Lecture #1: Simple Units and Flash

CHE 520 Processing Modelling and Simulation

NM Abukhdeir
Associate Professor

Department of Chemical Engineering

University of Waterloo
Waterloo, Ontario, Canada

Version 0.1 ©2024 NM Abukhdeir

1 / 38
Lecture Overview

▶ Introduction to the ethylene→ethanol process (BGW Chs. 1-2).

▶ Linear mass balance models for simple unit operations (BGW Chs.
3.2.1).
▶ Mixers
▶ Splitters
▶ Chemical reactors (?!)
▶ Linear mass balance models for flash units (BGW Chs. 3.2.2), the
building block for equilibrium staged operations.

NOTE 1: BGW is an abbreviation for the course textbook

LT Biegler, IE Grossmann, and AW Westerberg. “Systematic methods for chemical process
design.” (1997).
NOTE 2: I organize lectures based on the relation/dependence of course content and not the
individual lecture duration.

2 / 38
“Clean-sheet” Process Design

While the textbook has a more detailed treatment of the comprehensive

design of this process (see Ch. 1), time constraints, scope, and my
experience in professional practice make covering this prohibitive.

Instead, I will summarize some key details which both motivate this
process and the specifics of its implementation.

3 / 38
Motivation

While ethylene (CH2 CH2 ) is a useful precursor to many industrial

products, it is potentially not as profitable a product as those derived
chemicals, such as polyethylene or ethanol.

In our problem ethylene is being produced in excess for the existing

process demands. Thus the decision to be made is whether to:
▶ Sell excess ethylene as a low-value product (≈ $10−1 kg−1 ),
▶ Invest capital and engineering skill into using it to produce ethanol
(≈ $10kg−1 )

This is further motivated by the fact that our ethylene stream is actually
only 96%mole ethylene (EL), with impurities including 3%mole propylene
(PL) and 1%mole methane (M), since it is a byproduct of an existing
upstream process.

4 / 38
Motivation

Methane and propylene are difficult to separate from ethylene (why?),

resulting in a feed of impure ethylene which is even lower value than pure,
without this separation being performed.

Your supervisor, who has years of process design experience, with the
resulting knowledge base, insight, and intuition, has found that the most
likely chemical process that will be profitable is to produce 190 proof ethyl
alcohol (why?).

Apparently, this product is in high demand due to rampant global alcohol

abuse12 . Is this decision compatible with our professional ethics?

1
“Worldwide, 3 million deaths every year result from harmful use of alcohol. This
represents 5.3% of all deaths.” -World Health Organization
2
“In people aged 20–39 years, approximately 13.5% of total deaths are attributable
to alcohol.” -World Health Organization
5 / 38
Process Chemistry

CH2 −CH2 + H2 O −−→ CH3 CH2 OH R1

2 CH3 CH2 OH ←−→ C2 H5 −O−C2 H5 + H2 O R2

CH2 – CH2 , ethylene (EL, feed)

H2 O, water (W, feed)
CH3 CH2 OH, ethanol or ethyl alcohol (EA, feed)
C2 H5 – O – C2 H5 , diethyl ether (DEE, waste)

These reactions, particularly the desirable R1, occur at high temperatures

(535 − 575K) and pressures (≈ 6.9MPa = 68atm) without the need for
catalysts.

However, under these conditions reaction conversion is low 5 − 7% but

occurs under a broad range of water-to-ethelyne stoichiometric ratios
4 : 1 → 0.6 : 1.
6 / 38
How can we improve net conversion in this case?

7 / 38
How can we improve net conversion in this case?

A recycle stream is desirable to improve net conversion.

7 / 38
How can we improve net conversion in this case?

A recycle stream is desirable to improve net conversion.

What are the drawbacks to introducing a recycle stream?

7 / 38
How can we improve net conversion in this case?

A recycle stream is desirable to improve net conversion.

What are the drawbacks to introducing a recycle stream?

