Feature Report

Engineering Practice

Evaluating Green Projects:

Modeling Improves Economic Benefits
Allen Williams, Private Consultant
Ken Dunwoody, Retired The use of dynamic modeling can help to identify
he ability to model a process to the engineering and economic feasibility of any
T see if it meets project require-
ments and is economically viable
enhances the conceptual design.
Modeling gives engineers a means
to identify the most promising pro-
proposed project. This sample project, involving the
conversion of solid waste to synfuel, shows how
cesses, in terms of capital investment
SUPH 75,000
and return, and this can be particu- 575°F lb/h
larly important when the project is Steam
environmentally oriented (a so-called Drum CH4 in at
486°F. 600 psi 1,150°F
green project).
Methanol reaction
Did you ever consider producing a H2O
100°F, 200 – 250 psi
new product from your existing plant Reformer Separator
350 psi
300 –

inventory, or implementing a particu- Furnace

Makeup H2O
1,500 – CH3OH
lar green project that could help to re- 1,600°F
Conveyor
duce your tax liability but for which
high fuel costs ultimately killed the
overall economics of the project? DEA
Economizer
It is important to evaluate green CH4 Feed
298°F
ideas that could potentially improve 1,480 lb/h
your bottom line, but how do you sort
through a myriad of computations to FIGURE 1. Shown here is a typical process sketch for a solid-fuel-fed steam re-
screen all the options associated with former and methanol reactor. The hydrogen and carbon monoxide blend are reacted
such tasks, and evaluate the pros and to produce methanol (SUPH = super heater; DEA = deaerator)
cons of each proposed idea?
The answer is simpler than you Developing your own process The process
might think. The trick is to use a dy- model is not as daunting a task as it To begin, make a short sketch of the
namic process model that is equip- may first seem, especially if you de- selected process (such as that shown
ment-specific but still generic in the fine the modeling system and iden- in Figure 1). Identify the tempera-
way that it handles mass and energy tify what is required before tackling tures, pressures and flowrates for each
computations. Today, while there are a the process. And, your time invest- segment that is to be modeled. Such a
number of commercial models avail- ment will be paid off by increased conceptualization sketch must include
able, there are many benefits that analysis proficiency, speed and ca- key state variables, such as compo-
come from building your own real- pability. A model with greater ana- nent concentration, moisture content,
time dynamic model — one that can lytical capabilities allows users to levels of excess oxygen in the furnace
allow you to assess your process and explore more alternative scenarios, and so on.
its associated controls. in greater detail. The process model profiled here em-
Dynamic modeling platforms are Operating data and physical con- ulates a 100- to 125-million-Btu/h in-
available with most distributed con- stants define the boundary condi- dustrial combustor with feed preheat-
trol system (DCS) hardware, but tions that determine how well a model ing, a deaerator, an economizer and
in many cases, a PC-based system represents an actual process. But, a solids-conveying system. A green
model provides the greatest ver- don’t expect extraordinary or three- project idea proposes to use the sys-
satility and flexibility. Specifically, decimal-place accuracy in computa- tem to produce synthetic fuel from
building your own computer process tions — rather, such models are useful municipal waste.
allows you to customize the func- for identifying trends and direction of The process is represented by a set
tions and tailor the model to fit your changes. Most rudimentary modeling of ordinary differential equations that
site-specific application and this efforts will fall within ±10%. More- conserve mass and energy, and utilize
provides a smoother way to manage thoroughly-developed algorithms can thermodynamic relationships, reac-
project change. attain accuracy of ±1–3%. tion kinetics and authentic equipment
34 CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012
 Trend ptrs

Graphic Com ptrs

interface Ref ptrs
Control Ref ptrs
strategy
controllers Ref ptrs
Process Alarms
variables
Final control
FIGURE 2. This steam — passes from the reformer to
Module igure shows the
elements User interface a water separator. Steam condenses
with filter dynamic model
Common ‘n’
environment,
out in the separator and the remain-
Ref ptrs Methane
Process identifying dedi- ing dry gas flows from the separator
Ref ptrs Furnace output cated segments tank to a catalyzed reactor that oper-
and interconnec- ates at 100°F and 200 psi to produce
Com ptrs Gas bank Com ptrs tions within the methanol.
temps ‘User’ Interface.
The simulated Makeup water at 77°F cools the
Superheater Reformer process informa- methanol reactor and is pumped to
Com ptrs

