You are on page 1of 9

ENGINEERING PRACTICE

hen planning the design of a batch-type plant to produce commercial quan-tities of a product or a mix of producs, every full-size reactor in the plant will usually have a working volume that is some mltiple of the ca-pacity of a pilot plant reactor. Produc-tion capability of the larger plant is often predicted from cycle times expe-rienced in the small-scale equipment. Process engineers usually have little trouble maintaining the same cycle time for production steps such as the charging of liquids and sohds to the reactor. They are not so in control, however, when it comes to the times required to complete temperature adjustment steps in the larger units. Given the same vessel and jacket configurations and restricted to the same operating conditions as at the small scale, the same steps take longer at the larger scale because heat transfer rea does not scale-up at the same rate as does reactor volume. This series of four articles addresses four issues. Two are presented this month and two next month. Part 1, below, explains how the geometry of scale-up "short changes" the larger unit ofheat transfer rea. Part 2 will show the mathematics of scale-up to predict the times for temperature adjustment steps at the larger scale. Part 3 introduces changing the jacket to a zoned configura-tion. Part4 combines the aspect ratiocor-rections to the zoned configuration to reduce prolonged cycle times. Considering the aspect ratio Go to a meeting of process engineers to discuss scale-up. At least one od hand will remind others that scale-up geometry "short changes" the larger reactor of heat transfer rea. Next will come the intuitive judgment, "Sure, but if we make the new reactor skinnier and taller..." This sttement is a suggestion to maniplate the "aspect ratio" of the reactor. Define that as "the tangent-to-tangent length of the reactor divided by the reactor diameter." It is usually not good prac-tice to design reactors with large aspect ratios. Unfortunately, the mathematics of scale-up is too complicated to permit even the old-hand process engineer to offer an instantaneous intuitive reply as to the best ratio.
96

REACTOR CONSIDER
The reactor aspect ratio in scale-up
NOMENCLATURE
factor for determining the capacity (gal.) of tfie bottom head = 0.606 for ASME Standard F & D heads (3fg2 - d2)/g3, see Equation 15 total heat transfer rea, ft2 Absolute valu of p, see Equation 20 (3Pg - 2ad)/g3, see Equation 15 (f3 - a2)/g3, see Equation 15 5.875, see Equation 3 reactor inside diameter, ft. defined Equation 19 t + [4a(F-l J/7.48], see Equation 9 fraction of total volume oceupied by working volume. irF, see Equation 9 straight side length covered by VW,ft. reactor straight side (tangent-to-tangent) length, ft. (f+gRl), see Equation 14 (a+dR]), see Equation 14 defined Equation 17 defined Equation 18 aspect ratio, L/D, dimensionless volume scale-up factor = VWj/WV, heat transfer rea scale-up factor = A2/A1 factor for determining the surface rea (ft2) of the bottom head = 0.931 for ASME Standard F & D heads working volume, gal. Roots of the cubic equation 21 (SUFVM3/N2), see Equation 14 angle, see Equation 20, radians

This article develops the mathematics of reactor scale-up. We will show, when the aspect ratio of the two reactors is the same, the increase in heat transfer rea is only a fraction of the increase in volumetric capacity. We will also develop a shorteut to quickly compute the aspect ratio required to make the heat transfer rea increase proportionally to the volume. The mathematics of scale-up Consider the reactor in Figure 1; the working volume in that vessel is given by the expression: Wy=F [aD3 + 7.48(n/4)D2L] (1) with the conversin factor 7.48 gal = 1 ft3. Next, define aspect ratio: R = L/D (2) and define the constant, d: d = 7.48M4) = 5.875 (3) Substitute d and R into Equation 1: WV=F[aL)3 + dRD3] (4) Solving Equation 4 for reactor diameter gives:

aa =

A=
ABV = bb = ce = d = D = DV = f = F= 9= J= L= M= N= P= QV = R= SUFV = SUFA = t=

-W/-1/3
D= \ (a+dR)

(5)

Assume that the total rea of the bottom head is effective for heat transfer. The total surface rea of the reactor in Figure 1 that is avaable for heating or cooling the working volume is a sum: A = tV2 + nDJ (6) Working volume can be defined as in Equation 1 or it can be defined as re-lated to the straight side length that it filis. Make this expression in "J" equal to Equation 1 with the constants as ex-pressed by Equation 3.
F [aD3 + dD2L] = aD3 + dD^J (7)

wv =
YY = = 0=

Solving Equation 7 for J gives

CHEMICAL ENGINEERING / DECEMBER 1997

Part

DESIGN ATIONS
(8)
JaD(F-l)
+ FL

</ =

TABLE 1. THEEFFECTOF SCALIN UP VOLUME AT CONSTANT ASPECT PATIO SUFv 2 3 4 5 6 SUFA VALU SUFA Os %SUFV 1.59 2.08 2.52 2.93 3.30 10 4.65 79 69 63 59 55 47