Larger process flows, equipment, and costs.

7 / 38
How can we improve net conversion in this case?

A recycle stream is desirable to improve net conversion.

What are the drawbacks to introducing a recycle stream?

Larger process flows, equipment, and costs.

What water-to-ethelyne ratio is desirable and why?

7 / 38
How can we improve net conversion in this case?

A recycle stream is desirable to improve net conversion.

What are the drawbacks to introducing a recycle stream?

Larger process flows, equipment, and costs.

What water-to-ethelyne ratio is desirable and why?

The lowest to achieve the desired conversion, so 0.6 : 1, in order to

minimize process stream size and the consequences of that.

7 / 38
Process Chemistry

The second reaction is undesirable, thus maintaining a concentration of

DEE in the inlet stream to the chemical reactor is desirable (why?).

8 / 38
Process Chemistry

The second reaction is undesirable, thus maintaining a concentration of

DEE in the inlet stream to the chemical reactor is desirable (why?).

As long as we impose constraints so that this inlet concentration imposes

reaction equilibrium, we prevent the reaction of EA product to DEE waste,
further motivating a recycle stream (why?).

Our chemistry group advises us to limit the methane fraction (by mole) in
the feed to < 10% to prevent coking, which is possible at the high
temperatures and pressures at which the process operates.

They also advise that this reaction mechanism will produce a small
amount of a four carbon aldehyde, croton aldehyde
(CH3 CHE – CHCH – O), a waste product.

8 / 38
Process Chemistry
The propylene (PL) present in the feed will also react with water to
produce iso-propanol (CH3 CHOHCH3 ) with a conversion of 0.5 − 0.7%
under the given conditions.

9 / 38
Process Chemistry
The propylene (PL) present in the feed will also react with water to
produce iso-propanol (CH3 CHOHCH3 ) with a conversion of 0.5 − 0.7%
under the given conditions.

In summary, the chemistry associated with the feeds results in the

following product, byproducts, recycle, and waste:

9 / 38
Physical Properties

Note that the equilibrium vapor pressure or vapor pressure may be

approximated using the semi-empirical Antoine equation,
B
log10 P sat = A −


C +T
with P sat = mmHG and T = C for this particular set of coefficients.
10 / 38
Physical Properties

Be careful with these units in that Antoine coefficients are available for a
range of pressure and temperature units, using either log10 or ln!

11 / 38
Physical Properties

Be careful with these units in that Antoine coefficients are available for a
range of pressure and temperature units, using either log10 or ln!

You have been warned!

11 / 38
Process Topology

As you can see, M+EL+PL all have

very low boiling points compared to
the other species, with M+EL having
critical temperatures below ambient
temperature, making them
uncondensable given cooling water is
available at ≈ 310K (≈ 37◦ C).

Your supervisor synthesized the a

process topology (right) based on
this and other observations.

12 / 38
Process Design Paradigms

This is not a complete process flowsheet, it does not include compressors,

heat exchangers, pumps, and other necessary unit operations, but this is
why you are employed...

Before we start the difficult work of developing a model, mass balances,

determining temperature/pressure levels for separations, etc., let us
discuss our supervisor’s design choices.

13 / 38
Process Design Paradigms

This is not a complete process flowsheet, it does not include compressors,

heat exchangers, pumps, and other necessary unit operations, but this is
why you are employed...

Before we start the difficult work of developing a model, mass balances,

determining temperature/pressure levels for separations, etc., let us
discuss our supervisor’s design choices.

Recycle. This is clearly required to reach a net conversion of 100% of the

EL in the feed and to maintain DEE in the feed to avoid further reaction
into it.

13 / 38
Process Design Paradigms

Separations. The process species can be classified as:

Condensible: DEE, EA, IPA, W, CA
Non-condensible: M, EL, PL
based on their boiling points.

Units 3-4 (flash and gas/liquid absorber) are clearly present to separate
those two sets of species from eachother.