tion moves be- the separator tank, where it is mixed

Drum Separator tween segments with hot condensate from the HYCO
Ref ptrs
Economizer reformer and sent on to the deaera-
tor. Oxygen is steam stripped from the
Deaerator
feed water in the deaerator at 298°F
New ‘C’ code
module before entering the economizer.
Stack MeOH ..... The model unit is designed to oper-
reactor fcarb=c[9];
c[4]=feedwtr; ate at 1,500°F with 18% excess air and
Module ch4=c[1]; negligible air preheating.
.....
Variable Notes:
trending ptrs = Pointers The model
Ref = Reference
memory Once you have established a good pro-
Com = Common
memory cess definition, the next step is to begin
to develop an outline of the model ‘C’
coding, adding external references and
ECONOMIZER SEGMENT variables, as needed. ‘Pseudo code’ is
/*......................................Economizer Segment....................................*/
/* Source code as it should appear in the ‘C’ editor */ preferred at this stage. Be certain that
Re = 4 * Wfw/(*(Di/12)*Mu*ntube); /* Liquid Reynold’s No */ supporting definitions and library
Pr = cph2o*Mu/Kl; /* Prandtl No. */ functions are loaded.
tsat = a+dp*(b+dp*(c+dp*(d))); /* Saturated temp at drum pres */ When the pseudo code is complete,
tavg = (tw + tdea)/2; /* Avg temp across economizer */ formalize and code the general and li-
kw = 345 -.077*tavg; /* TC data - .23 C.S. , 1972 B&W */
brary routine statements into a main
/* STEAM, p. 4-2, Fig. 1 */
Rtube = Do/12 * log(Do/Di)/(2*kw); /* Tube therm resistance */ process module. Add separate seg-
/* from 1972 STEAM p.4-9 */ ments, as needed, and name them ac-
Re = wflue/Af*(Di/12)/ugas; /* Gas Reynold’s No, STEAM, Chap 4 */ cording to their intended function (for
/* Where: wflue/Af is the bulk gas velocity */ example,the economizer segment, and
Pr = cpflue * ugas/kgas; /* Gas Prandtl No. */ so on).
Hid = 0.023*Kl/(Di/12)*(4*wfw/(*(Di/12)* uliq * ntube))^.8 *Pr^.4;
Library routines containing other
kair = (a + b*dty)/12; /*Regress fit of air thermal cond */
/* Where: ‘dty’ is the differential gas bamk temp entering the economizer */ supporting functions must be called
kgas = xrh2o* kh2o +(1-xrh2o)*kair; /* flue gas therm Cond */ with an #include statement. For in-
Ucg = a1*kgas*Re^.61*Pr^.33; /* Gas ht-trans, 1972 STEAM, Chap 4 */ stance:
U = 1/((Do/Di)* 1/Hid + Rtube + Ucg; /* Overall Heat Trans coeff */
q = wflue * cpflue *(ty - ta); /* Heat added by flue gas */ /* MAIN SEGMENT */
area = ntubes * Ltube * Ao; /* Gas outside tube area A0=ft2/ft */
dtw = tdea – tw +q/(area*U); /* T(i+1) = F(n) T(i) / F(n’) T(i) */
tw = tw + dtw; /* Integrate Economizer Outlet temp */ #include <math.h> /*
cph2o = a + b* tavg + c * tavg^2; /* Economizer ht-capacity */ math subroutine library */
q = wfw * cph2o*(tw - tdea); /* Heat added by the flue gas */ float Di, Do, Q, V, ts, tg, tsat; /* local
tavg=(tw+tdea)/2; floating pt single */
cph2o = a + b * tavg + c * tavg^2; /* Economizer ht-capacity */
q = wfw * cph2o*(tw - tdea); float Hg, Hid, Hod, Lhv, Hhv, Mu,
dta = ta - ty -q/(area * U); /* T(i+1) = F(n) T(i) / F(n’) T(i) */ Ntube; /* precision variables */
ta = ta + dta; /* New Econ Bank outlet temp
….