Edward H. Steve Day& Zimmermann International

Substitute Equation 8 into Equation 6. Also use the definitions of Equations 2 and 3 and let g = nF. A number of terms are collected into fas follows: f=t + [4a(F-l)/7.48] After suitable rearrangement, this is a short equation of heat transfer rea: A = fD2+gRD2 (9) Defning scale-up factors Scale-up is a ratio calcuktion. Define scale-up factors as: SUFy = volume scale-up factor =

FIGURE 1.

These geometrlc parameters define a batch reaction vessel

wv2/wv
SUF = heat transfer rea scale-up factor =A2/Aj where the subscript 1 refers to the pilot scale reactor and the subscript 2 refers to the large scale reactor. A common restriction is usually made that the scaled-up reactor has the same bottom head type (e.g., ASME Standard F & D [1]) as the pilot-scale reactor. This keeps g and fthe same for both sizes. Otherwise, the math gets impossibly messy. Calclate the following expression for SUF using subscripts in Equation 9 to denote the reactor scale. DJif+eR*) A (10) Dfif + gRj Substitute Equation 5 versions of D and D2 into Equation 10. Use the defin-ition of SUFy to eliminate WV. F will remain the same in scale-up. The re-arranged mathematical expression relates the magnitude of SUF to the valu of SUFV:
a + dRi SUFA=\SUF) dR f+i f +gRi
DAVID WHITCilER

TABLE 2. EXAMPLE OF SOLVING CUBIC EQUATION BASIC PROCESS DATA Input (L/D), SUFV F a t d f g N M aa bb ce Valu 4 2 0.8 0.606 0.931 5.875 0.866 2.512 24.09 10.91 4.479 -8.708 -1.653 -0.0627 (R2)i (R2)2 (R2>3 8.895 -0.05256 -0.1341 (Root #1) (Posltive root) (Root#2) (Root #3) SOLVING 8c CHECKING OF CUBIC EQUATION Output P QV DV ABV PHI PHIR YY, YY2 YY3 Valu -26.93 -53.78 -0.4019 26.93 1.350 0.02357 5.992 -2.955 -3-037 (Angle Phi n degrees.) (Angle Phi in radlans.) Roots are fhree, real and unequal

(11)

ROUNDED RESULT FOR (L/D), = 4 and SUFV = 2 is (L/0)2 = 9

Equation 11 is a general expression

CHEMICAL ENGINEERING / DECEMBER 1997

97

ENGINEERING PRACTICE
-^-~-

Part SOLVING THE CUBIC EQUATION

An Excel 5.0 spreadsheet is the easy way fo solve Equation 15. Two key constants based on ASME Standard F & D vessel heads are a= 0.606 and f = 0.931 []]. It will be easy to find three roots for nu-merous inpuf vales of R\, the aspect ratio. Table 2 shows the example calculation for R] = 4. For this particular cubic equaiion, the following traditional solution method applies [2], Read the spreadsheet this way: 1. Design choice inputs: R, = 4, SUFV = 2, F = 0.8 2. Constants:.a = 0.606, t = 0.931 3. Calclate d, f and g 4. Calclate N, M and 5. Calclate aa, bb and ce. These are the coefficients of the cubic equation

9. Compute the angle 0 (BV = the absoluto valu of the pararneter p]

QV2 0 = arceos ABV


27
10. Calclate the three roots, YY,, YY2 and YY3

(20)

-f
eos ABV
0 + 7t

(21)
(22) ( 23)

p = j[3bb~aa2]
(17): QV = ~ [2aa? - 9aa x bb + 27cc] (18):

yys=-2^
11 .Calclate possible Rj vales (R21 =YY - (aa/3) (24)

7..

Dv.
2
In this example, DV<0, The three roots are real and unequal. Because DV<0, a trigono-metric solution is preferred.