14 / 38
Process Design Paradigms

Separations. The process species can be classified as:

Condensible: DEE, EA, IPA, W, CA
Non-condensible: M, EL, PL
based on their boiling points.

Units 3-4 (flash and gas/liquid absorber) are clearly present to separate
those two sets of species from eachother.

Focusing on non-condensibles, a purge was added to remove M+PL from

the process, at the cost of wasted EL feed (why?).

14 / 38
Process Design Paradigms

Focusing on condensibles, our product EA has a boiling point above DEE,

but below W+CA, thus we introduce the first column to remove the
majority of the water, with most recycled but a small amount purged
(why?).

15 / 38
Process Design Paradigms

Focusing on condensibles, our product EA has a boiling point above DEE,

but below W+CA, thus we introduce the first column to remove the
majority of the water, with most recycled but a small amount purged
(why?).

The second column separates DEE which is recycled (why?).

15 / 38
Process Design Paradigms

Focusing on condensibles, our product EA has a boiling point above DEE,

but below W+CA, thus we introduce the first column to remove the
majority of the water, with most recycled but a small amount purged
(why?).

The second column separates DEE which is recycled (why?).

Finally, the last column “finishes” the EA+W (190 proof, 95% EA by
volume) by removing IPA and W, with the product EA+W mixture having
a composition very close to the azeotrope (why?), where the compositions
of the vapor and liquid phases are equal (at equilibrium).

15 / 38
Process Design Paradigm

We now have some basic insight into our supervisor’s design paradigms,
however, some details about why this concept was chosen over some
others is missing:
1. separate M+PL from impure feed before reaction,
2. use membrane separations versus distillation and purge,
3. use a catalyst to reduce the temperature/pressure requirement, avoid
compressors.

16 / 38
Process Design Paradigm

We now have some basic insight into our supervisor’s design paradigms,
however, some details about why this concept was chosen over some
others is missing:
1. separate M+PL from impure feed before reaction,
2. use membrane separations versus distillation and purge,
3. use a catalyst to reduce the temperature/pressure requirement, avoid
compressors.

However, mass and energy balances must be formulated, which is not a

simple task, but needs to be done to inform sizing and rigorous economic
analysis (for a final decision).

16 / 38
Motivation

In this first part of the course content, we are interested in process model
formulations which are simple (linear, why?) and fast (hand or
spreadsheet calculations, why?), prior to devoting time and effort to
developing a rigorous (nonlinear) process model.

To achieve our goals, we will assume ideal mixtures along with process
streams being saturated vapor or liquid (why?). These assumptions
enable us to use simple equilibrium thermodynamic relations to develop
linear models for simple-to-complex unit operations (distillation!).

We will focus first on developing mass balances, then determining

temperature/pressure levels for streams and units so that we may finally
formulate energy balances to determine heating/cooling duties,
compressure duties, etc.

17 / 38
Simple Unit Operations – Mixers

We will begin with simple units: mixers, splitters, and (fixed-conversion)

chemical reactors. A mixer unit simply sums all of the input streams into a
single output stream, which requires the use of some reasonably
“interesting” notation.

µkij ≡ the molar flowrate of component k of the j th output3 of unit i.

X
µkM = µkM1 = µkij
i

3
If the unit has a single output j is surpressed, µki
18 / 38
Simple Unit Operations – Splitters
A splitter unit simply divides a given feed into specified fractions, ξj for
each output stream j, where all output streams have the same
composition of the input stream

µkSj = ξj µkIN for j = 1 → NS − 1 µkSj = ξj µkIN for j = 1 → NS

OR
S −1
NX NS
!
X
µkSNS = 1 − ξj µkIN ξj = 1
1 1

19 / 38
Simple Unit Operations – Fixed-conversion Reactor
A reactor is by no means a simple unit operation, but a fixed-conversion
approximation is both simple and quite accurate in some cases (which
ones?).