specifications to develop the model preheaters at 100+°F and is heated extern double c[9], w[9]; /*
profile. In our example, solid fuel pel- successively in two stages from global double precision */
lets are conveyed from a feed bin onto 950°F to 1,150°F. The hot methane (Note: All terms are defined in the
a furnace-traveling grate. Calcium feed enters a HYCO reformer where Nomenclature box on p. 37.)
carbonate is injected above the com- 600-psi steam is added. The HYCO Entering comments or reminders
bustion area as an SO2 absorbent. gas — which is comprised of carbon into the code language for future ref-
Methane feed enters the first of two monoxide, hydrogen, and unreacted erence is highly recommended. For
CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012 35
 STEAM SEGMENT
/*....................... Steam Drum segment........................ */
Hg =a + dp *(b + dp *(-c + dp *(d)); */ Stm enth @drum pres curve fit */
Ub = 4000/(2.515*dtsat^.33)); /* Peters & Timmerhaus, 2nd edition */
dw =(Ub*wfw/Rf)/(Hv*b1-2*c2*tb)*dtsat/(tsat-tw); /* diff lb/hr evap */
tavg =(tsat + tw)/2; /* Avg mud drum and Sat liq temp */
Kw = 345-.1*tavg; /* .23 CS, STEAM - p. 4-2, Fig 1 */
instance, comments can
area = ntubes*Ltube*Ao; /* Gas outside tube area Ao=ft2/ft */
be added directly into the R = dx/(kw*Alm); /* Perrys 4th Ed, Tube Chart - Table 11-2 */
computer code as shown Uo = Ub +1/R; . /* Overall Boiling Ht Transfer Coefficient */
above, according to the wstm = wstm + dw; /* Steam mass evaporated */
software vendor’s pro- q = wstm * (Hg - Hf); /* where Hg - Hf is the ht of vaporization */
gramming convention. qfr = wflue * cpflue *(tf - t0); dt0 = tf - t0 - (q + qfr)/(area*Uo); /* T(i+1) = F(n) T(i) / F(n’) T(i) */
Double-precision vari- t0 = t0 + dt0; /* New water drum gas bank outlet temp */
ables should only be used
when a floating-point ac-
curacy of greater than SUPH STEAM SEGMENT
8 bits is required — for /*....................... Super heater segment ....................... */
instance, in the calcula- Rtube = dx/(kw*Ao); /* Perry’s 4th Ed, Tube Dia, Table 11-2 */
tion of heater-tube fouling Re = wstm/Af *(Di/12)/uvap; /* Steam Reynolds No. */
coefficients that involve Pr = cpstm * uvap /kv; /* Steam Prandtl No. */
very low numbers (that is, Hid=0.023 * RE^.8 * Pr^.4)/(Di/12)^.2; /* STEAM, 1972 - Fig 6, p. 4-8 */
RE = wflue/Af *(Di/12)/ugas; /* Gas Reynold’s No */
less than 1E-04).
Pr = cpflue * ugas/kgas;
Calculated heat and Ucg = a2 *kgas*pow(RE,0.61)*pow(Pr,0.33)*Fa; /* 1972 STEAM, Chap 4 */
mass transfer data should U = 1/Rtube + 1/Hid + Ucg; /* Overall Heat Trans coeff */
reside in common mem- q = wflue * cpflue *(t0 - ts); /* Gas bank energy input */
ory. State variables that dtw2 = tb – tw2 +q/(area*U); /* T(i+1) = F(n) T(i) / F(n’) T(i) */
are to be used as process tw2 = tw2 + dtw2; /* Integrate Suph Outlet temp */
inputs must be linked to cpstm = a + b * tavg + c * tavg^2; /* Integrate Steam Ht capacity */
the controller-measured /* where a, b, c are dimensionless ht capacity coefficients */
variable.
Whatever dynamic
modeling platform you FURNACE SEGMENT
select should support this /*....................... Furnace segment ....................... */
approach. The ability to Hhv = 10221; /* Waste-Blend heating value */
make coding modifica- xc = 0.5297; /* fuel carbon content */
tions and additions within xh2 = 0.1048; /* fuel hydrogen content */
a model are greatly fa- xsul = 0.0238; /* fuel sulfur content */
cilitated when you do not xo2 = 0.1300; /* fuel oxygen content */
xn2 = .0017; /* fuel nitrogen content */
have to re-allocate vari-
xrh2o = 0.1467; /* wt frac of H2o in the waste fuel */
ables. Figure 2 shows how fuel = S ft/hr * w ft2 wide * d lbs/ft3 /* Solid waste feed - lb/hr */
the code segments are Lhv = hhv -(1040*8.9399*2*xh2o+18*xh2o); /* Lower Heating Value */
configured to pass infor-
mation to one another.
manipulation. For example, heats of
Building what you need reaction can be found by adding and
Once you have identified a green proj- subtracting chemical equations from
ect to evaluate, pick the most com- standard texts until suitable heat
mon service elements. In the example data are obtained.
profiled here, these service elements Equilibrium rates for combustion
are the furnace, reformer and reactor. reactions that generate emissions
Other components can be developed may be estimated from the catalyst
and added later, as required. manufacturer’s Kf formation data. For
Process data should be obtained example, in the case of predicting NOx
from plant operations to develop the emissions, Equation (1) can be used The computed Kp values may be loga-
physical characteristics of the plant to calculate the heats of reaction for rithmically plotted against tempera-
equipment being modeled. Profes- nitrogen dioxide by manipulating the ture to give the reaction equilibrium
sional handbooks and equipment heat of formation data to determine constant, K, over the anticipated oper-
specification sheets are good sources the combustion equilibrium constant ating conditions.
to provide representative data related Kp. (This approach lets users get the
to materials of construction, sizes, required rate data when exact infor- Arranging code segments
flowrates, and occasionally tempera- mation is not shown): Identify the process control streams
ture and pressure. that drive the model calculations. In
When you cannot locate the (1) this example, they are the solid waste
exact information desired, relevant The reaction equilibrium constant, Kp, fuel, feed water and methane flows.
numbers can often be generated by is defined below: Define the process segments, how
36 CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012
 NOMENCLATURE
A, Af, ALM tube areas, ft2 Hv liquid heat of vaporization, tavg average temperature across
Ao tube outside area, ft2/linear ft Btu/lb economizer, (inlet + outlet)/2
CaO carbon monoxide initial kair thermal conductivity of combus- tb saturated liquid boiling temp
concentration, lb-moles/ft3 tion air, Btu/h-ft-°F tdea liquid temperature in the
CH4 methane feed to the HYCO kgas thermal conductivity of fluegas, deaerator
reactor Btu/h-ft-°F tf fuel flame temperature
Cpco heat capacity of carbon kh2O thermal conductivity of the feed tj, dtj methanol jacket outlet temp and
monoxide product water, Btu/h-ft-°F temperature differential
CpH2 ht capacity of hydrogen product K, K(1), K(2), Kp equilibrium constants, t0, dt0 water drum gas-bank outlet
Cpflue heat capacity of the flue gas dimensionless temperature and temperature
Cpmeoh heat capacity of methanol product kl liquid thermal conductivity, differential
Cpstm heat capacity of steam Btu/h- ft2-°F/ft tsat, dtsat saturated steam temper-
CV valve flow coefficient, kw thermal conductivity for carbon ture and temperature differential
dimensionless steel, Btu/h- ft2-°F/ft tw, dtw economizer liquid outlet tempera-
DEA deaerator kv steam thermal conductivity, ture and temperature differential
Di inside tube diameter, in. Btu/h- ft-°F tw2, dtw2 steam superheater outlet
Do outside tube diameter, in. Ltube length of tube, ft temperature and temperature
dhwgs heat of HYCO gas reaction Lhv fuel low heating value, Btu/lb differential
corrected to reaction conditions Mu liquid viscosity, centipoise ts superheater gas-bank inlet tem-
fco carbon monoxide feed to ntube number of tubes perature
methanol reactor Opn model methane valve percent ty, dty economizer gas-bank inlet
feedwtr feed water flow to the deaerator open, % temperature and temperature
Fa tube bank arrangement factor, differential
P, dp drum pressure and pressure
dimensionless U, Ub, Ucg, Uo Unit heat transfer
differential
Fao carbon monoxide initial feedrate, coefficients, Btu/h-ft2-°F
PIDout model controller output, %
lb-moles/h ugas fluegas viscosity, centipoise
Pr Prandtl number, dimensionless
Fcarb steam to carbon ratio, dimension-
Q heat added to liquid, Btu/h uvap steam viscosity, centipoise
less
q heat lost by fluegas, Btu/h V reactor volume, ft3
Fm methanol produced in reactor,
qfr heat generated in furnace wflue fluegas mass flow
lb/h
combustion, Btu/h wfw feedwater flow from the deaera-
Hg saturated vapor enthalpy, Btu/lb
R[dx/KALM], Rtube thermal resistance to tor to the economizer
Hf saturated liquid enthalpy, Btu/lb
Hhv fuel high heating value, Btu/lb heat transfer, hr-°F/Btu wstm, dw steam generation flow and
Hid tube inside heat transfer coeffi- Re Reynolds number, dimensionless steam mass differential
cient, btu/h-ft2°F Rf saturated liquid specific volume, xc fuel carbon content, wt. %
Hod tube outside heat transfer coef- ft3/lb xh2 fuel hydrogen content, wt. %
ficient, btu/h-ft2°F rxn HYCO gas reaction rate, h-1 xo2 fuel oxygen content, wt. %
hrxnp Heat of reaction of HYCO gas stdh standard heat of reaction xn2 fuel nitrogen content, wt. %
products ta, dta economizer gas-bank outlet tem- x, xm unit conversions, mole %
hrxnr Heat of reaction of HYCO gas perature and temperature xrh2o fuel moisture content, wt. %
reactants differential xsul fuel sulfur content, wt. %

Note: All simulation temperatures are expressed in °F, pressures are in psi and liquid flow is in lb/h. Heat capacity is expressed in
Btu/lb-°F, and heats of reaction are shown in Btu/lb-mole units.