12. Select the positive valu for R2 as the desired solution. (19) Table 3 shows the results obtained for Rj when sofving Equation 15 for various typical vales of R) and SUFv- For TABLE 3. REQUIRED R2 VALES FOR SUFA completeness, it also lists the raw vales for R2 obtained in the spreadsheet calculation. In = SUFV ATVARIOUSR, VALES each case, these were the positive roots. Required R Valu
2

SUFU

Rounded 3 5 9 4

1 that describes how the heat transfer rea increases when increasing the vol-ume of a reactor from the pilot to the production scale.

2 3

Raw Valu
3.11 5.13 9.14 4.01 6.47 4.96 7.89 10.81 6.92

1.5 2

6.5 5

bb = (3pg-2ad)/g3 cc = (fi-a2)/g3 Select a valu for R and use Equation 15 to calclate the R2 for SUFA = SUFy. Solving this cubic equation by spreadsheet is covered in the box. Handy rule of thumb There are times when the computer-ized method for solving Equation 15 may not be practical. Suppose you are away from your desk and need to quickly derive the valu for R2 needed to make SUFA equal to the SUFy. The rounded data in Table 3 can be used to deduce a handy rule of thumb: Pilot Scale-up aspect ratio Rj = R2 = (2xSUFv)-l ij = 1.5 = (2.5 X SUFy) -1 R1 = 2 R2 = (3x SUFy) 1 R1 = 3 R2 = (4X SUFy) -1 R = 4 R2 = (5x SUFy) -1 For every case, the multiplier for SUFy is (Rj + 1). Based on this, use the next equation to approximate the L/D required for a scaled-oip reactor that will haveSUFA=SUFy. When circumstances and time permit, use an exact calculation. R2 = ([R, + 1]X SUFy) -1 (16) Edited by Peter M. Silverberg References aspect ratio l

8 2 3 Main tainir.g aspect ratio 11 4 7 Because there are some advantages to 3 . maintaining geometric similarity be-tween 10.8 11 2 the plot and the production scale reactors 1 9 (discussion of these is outside the scope of 3 this article), the larger unit often has the tion) substitute SUF for SUF on 8.9 the left side V A same aspect ratio as the smaller unit. In this of Equation 11. Then take both sides to the 4 case, R2 = R and Equation 11 reduces to: third power to clean up fractional exponents: SUFA = (SUFyl2'3 (12) Ja +2 dR^2 jf+gRj]3 f + gRil (13) For example, if the larger reactor is twice SUFV = 3 a + dR the size of the pilot unit (SUFy=2) and the LH two units share the same aspect ratio, the Define the following constants to make the heat transfer rea only increases by 1.59 (= rest of the ____ discussion simpler: 22/3). The SUFA would, therefore, be 79% L of the SUFy. As the numrica] vales in N=(a+dRj) Table 1 show, the SUFA becomes an M=(f+gRt). increasingly smaller percentage of SUFy as = (SUFvM3/m the latter be-comes larger at a constant Make substitutions into Equation 13 and aspect ratio. rearrange: Intuitively, those who are involved in the 3 2 scale-up exercise know that a full-scale (SUFyMZ/WMa+dR^ (M /N2)x(a+dR2) = 3 (14) reactor with a large aspect ratio will have a (f+gRJ total heat transfer rea that makes up for the Rearranging this and grouping all the "shortage" that would oceur at constant constants together, results in the following aspect ratio. cubic equation for calculating the R required
2

R2

"Largest" aspect ratio required 3 To determine what aspect ratio is needed to R2 + aaR^ + bbR2 + cc = 0 cause SUFA = SUFy (al-though it may not where aa = (3fg2 - d2) Ig3 be a practical solu-

to make SUFA = SUFy:

(15)

1. Dimoplon, W., Jr., How to Determine the Geometry of Pressure Vessel Heads, Hydro-carbon Process., pp. 71-74,
98 CHEMICAL ENGINEERING / DECEMBER 1997

August 1974! 2. Turna, J.J., "Engineering Mathematics Handbook," p.7, McGraw-HiB, New York, 1970.