Given NR parallel reactions labelled by index r with fixed-conversions ηr ,

we formulate normalized stoichiometric cooefficients w.r.t the limiting
reactant/component kr ,
Crk
γrk =
Crkr
where Crk are the stoichiometric coefficients of component k in reaction r .

20 / 38
Simple Unit Operations – Fixed-conversion Reactor
A reactor is by no means a simple unit operation, but a fixed-conversion
approximation is both simple and quite accurate in some cases (which
ones?).

Given NR parallel reactions labelled by index r with fixed-conversions ηr ,

we formulate normalized stoichiometric cooefficients w.r.t the limiting
reactant/component kr ,
Crk
γrk =
Crkr
where Crk are the stoichiometric coefficients of component k in reaction r .
What is Crk for components that are inert?

20 / 38
Simple Unit Operations – Fixed-conversion Reactor

Noting that the normalized stoichiometric coefficients are classified,


 > 0, product k
γrk = < 0, reactant k
0, inert k


The fixed conversion reactor model is,

X
µkR = µkIN + γrk ηr µkINr
r

21 / 38
Simple Unit Operations – Fixed-conversion Reactor

Noting that the normalized stoichiometric coefficients are classified,


 > 0, product k
γrk = < 0, reactant k
0, inert k


The fixed conversion reactor model is,

X
µkR = µkIN + γrk ηr µkINr
r

µkIN = µkR ?
P P
Does k k

21 / 38
Simple Unit Operations – Fixed-conversion Reactor

Noting that the normalized stoichiometric coefficients are classified,


 > 0, product k
γrk = < 0, reactant k
0, inert k


The fixed conversion reactor model is,

X
µkR = µkIN + γrk ηr µkINr
r

µkIN = µkR ? No! Moles are not conserved!

P P
Does k k

21 / 38
BGW Example 3.1 – Fixed-conversion Reactor

22 / 38
Motivation

We will spend quite a bit of time on flash units due to the importance of
vapor-liquid equilibrium calculations for:
▶ flash and condensing separations
▶ equilibrium staged separations including:
▶ distillation
▶ absorption
▶ stripping

These calculations have a significant effect on the accuracy of our process

models. However, competing with accuracy is the complexity of solving
the resulting model.

23 / 38
In this part of the course, we will bias towards simplicity of solution
through poor accuracy. Two schematics of the liquid vapor flash unit are,

Ncomp ≡ number of components

n ≡ key component
ϕ = VF ≡ vaporization (fraction)
ξk = vfkk ≡ overhead split fraction for component k
k ′ ≡ major component (by mole) in liquid phase

24 / 38
Flash Model
Why can the n key component and k ′ major component in the
liquid phase potentially be different?

25 / 38
Flash Model
Why can the n key component and k ′ major component in the
liquid phase potentially be different?

The flash model (below) is well-specified with ξn , ϕ, P, T , and Q.

However, these are not all free variables.

fk = lk + vk , for Ncomp
yk = Kk |P,T xk , for Ncomp
X X
lk = L and vk = V
k k

25 / 38
Flash Model
Why can the n key component and k ′ major component in the
liquid phase potentially be different?

The flash model (below) is well-specified with ξn , ϕ, P, T , and Q.

However, these are not all free variables.

fk = lk + vk , for Ncomp
yk = Kk |P,T xk , for Ncomp
X X
lk = L and vk = V
k k

Note that the second equation above can be conveniently reformulated,

vk lk
= Kk |P,T , for Ncomp
V L
25 / 38
Flash Model

Given that the feed is specified and Kk |P,T is constrained by our

thermodynamic model, it can be shown that we have two less equations
than unknowns.

Neglecting Q as a variable until we address the conservation of energy

equation, we may choose two of the following: ξn , ϕ, P, and T .

26 / 38
Flash Model

Given that the feed is specified and Kk |P,T is constrained by our

thermodynamic model, it can be shown that we have two less equations
than unknowns.