they will be structured and how they of heat energy, in the data passed, perheating components — may find it
will interact in the simulation. Figure as the temperature of the fluegas expedient to dedicate an entire mod-
2 shows the methodology. reduces through this series of coded ule to one process function alone (that
Next, assign inlet and outlet num- segments. The magenta arrows in is, all code segments in that module
bers or letters to the furnace banks Figure 2 show the HYCO gas infor- are dedicated to a single task).
for the estimates of model gas tem- mation progression from the meth- The selected simulation user in-
perature so that model heat flow can ane preheat units to the methanol terface should be able to contain mul-
be developed in the proper direction. reactor segment. The model can be tiple user blocks with individual dis-
The smaller arrows in Figure 2 depict set up so that a particular segment’s crete segments. Multiple user blocks
information flow between the code calculated information appears in a must be able to pass information to
segments. realtime graphical display. other user blocks, just as if they were
The red arrows within the Figure segments within the same module.
2 process module represent com- Implementing the model code The modules must be linkable to
bustion gas temperature and infor- The user interface of Figure 2 is a your control strategy in the chosen en-
mation flow, while the blue arrows module made up of one or more ‘C’ vironment.
provide feed water condition infor- code segments. Segments in the exam- Once the model is structured and
mation progression from the econo- ple module, because of their relatively ready for coding, initialize all variables
mizer to the steam drum and steam small size, comprise the entire model used in that segment before develop-
to the super heater (SUPH). The user interface shown. ing detailed computations, as shown
arrow color for fluegas decreases in However, users working with larger in the furnace segment example for
intensity (from darker red to lighter furnace simulations — for instance, updating fuel composition.
red), to depict the corresponding loss those with reheat and additional su- One can develop the model for en-
CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012 37
 Model Temperature Profiles
Operating conditions: 1,500°F — 1,600°F
Engineering Practice Heat content, Btu/lb: 5,040 — 10,221
Steam temperature Furnace gas bank
Methane feed temperature temperature
2,800
Incorrect initial estimate
ergy flow in either direction. In our of second CH4 preheater
example, we move from the coolest 2,400 gas-bank outlet temperature

streams to the hottest ones.

Next, code the process variables that 2,000 HYCO
will be input to the model controllers. reactor
Begin to develop dimensional data

Temperature, °F
1,600 Second CH4
from unit hardware characteristics. preheater
Incorrect
The model should utilize the same low temperature
economizer tube
1,200
number and size tubes as the proto- heat transfer
type. The economizer segment typically coefficent –1/hi
Temperature
800
shows how tube diameter, quantity and gain from
material of construction develop the Deaerator CH3OH
Change kw reactor
heat-flow characteristics of the equip- 400
to C.S. and
ment being modeled. correct tube size
Develop the liquid heat transfer co- 0
efficients using Dittus-Bolter or other xf x0 xs xg xy xa xm amb
accepted heat transfer expressions.
Hottest Coolest
Thermal conductivity data are readily
Gas bank position
available in physical property or com-
mercial refrence texts and only need FIGURE 3. This igure is the model calculated temperature proiles for the process
to be regression fit to use. However, for reference condition. The model temperature proiles determine whether or not the
model material and energy balances are sufficiently representative of the prototypical
a smoother and quicker initial startup, process so as to be reasonably predictive. They are an essential tool for developing
develop gas bank heat transfer coeffi- and troubleshooting any model process
cients on thermal resistance alone, for
example dx/KA until operational (see due to heating the feed water. The up- TABLE 1. STEAM/CARBON RATIO
Perry’s 4th Ed., Tube Char-Table 11-2, dated economizer outlet temperature TO CONVERSION
p. 7) and then enhance the overall heat (the term ‘tw’) becomes the input to
370 lb/h 18.0% excess air
transfer coefficient, U, with additional the furnace steam drum.
relationships. Estimate the differential steam T=1,500°F 9.3 MJ/kg =
The thermoconductivity (kgas) of the mass ‘dw’ evaporated by the furnace. 4,466 Btu/lb
fluegas is dependent on the fuel mois- The overall heat transfer coefficient Steam/carbon CH4
ture content (xrh2o) and water con- for the drum must include the inside ratio conversion, %
ductivity (kh2o). tube boiling film coefficient ‘Ub’, as 1.9 96.96
Furnace gas flow, wflue, can be de- previously described, so that the fur- 2.2 97.05
veloped from combustion reactions, nace exit-gas-bank temperature var-
2.7 97.09
C + O2 —> CO2, using the composition ies with the quantity of steam gener-
data shown in Table 3. ated. This variance is then propagated 3.2 97.12
Sum the fluegas components, (such to subsequent bank temperatures to 3.7 97.15
as ‘Co2’, ‘H2o’ and so on) to get an up- develop an overall dynamic gas tem-
dated wflue based on model control- perature profile (shown by the curved drum ‘wstm’ is the input to the super
ler fuel flow. Integrate the individual black line of Figure 3). heater.
fluegas heat capacities, at gas bank The overall heat transfer coefficient Thermal conductivity of the super-
temperature ty (to get the new cpflue) for this case is Uo = Ub + 1/R. heater steam tubes is the same carbon
before calculating the gas heat loss at Calculate the heat required for the steel material ‘kw’ as in the econo-
the economizer. steam vaporization: mizer tubing wall resistance, ‘Rtube’.
Heat transfer area is a product Continue updating other flue gas
q = wstm * (Hg - Hf)
of the number of tubes, length and temperatures for energy lost due to
the external square feet of area per Calculate the heat produced in the heat transfer to the process, using
foot of tubing. furnace combustion, ‘qfr’ where ‘tf’ is the proper gas-bank number or let-
Derive the necessary mathematical the flame temperature. ter codes, as was done for the outlet
expressions and formulate the differ- Derive the mud drum gas bank out- temperature of the economizer gas
ential temperature relationships. To let temperature expression, ‘dt0’ and bank.
calculate the heat gained, integrate formulate the differential tempera-
up the liquid heat capacity based ture relationship: Kinetics
on the average temperature of Kinetic data for equilibrium reaction
T(i+1) = F(n) T(i) / F(n´) T(i)
the economizer inlet and the new out- rates can be obtained from catalyst
let temperature. As developed in the economizer manufacturers and from thermody-
Update the economizer outlet-gas- segment. namic texts. There are two competing
bank temperature for the energy lost The steam evaporated from the reactions in the HYCO process:
38 CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012
 REACTOR SEGMENT
/*....................... HYCO Reactor segment ....................... */

