating that the model may be inadequate representation of physical system. For details of how the model, process simulation, and parameter estimation were set up, the reader is referred to the entities Models gas_phase_CSTR_question5, Processes Simulate_CSTR_question5, and Parameter Estimations Parameter_Estimations_question5 in the file CSTR Model.

6. Estimation of Kr, KCO, KH2, and KCH3OH

For this part of the project, it was believed that the value of was already known with sufficient accuracy from independent thermodynamic measurements. However, the values of , , and were less reliable, probably varying by up to 50% from those mentioned in part (1) of the project. This in turn affects the value of . Hence, this part of the project involved re-estimating the values of , , , and .

6.1. Method
First of all, using the previous experimental data, these values were hence estimated. The model developed for part (3) of this project was modified for this parameter estimation; the reaction kinetic parameters , , , and were moved to the variable section and was assigned -4 -2 initial guesses of 1.434x10 , 4.111x10 , 1.227x10-2, and 4.367x10-2, respectively, in the assign section of the process simulation entity for this part. Since the estimated values of the above kinetic parameters may differ from those previously used by 50%, the upper and lower bounds of the parameters were assumed to be 50% and 150% of the initial guesses. These calculated values are shown in the table below:
Table 12: Initial guesses and lower and upper bounds of the parameters to be estimated.

Parameter to be Estimated Reactor.KCH3OH Reactor.KCO Reactor.KH2 Reactor.Kr

Initial Guess

Lower Bound

Upper bound

0.04367 0.04111 0.01227 1.434E-4

0.021835 0.020555 0.006135 1.434E-5

0.065505 0.061665 0.018405 0.001434

DBPS project

Alastair Wong and Iniobong Akpan

15/01/2012

6.2. Results analysis

Similar to that in part (5) of this project, the table below shows the results of the model parameter estimation.
Table 13: Results of the model parameter estimation for part (6) of project.

Model Paramete r Reactor. KCH3OH Reactor. KCO Reactor. KH2 Reactor. Kr

Final Value

Initial Guess

Lower Bound

Upper Bound 0.06550 5 0.06166 5 0.01840 5

Confidence Interval 90% 95% 99%

95% t-value 0.005081 * * 0.005071 * * 0.004924 * *

Standard Deviatio n 4.029

0.040793

0.04367 0.021835

6.711 8.029 10.65

0.022264

0.04111 0.020555

3.67

4.39 5.825

2.203

0.0069093

0.01227 0.006135

1.173 1.403 1.862

0.7042

0.0002810 0.000143 1.43410- 0.00143 0.134 0.161 0.213 0.001743 * 6 4

5

0.0809

Reference t-value (95%): 1.6655

It can be seen from the table the results of estimation of , , , and , shown to be 0.00028106, 0.022264, 0.0069093, and 0.040793, respectively. However, the 95% t-values are all smaller than the reference t-value as stated in the table. This was probably due to the fact that too many parameters were under estimation for the small amount of data, as a result leading to inaccurate parameter estimation.

6.3. Correlation matrix

The table below shows the correlation matrix which was calculated from the variance-covariance matrix:

10

DBPS project
Table 14: Correlation matrix.

Alastair Wong and Iniobong Akpan

15/01/2012

Correlation Matrix Parameter Reactor.KCH3OH Reactor.KCO Reactor.KH2 Reactor.Kr No. 1 2 3 4 1 1 1* 0.998* 1* 1 0.999* 1* 1 0.998* 1 2 3 4

From the table, the values in the off-diagonal are close to 1, indicating strong correlations between any of two of the parameters , , , and . However, in an ideal case these values should be closed to 0.

6.4. Lack of fit test

The lack of fit test results and conclusion is shown in the table below:
Table 15: Results of the lack of fit test.

Weighted Residual 93.567

X2-Value (95%)

Comment

295.081 Good fit: weighted residual less than X Value

Since the weighted residual is lower than the X2-value, there is good fit between the model and the representation of the physical system. For details of how the model, process simulation, and parameter estimation were set up, the reader is referred to the entities Models gas_phase_CSTR_question6, Processes Simulate_CSTR_question6, and Parameter Estimations Parameter_Estimations_question6 in the file CSTR Model.

11