Neglecting Q as a variable until we address the conservation of energy

equation, we may choose two of the following: ξn , ϕ, P, and T .

The three main cases for flash specifications are as followed:

Case 1: Specify ξn (key component overhead recovery) and
T or P.
Case 2: T and P specified (isothermal flash)
Case 3: ϕ specified and T or P

26 / 38
What is the significance of each case?

27 / 38
 What is the significance of each case?

Which do you think will be most useful in developing our linear

process models?

27 / 38
 What is the significance of each case?

Which do you think will be most useful in developing our linear

process models?

27 / 38
Closing the Flash Model

In our quest for linear models, we will use the assumptions of ideal
mixtures for gases/liquids and Dalton’s Law,
X
P= Pk
k
Pk
yk =
P
where Pk is the partial pressure and yk is the mole fraction in the vapor
phase of component k.

28 / 38
Aside: Ideal Mixtures and Gases

Partial pressure is defined for each component in a multicomponent gas

mixture as the pressure that each component would exert on the volume
at the same temperature, in the absence of the other components.

Dalton’s Law states that the the pressure which a multicomponent gas
mixture exerts on a volume is the sum of the partial pressures of each
component.

Raoult’s Law states that partial pressure of a component in a

multicomponent gas mixture in thermodynamic equilibrium with a liquid
mixture is the product of its vapor pressure and liquid-phase mole fraction,

Pk = Pkvap xk

29 / 38
Vapor-Liquid Equilibrium Relation for Ideal Mixtures

Putting all of these ideal mixture relations together yields a simple linear
VLE relationship,

Pk Pkvap vk lk
yk = = xk → =K (1)
P P V L
Pkvap
where the Kk = P .

30 / 38
Vapor-Liquid Equilibrium Relation for Ideal Mixtures

Putting all of these ideal mixture relations together yields a simple linear
VLE relationship,

Pk Pkvap vk lk
yk = = xk → =K (1)
P P V L
Pkvap
where the Kk = P .

We may approximate the vapor pressure of each component using the

Antoine equation,
Bk
log10 Pksat = Ak −


Ck + T
as long as we are careful with units and validity ranges of the empirical
constants we use!

30 / 38
Relative Volatility
We have “closed” our flash model by finding a convenient linear
approximation for Kk |P,T , however, remember that in Case 1 both ξn and
T or P are specified.

A simple reformulation of the original model introducing relative volatility

is useful,
Kk yk /xk vk /lk
αk/n = = =
Kn yn /xn vn /ln
X Kk X 1
α= xk = αk/n xk =
Kn Kn
k k

where α is the average relative volatility.

31 / 38
Relative Volatility
We have “closed” our flash model by finding a convenient linear
approximation for Kk |P,T , however, remember that in Case 1 both ξn and
T or P are specified.

A simple reformulation of the original model introducing relative volatility

is useful,
Kk yk /xk vk /lk
αk/n = = =
Kn yn /xn vn /ln
X Kk X 1
α= xk = αk/n xk =
Kn Kn
k k

where α is the average relative volatility.

The relatively volatility of each component varies less w.r.t. temperature

than the volatility, which is frequently leveraged in hand calculations.

31 / 38
Flash Model using Feed/Split Basis

Now we need to reformulate our flash model in terms of feed composition

and split fractions (why?),

Kk vk /lk
αk/n = =
Kn vn /ln

32 / 38
Flash Model using Feed/Split Basis

Now we need to reformulate our flash model in terms of feed composition

and split fractions (why?),

Kk vk /lk
αk/n = =
Kn vn /ln

Now note the relations between the vapor and liquid outputs to the feed,
vk = ξk fk and lk = (1 − ξk ) fk

32 / 38
Flash Model using Feed/Split Basis

Now we need to reformulate our flash model in terms of feed composition

and split fractions (why?),

Kk vk /lk
αk/n = =
Kn vn /ln

Now note the relations between the vapor and liquid outputs to the feed,
vk = ξk fk and lk = (1 − ξk ) fk