K(1) = Exp(log(a)+ dt * log(c)); /* Log plot, Ka = k1/k2 */

K(2) = a - b * tout; /* Regress fit of catalyst data, Kb = k3/k4 */

ufacturer data, or from kinetic
V = ntubes * Ltube * Area; /* ft3 HYCO Reactor Volume */ expressions, if real data are not
rxn =a*(K*V*CaO)/FaO*c2 - log(b*c); /*(a,b,c from 2nd Order kinetic Rx */ accessible.Gas temperature at
Hrxnp =(fco*cpco+fh2*cph2)*1.8*(tout-tref); /* Prod x 1.8 = Btu/lbmole */ the reformer and the HYCO
Hrxnr =(fco*cpco+fh20*cpstm)*1.8*(tref-tout); /* React x 1.8 = Btu/lbmole*/ heat of reaction determine the
Dhwgs =stdh + hrxnp + hrxnr; /* Std ht of react, Btu/lb-mole */ reformer outlet temperature.
/* corrected to reaction temp */

/*....................... Methanol Reactor segment ....................... */ Methanol

Methanol is produced from a
float ch4, feedwtr, fcarb ..; catalyzed HYCO gas reaction
fcarb = c[9]; /* Steam to Carbon ratio */ according to:
c[4] = feedwtr; /* feed water flow */
cat
ch4 = c[1]; /* methane feed to HYCO Rx */ CO + 2H2 —> CH3OH
x = c[7]; /* methane conversion from HYCO */
Fm = xm * fco; /* lb/hr MeOH from HYCO Feed */ After the furnace and re-
/* Where xm is methane conversion */ former segments are opera-
cpco =a + b*tout-c*tout^2; /* ht capacity CO at HYCO outlet temp */
tional, add the methanol reactor
cph2 =a1 + b1*tout + c1*tout^5;
cpmeoh =a2 +b2*tj -c*tj^2; /* MeOH liq assumed at H2o Jacket temp */ to the USER module, as shown
hrxnr2 =fco* cpco + fh2 *cph2)*1.8*(tref-tj); /* React @ Rx cond */ in Figure 2, beginning with
hrxnp2 =Fm* cpmeoh *1.8*(tj-tref); /* Prod @ Rx cond */ common memory, c{*], as illus-
dhwgs = stdh2 +hrxnp2 +hrxnr2; trated in the methanol reactor
code segment.
U = 1/(dx/(kw*Ao)); /* Perry CH.E, 4th Ed, tube Table 11-2 */ Equilibrium rate constants
and heat of reaction are available
dtj = (rxn * dhwgs)/(U*area); /* T(i+1) = F(n) T(i) / F(n’) T(i) */ from catalyst manufacturers and
were curve fit for this example.
tj = tj + dtj; /* H2O temp to the separation tank */
The overall heat transfer coef-
ficient is similar to that of the
TABLE 2. SHREDDED TIRE HEAT CONTENT ESTIMATOR HYCO reactor, except there is no ex-
Basis Comp, mol Lb H Btu/lb-mole Moles ternal source of heat other than what
fraction the gas entering the methanol reactor
C 0.5794 60 98095.51 0.0483 contains.
The estimated methanol-reactor
H 0.1558 16.128 19153.10 0.0773
water-jacket outlet temperature is
S 0.0464 4.8 5903.80 0.0014 the product of the heat of reaction
O ‘rxn’ and the standard heat divided
by the transfer coefficient and the
N
jacket area.
H2O 0.1584 16.398 The methanol exothermic reaction
Iron 0.0601 6.2222 provides about 109.8°F–77°F=32.8°F
Total 0.1270 of temperature gain to the water sepa-
moles rator mix of Figure 1.
Total 1,000 103,55 123,152 Btu/lb-mole After the ‘C’ process model has been
written, it must be compiled and linked
7,87 lb/lb-mole
to the simulator’s graphical interface
15,640 Btu/lb according to the simulator vendor’s
procedures. Unit controls may then be
(1) CH4 + H2O —> CO + 3H2 and feed to the methanol reactor. The configured, and testing can begin.
value for ‘rxn’ should be determined The chosen vendor’s model platform
(2) CO + H2O —> CO2 + H2 from appropriate kinetic expressions. should also support logic functions,
Equilibrium reaction-rate data are Next, develop the standard HYCO such as digital input and output mod-
readily available. Log plots of the heat of reaction (‘Dhwgs’) from ther- ules for development and testing of
rate data yields the equilibrium con- modynamic data at reference condi- process and safety interlocks.
stants ‘K1’ and ‘K2’ for the compet- tions plus the reaction enthalpy of the
ing reactions. The reaction area is products minus the reactants at reac- Control blocks
the inside cross-section of the tube tion conditions. The selected dynamic simulator plat-
(ft2/ft). The reaction rate ‘rxn’ is a func- The heat of reaction, endothermic in form should permit manual or auto-
tion of the equilibrium constant, ini- this case, should be developed from a matic control with loop tuning and
tial concentration of carbon monoxide plot of dH versus T from catalyst man- choice of proportional (P), propor-
CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012 39
 Results from MSFB Results from MSFB Calculated by model
-1 x E6 Btu/h -0.4 x E06 Btu/h -115 x E06 Btu/h
Engineering Practice 0.7
Temp., °F NOx, ppm
1,500 249
0.6

Emission level ppm at 18% excess air

1,515 253
tional-integral (PI), and proportional- 400

NOx emission, lb NO2/million Btu

integral-derivative (PID), and ramp 1,525 269
0.5
controllers along with variable trend-
ing. Control blocks must have configu- 0.4 300
rable scan rate settings.
The simulated process should
0.3
react identically to the prototype 200
with proper scan settings. The rule of
0.2
thumb is 1/4 second for flow loops, 1/2
second for non-critical loops, and 1 100
second for calculation blocks. Control 0.1
loops must be faster than the simu-
lated process.
0
Configure single-loop control ini- 0 0.2 0.4 0.6 0.8 1.0 1.2
tially in your process, adding any com-
Primary air stochiometric air ratio
plex loops after the model is working
properly. Our example has only one FIGURE 4. ‘Effect of Excess Air on NOx Emissions [4]’ shows the model-predicted
cascade loop on the furnace-conveyor NOx concentrations compared to actual pilot-plant furnace emissions levels. The pre-
dicted data correlate well with the prototype’s 18% excess-air plotted conditions
feed; the rest are all single loops.
The differential equations will in-
tegrate to dynamic equilibrium from struction should reflect real compo- Figure 3) created errors. For instance,
controller action, usually within three nent hardware. a wrong initial pre-heat gas-bank tem-
to five scan cycles. Calculated values from model runs perature estimate (shown in the upper
Significant improvement in the op- should be compared to equipment dashed curve of Figure 3) caused the
eration of an actual refinery reformer data-sheet values or literature pilot methane feed inlet temperature to be
unit was achieved using Energy Man- studies for similar operations (if avail- higher than the outlet, and, in turn
agement Control (for details, see Ref. able) to verify that compatible data the pre-heat gas-bank outlet tempera-
[4]) over competing schemes. are produced. ture to be higher than the gas bank
Plotting the model-generated inlet temperature. Lowering the ini-
Final control elements temperature profiles is essential to tial gas-bank temperature estimate
Control loops should utilize a final ensure that the model provides a and raising the methane feed-temper-
control element to the process. Valve reasonable representation of unit op- ature estimate corrected the problem
recursion formulas can be determined eration. Temperature-enthalpy dia- and a workable temperature profile
from vendor-specific, equal-percent- grams from manufacturers are also was achieved.
age data (available from Fisher, useful tools for establishing equip- After correction, the second meth-
Valtek, Masoneilan or other compa- ment-operating lines. ane pre-heater temperature varied
rable valve manufacturers). Vendor Figure 3 shows the model develop- from 1,134°F to 1,175°F, depending
software should be used to size model ment progression and corresponding on steam production.
control valves just as if they were ac- temperature profiles. The upper left- The model’s superheater outlet
tual valves; real ‘Cvs’ give loops a bet- most point in the figure is the esti- temperature was found to be 700°F,
ter prototypical response. mated flame temperature. well above the 575°F generated by
As a rule, valve equations for con- Process operating lines should the prototype. As shown in Figure 3,
trol loops should be implemented ex- be relatively linear when the heat a faulty tube size and material of con-
ternally to the process in a calculation transfer relationships are correct. struction led to an excessive econo-
block with a small time filter. This The dashed lines in Figure 3 indicate mizer heat-transfer coefficient, which
allows the process variable to change computation errors discovered in cod- produced an unrealistic economizer
slowly enough for the controller to ing execution (the lower half of the inlet temperature (700°F). The ther-
respond properly. It is much more dif- methane line was omitted for clarity), mal conductivity of the exchanger
ficult when final control elements are while solid lines indicate corrected tubing was changed in the coding
coded into a ‘C’ process module. model performance. to carbon steel and the tube size in-
The option of ‘X-Y’ plotting of model The Newton-Raphson numerical creased to correct the discrepancy.
variables from a trend chart display method requires an initial estimate Unit prototype data indicate that
is desirable. of the state variable(s) prior to execu- the reformer is natural-gas fired,
tion. For instance, guessing an initial with the second methane pre-heater
Model verification value within +/- 200 units of the ac- operating at 1,150°F. The unit typi-
Gas bank temperatures should be tual number is often satisfactory. cally produces 75,000 lb/h of 600-psi
obtained from design specifications However, incorrect initial tempera- superheated steam at 575°F.
for the particular furnace unit mod- ture estimates for the second methane The corrected model predicted
eled, and values for materials of con- pre-heater temperatures (shown in 75,470 lb/h of superheated steam at
40 CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012
 Tons Not recommended
waste/h 6.3 9.3 12.8 16.2 19.4 21 $9.00/1,000
50
Adj. inflation
$8.58/1,000 lb

40
primary air in our HYCO-methanol
2002 rate model process to the same 18% ex-
$6.95/1,000 lb cess air conditions reported in the lit-
30
erature [5] to see if our NOx emissions
Profit, %