Substitution yields,

vk /lk ξk fk / (1 − ξk ) fk ξk (1 − ξn )
αk/n = = =
vn /ln ξn fn / (1 − ξn ) fn ξn (1 − ξk )

32 / 38
Flash Model using Feed/Split Basis

However, we need this relation to be w.r.t. αk/n and ξn ,

ξk (1 − ξn )
αk/n =
ξn (1 − ξk )

so after simple rearranging,

(1 − ξk ) αk/n ξn
ξk = αk/n ξn → ξk =

(1 − ξn ) 1 + αk/n − 1 ξn

33 / 38
Flash Model using Feed/Split Basis

However, we need this relation to be w.r.t. αk/n and ξn ,

ξk (1 − ξn )
αk/n =
ξn (1 − ξk )

so after simple rearranging,

(1 − ξk ) αk/n ξn
ξk = αk/n ξn → ξk =

(1 − ξn ) 1 + αk/n − 1 ξn

Thus for a specified ξn we may now compute {ξk } and formulate a linear
model
 for the flash, except for the inconvenience that we need to find
αk/n given T or P.

33 / 38
Bubble and Dew Point Equations


In order to find αk/n given T or P, we must satisfy the ideal mixture
thermodynamic constraints,

Pkvap αk/n
Kk = =
P α
which we rearrange depending on if T or P is specified,
α αk/n
P= Pkvap OR Pkvap = P
αk/n α

noting that Pkvap is a function of temperature only. This equation must be

satisfied for the majority liquid phase component (by mole), and is referred
to as the bubble/dew point equation for flashing/condensing in this case.

34 / 38
Case 1 Flash: ξn and T (or P) specified
While the resulting model for the flash unit is linear, given that the split
fractions are known, to find split fractions consistent with the specified
operating condition actually requires the solution of a nonlinear equation
system!

You will develop simple numerical computing tools to solve these

equations in HW1, with the general algorithm for Case 1 is as follows:
1. For a specified ξn and P (or T ), guess T (or P).

2. Evaluate {Kn } → αk/n and find the resulting {ξk }, {(xk , yk )}, and
α for the specified P (or T ).
3. Use the bubble point equation to find the consistent T ′ (or P ′ ) and
compare to your guess, if not consistent, repeat.
▶ For fixed P, T ′ from Pkvap = αα k/n
P.
▶ For fixed T , P from P = α Pkvap .
′ α
k/n

35 / 38
Case 2 Flash: T and P specified

1. For a specified T /P, choose key component n and guess ξn .


2. Evaluate {Kn } → αk/n and find the resulting {ξk }, {(xk , yk )}, and
α for the specified P (or T ).
3. Find the consistent ξn′ as follows,

vn Kn ϕ θ
θ= = → ξn′ =
ln 1−ϕ 1+θ
and compare to your guess, if not consistent, repeat.

Note: The equivalent expressions for θ are only valid for the
consistent/correct value of ξn ! Which expression should we use?

36 / 38
Case 3 Flash: ϕ and T (or P) specified

1. For a specified ϕ = VF and P (or T ), guess T (or P).


2. Evaluate {Kn } → αk/n and ξn as follows:

Kn ϕ vn θ
θ= = → ξn =
1−ϕ ln 1+θ

3. Find the resulting {ξk }, {(xk , yk )}, and α for the specified P (or T ).
4. Use the bubble point equation to find the consistent T ′ (or P ′ ) and
compare to your guess, if not consistent, repeat.
▶ For fixed P, T ′ from Pkvap = αα k/n
P.
▶ For fixed T , P from P = α Pkvap .
′ α
k/n

37 / 38
BGW Example 3.2

Case 1: ξ2 = 0.9 and Case 2: T = 390K Case 3: ϕ = 0.8 and

P = 1bar and P = 1bar P = 1bar
38 / 38