would be comparable to a one-million-

20 Btu/h pilot waste combustor.
Model results ranged from 269 ppm
to a low of 178 ppm, denoted by the red
10 squares in Figure 4. The Model NOx
results obtained as a function of tem-
perature also appear in red. Predicted
0 emission data correlate well with pre-
0 5,000 10,000 15,000 20,000 25,000 vious literature pilot studies.
Steam flow, lb/h Calculated SO2 levels for the model
FIGURE 5. The model-predicted steam low and margins (combusting 4,466 Bt/lb ranged from 24.5 to 74.08 ppm for
solid fuel) illustrate steam-pricing strategies applicable to the prototype the blended fuel tests. In the 1,600°F
range, the model predicted SO2 was
TABLE 3. BLENDED FUEL FOR MODEL EVALUATION 88.32 ppm. The literature [5] reports a
Basis EPA 2005 Tires 2011 Blend level of 82 to 91 ppm for the two com-
Split 0.5 0.5 bustors tested at 1,550°F–1,600°F.
Model operation between 1,500°F
Waste Tires Composition H Btu/lb-mole Moles
and 1,600°F with 18% excess air
C 0.48 0.5794 0.5297 89,674.82 0.0441 matches the literature claim for con-
H 0.0538 0.1558 0.1048 12,887.32 0.0520 trol of NOx and good absorption
S 0.0011 0.0464 0.0238 3,024.80 0.0007
of SO2.
Calculated data that are close to
O 0.26 0 0.1300 literature values suggest the model
N 0.0034 0 0.0017 material and energy balance relation-
H2O 0.135 0.1584 0.1467 ships are reasonably accurate and rep-
resentative. Proper directional trends
Ash/solids 0.068 0.0601 0.0641
in any model are essential.
Total, lbs 1.001 1.000 1.001 105,586.94 0.0969
105,587 Btu/lb-mole Economics
Avg mol wt: 10.3 lb/lb-mole Green projects are capital intensive
and can take years for investment
Density: 47.7 lbs/ft3
recovery. As government regulations
HHV: 10,221 Btu/lb continue to play an increased role
in national economies, new energy
560.5°F and the second methane pre- boundary conditions. Model identifica- sources are needed to sustain growth.
heater operating at 1,142°F. tion of these break points early on can There is a vast untapped reservoir of
lead to better-recommended prototypi- man made waste, lying in the ground,
Results cal operating points. that may provide the solution to the
Five steam-to-carbon ratios were When fuel heat content is not avail- nation’s energy appetite.
tested —1:9, 2:2, 2:7, 3:2 and 3:7. Table able, as in this example, it can be esti- A 2010 study [1] has found that, in
1 indicates the predicted water-gas mated from the chemical composition of comparing emissions from landfills
shift conversions (xm). the fuel components using traditional versus municipal waste combustion
Methane conversion increased, as thermodynamic relationships. Use a using the U.S. Environmental Protec-
expected, with higher ratios. However, suitable spreadsheet for calculating tion Agency’s life cycle assessment
methane conversion above 2.7 was model fuel composition, as in our rub- (LCA) model for the range and sce-
nominal. Greater ratios diverted ad- ber blending computation of Table 2. narios evaluated, waste combustion
ditional steam to the HYCO reaction, The heat content for tire-grade rub- outperforms land filling in terms of
which increased the condensate re- ber is approximately 15,500 Btu/lb, greenhouse gas emissions regard-
cycle burden. The increased methane which compares favorably with our less of landfill gas-management tech-
conversion at higher steam/carbon estimated 15,640 Btu/lb. niques.
ratios did not offset the resulting loss Better methods are needed to evalu-
of salable contract steam at our arbi- Using the literature ate alternative energy sources and
trary 30-ton/h fuel limit. Emissions data on SO2 and NOx may processes that can increase competi-
Product flows are often sensitive to be found in the professional litera- tiveness in a global market. A good
small changes in feedrates or other ture. In our example, we set furnace process model can predict unit prof-
CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012 41
 Model predicted steam generation
by fuel heat contentm, HHV — Btu/lb
Engineering Practice
Tons waste combusted per hr: 24 – 30 Temperature: 1,500°F Methane feed, lb/h: 1,480
Profitability Landfill heat values Methane production
Shredded tire fuel blending Methane feed, lb/h
itability under changing market and 100
HHV Gross profit Gross profit
energy conditions.
90 4466 44.60 31.71
A good rule of thumb is to begin a 6651 50.72 38.29
10,221
(50%)
profitability study at the lowest set of 80 7480 51.18 39.41 (33%)
process conditions attainable and then 8351 51.57 40.46
9265 51.96 41.46
work upwards until a positive result 70
10221 52.28 42.43
9,265

Product, lb/h (x 1,000)

(40%)
can be obtained. Projects that require (29%)
60 Table 4 Steam margin
extreme conditions to give a positive Steam adjusted for 8,351 (30%) 60
return aren’t worth the time involved. profitability inflation (24%)
50 (1480)
50
If it’s profitable under minimum con- 7,480 (20%)

Gross profit, %
(1465)
(17%)
ditions, it’s likely to be profitable 40 (3701) 2005 6,651 40
throughout. It’s a handy means for (10%)
5,865 (9%)
eliminating less-promising projects. 30 5,305 30
5,040
Most model packages permit data 4,466 5,103 End EPA heat 20
20 content chart data
export directly into a personal spread-
sheet already containing your project 10 10
cost macros, which greatly facilitates (370)

economic analysis. 0 0
4,000 5,000 6,000 7,000 8,000 9,000 10,000

Using the model Fuel heating value, Btu/lb

EPA solid waste heating values [2]
FIGURE 6. The model-predicted steam generation by fuel heat content (HHV-Btu/lb)
ranging from 5,040 to 5,865 Btu/lb suggests the best steam and methanol generation production points as a function of
were model tested to determine the fuel heat content and a 30-ton/h fuel limit. The model-calculated product quantities at
potential profitability of the example 1,500°F are converted into proit margins (curved dashed lines) help the engineer as-
synthetic fuel from waste process. sess project viability (Source: Heat Content from Municipal Waste in the U.S., http://
www.epa.gov/msw/msw99.htm)
Model tests utilize the heat content
of EPA landfill fuels up until 2005 (the
year the EPA data tabulation ends), as value is based on the standard heats Combustion
indicated in Figure 6. Although greater of combustion for the elements and Add new fuel compositions and the
temperatures than 1,600°F increase blend composition. updated heating value to the Furnace
yield, they also increase emissions. A solid waste heating value of 5,865 segment of the model, based on blend-
Values above 1,900°F cause disasso- Btu/lb was selected for 2005 [2] from ing computations:
ciation of CaSO4 formed during sul- EPA’s 1989-2005 Solid Waste Heat Calculate the fuel fed to the furnace
fur dioxide removal from the fuel, and Content chart to represent the model at mid-range conveyor belt speed, ‘S’
2,000°F and above causes the NOx fuel base. The model fuel-base compo- as a function of belt width ‘w’ and fuel
concentration to go up exponentially sition matches the EPA determined density ‘d’:
in the combustion reaction. heating value.
Figure 5 plots three projected steam The base heating value is EPA’s fuel = S (ft/h) * w (ft2 wide) * d (lb/ft3)
margin lines for a low-heat content 2005 [2] reported land-waste heat
waste fuel (4,466 Btu/lb). The lowest content to which scrap tire rubber Correct the blended fuel heating value
line represents arbitrary steam pricing was added. (Lhv) for the energy lost from fuel
of $6.95 per 1,000 lb, based on a 2002 Shown below is the chemical compo- moisture and hydrogen content.
commercial contract. The model identi- sition of tire-grade rubber: The heating values of Figure 6
fied a 46% gross profit using this fuel. were determined by fuel blending.
However, the model profit margin de- CH3 Calculated heat content appears next
creased with additional steam genera- | to each point with blend composition
tion. Low fuel-heating values restrict (CH2-C=CH-CH2)n below in parenthesis. The dashed
steam generation and demonstrate the | lines represent the model-predicted
need for a higher energy content fuel. S pre-tax profits, adjusted for inflation.
The inflation-adjusted line demon- Scrap tires are assumed to contain The upper dashed line represents
strated a satisfactory rate of return high moisture. 30 tons per hour fuel consumption.
using the same fuel. If attainable, The model-blended fuel contains The lower dashed line demonstrates
steam contract pricing should be set 2.38% sulfur, 14.67% moisture and the combined effect of 20% less
at a rate that helps reduce pay out. has an ash content of 6.41% as shown throughput and a $3.56/ton fuel in-
in Table 3. crease, resulting in a 10% profitabil-
Fuel blending The mix is a simple ratio. For car- ity loss.
Table 3 indicates a typical spreadsheet bon content, the 2011 blend is: Fuel moisture for the tests ran be-
calculation that develops fuel energy C: = 0.5 * 0.48 + 0.5 * 0.5797 tween 13% and 16%, depending on
content for the simulation. Heating = 0.5297 blend composition.
42 CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012
 Methanol production remained near Pay out rises to 16.9 years at 20% less Dynamic models also give engineers
maximum at 8,760 gal/d throughout throughput and a $3.56/ton fuel in- a development tool for determining
the tests. But at 1,600°F, there was a crease. the suitability of solid waste composi-
0.9% increase in methanol output. Scrap steel sales from the tire tions as a fuel source for a particular
The upper dashed line represents blend were not used to calculate the process. As demonstrated here, en-
the best model-predicted pre-tax prof- pay out in our example. If the scrap gineered fuel pellets produced from
its, adjusted for inflation. There is a metal composition for a green project municipal waste sources can provide
noticeable slope change in the curve is unknown, EPA recycling data [6] a sustainable and reliable fuel, whose
at 7,480 Btu/lb, suggesting that lower can provide a workable estimate of use can help to reduce dependence on
heat content will be insufficient for glass, metals and other salvageable petroleum-derived fuels well into the
profitable steam production. materials in municipal waste. Resale future. n
of these commodities can reduce fuel Edited by Suzanne Shelley
Capital costs production costs, improving overall
A feasibility estimate of the mod- economics. References
eled process capitalized equipment, Salable scrap from solid waste often 1. Brian Bahor, Michael Van Brunt, P.E., Keith
exclusive of environmental permits can amount to tons/d. So salvaging Weitz, and Andrew Szurgot, Life-cycle as-
sessment of waste management greenhouse
indicates a cost of $108,479,360 U.S. item ‘A’ for example, produces ‘tons/d’ gas emissions using municipal waste com-
bustor data, J. Envir. Eng., 136, 749 (2010);
dollars as of March 31, 2011 [3]. This x ‘operating d/yr’ x ‘$/ton’ = ‘$A/yr’ doi:10.1061/(ASCE)EE.1943-7870.0000189.
includes the cost of site excavation, of additional revenue. Assuming, fuel 2. Methodology for Allocating Municipal Solid
buildings, parking, concrete, steel, con- production costs are ‘$C/yr’: Waste to Biogenic and Non-Biogenic En-
ergy,, Energy Information Administration,
struction labor, process equipment, in- Office of Coal, Nuclear, Electric and Alter-
stallation, piping, distributed control Adjusted fuel cost: nate Fuels U.S., May 2007 Report.
(DCS), instrumentation, pneumatic 3. Peters, Max S., and Timmerhaus, Klaus D.,
“Plant Design & Economics for Chemical
valves, and administrative costs, in- Engineers,” 2nd Ed., MCGraw-Hill, 1968.
cluding a 10% contingency. Taxes and (7) 4. Williams, A., Wilson, R.G., and Chang, S.,
interest charges were not considered. Reformer model improves control, A. Wil-
liams, R.G. Wilson. S. Chang, Hydrocarbon
The process equipment that is de- The example process has potential Processing, Nov. 1994.
preciated (but not limited to) includes given the fuel tests were restricted to 5. Nack, H., Litt, R.D., and Kim, B.C., Multi-
electronics, instrumentation, vessels, 30 tons/h maximum of blended waste. solid fluidized bed combustion, CEP, Janu-
ary 1984.
motors and vehicles. The total depre- It is reasonable to expect that much
ciable sum is $35,782,083. higher steam outputs are attainable 6. Municipal Solid Waste Generation, Recycling,
and Disposal in the United States, Tables and
Assuming a 10% salvage value for since the blended heating values are Figures for 2010, U.S. Environmental Pro-
tection Agency Office of Resource Conserva-
the process equipment and a 40-year comparable to coal. tion and Recovery, http://www.epa.gov/osw
useful plant life, the annual straight- Payout is acceptable for the green /nonhaz/municipal/pubs/msw_2010_data_
tables.pdf. Tables 1–29, Nov. 2011.
line depreciation is: project and it shows promise, demon-
strating profitability under minimum
(3) throughputs. Authors
The engineer should consider fur- Allen Williams is a private
ther investigations of this process consultant with more than 30
years of experience in power
For 30 ton/h of fuel consumed and a based on model predictions, with an generation and petrochemi-
cals design (Email: allen_k_
50% fuel blend (10,221 Btu/lb), metha- eye towards improving conversion and [email protected]). He
nol production is: plant throughput for enhanced profit- has developed statistical con-
trol algorithms for the power
ability and better pay out. industry and has two previ-
Higher methanol and steam pricing ous technical publications. He
holds a B.S.Ch.E. with gradu-
is not recommended as a means to- ate study in kinetics from the
(4) wards this end as market factors play University of MO-Rolla and New Mexico State
University. He is a registered professional engi-
The predicted steam generation gross an increased roll in product prices. neer in the state of Michigan.
profit is:
Conclusions Ken Dunwoody (Email:
[email protected]) prior
Dynamic models are useful tools to to his retirement, chairman
evaluate any process that can be repre- and chief operating officer
of RCR Systems, Inc., where
(5) sented mathematically. Engineers can he and the firm’s general
partners developed the RCR
Annual revenue is: ($14,129 + $8,760) use them to not only emulate a given solid waste handling system
x 313 working d/y = $7,164,257. process but to test competing control prototype. His firm holds five
patents in waste-separation
The predicted pay out period, as- strategies, trend cost variables, and machinery. Prior to that,
suming no interest charges: identify and implement process im- Dunwoody was general manager of the glass
manufacturing division in charge of production
provements. Modeling a process pro- at Coors Container. He holds a B.S. in business
vides the best means of culling out the administration and Associate degrees in digital
electronics and industrial management the Univ.
(6) most promising projects from the pack. of Colorado and the Denver Technical Institute.

CHEMICAL ENGINEERING WWW.CHE.COM JULY 2012 43