ENGINEERING PRACTICE

MORE

his is the second in a series of ar-ticles that addresses selected reactor design considerations. The first (p. 96) discussed how the geometry of scale-up "short changes" the larger unit of heat transfer rea. This article will develop how the math-ematics of scale-up can be used to pre-dict the heat loads and times required for completing temperature adjust-ment steps at the larger scale. Traditional Heat Transfer Model The figure shows the traditional model used for unsteady-state heat transfer analysis. Components are an agitated reactor and a non-isothermal jacket fluid. The mission is to cool or heat the contents of the vessel from initial to final temperature. Note 2 with the drawing recom-mends including the weight of the ves-sel in an analysis. The temperature of the reactor walls must change along with the contents. For simplicity, how-ever, this feature is not included in the mathematics in this article. The as-sumption is plausible based on relative heat capacities and weights. Calculating the time required to change the internal temperature of the reactor starts with differential heat balance equations: Por Cooling: = wC(t2-t1) = UA(LMTD) (1) (b)
-**.%

REACTOR CONSIDER
LMTD ln (t2-T)

(*x-n

Predicting temperature loads and times for scale-up


NOMENCLATURE All items below are introduced for Part 2. If common to both parts, they will not be repeated. ai = A= Aj = A$ = b= C= Cp|_ = k= UA]/wiC, see Equation 18 total heat transfer surface rea = As + ABift2 heat transfer rea on the bottom head of the reactor, ft2 heat transfer rea on the straight side of the reactor, ft2 (SUFv)2/3/k, see Equation 18 heat capacity of acket fluid, Bfu/lb-F heat capacity of mass of material in the reactor, Btu/lb-"F scale-up terrn for acket How, dimensionless

(c)

LMTD= log mean temperature differ-ence, see Equations 3 and 4 ML = t] = t2 = T= U= (3) w= X= mass of material in the reactor, Ib acket inlet temperature, F acket outlet temperature, F temperature of mass of material in the reactor, F overall heat transfer coefficient, Btu/hr-ft2-F acket flow, Ib/h
eUA/wC see Eauation 8. Subscripts explained Equations 13, 14andl5.

(a) For Heating: LMTD MLCpLd= = wC(.t1-t2) = UA(LMTD) (d) da


(e)

(2)

()

Then compute the log mean tempera-ture differences

For coohng: For


(4)

fl =

time required to complete the step, h

heating:

_ [(h-ry^-T)]

We will retain the lower case "t" for representing jacket temperature and the upper case "T" for representing internal reactor temperature. It will not matter if the temperature adjustment operation is for cooling or for heating. Another simplification we will assume is that both U and A remain constant over the temperature adjustment cycle. Jacket Outlet Temperature Equate parts (b) and (c) in Equation 1. Use the definition of LMTD for the cooling mode. This equation allows the di-rect calculation of the jacket outlet temperature, t2, in the figure.

[T-tt)-(T-h
100

temperature at the start of the coohng cycle, T, to the temperature at the end Clearing and of the cooling cycle, T2. When cooling, T2 < T. The (g) form of rearranging... Equation 8 shows clearly that the JA ln (6) pre-dicted t at the start of the cycle will wC 2 be larger than the predicted t2 at the end (T-h), of the cycle. This is an important aspect of (T-H) un-steady state heat transfer. Don't Therefore, in exponential form: overlook it when completing heat load (T-t1)ICT-t2) = eAIwC (7) calculations. Equate parts (e) and (f) in Equation 2 LettingX= eUA!wC and and use the full definition of LMTD for substtuting gives the heating mode. The result will be t2 = mX-V+tJ/X =T + [(tj-T)IX] (8) (g) (h) The vales for t and X in Equation 8 are constant. Temperature T vares from the u>C(t2-t1)=UA
ln (5)

CHEMICAL ENGINEERING / DECEMBER 1997

(T-h)/ ". /(T-h)

Part2

DESIGN ATIONS
Equation 8 for the direct calctdation of the jacket outlet temperature, t2. It should be evident that the predicted t2 at the start of the heating cycle will be smaller than the predicted t2 at the end of that cycle. Temperature -adjusting time Equate parts (a) and (b) in Equation 1 and substitute t2 from Equation 8. Steps to Equation 11 will be shown. It allows the direct calculation of the time required to complete a cooling tempera-ture-adjustment step in the reactor. X M,C
t.
L^PA

TABLE . PARAMETERSFORTHi AGITATED REACTOR


:. As

.c
:
.:

Edward H. Steve Day& Zimmermann International

75 ft2 188.9 (t2 = 1.0Btu/lb-F CpL = 0.516 Btu/lb-F ML = 29,962 Ib ti = 104F T, = 221F h = 158F 2 u = 73.5 Btu/hr-ft -F w = ; 28,798 Ib/h AB
s s

1 Vw=waterf low needed to achieve 5 ft/s (usual mximum valu) in the partial pipe jacket. 2. ML has been simplified forthls article. DZI recommends including the heat required to adjust the temperature of the vessel Itself. 3. Overall heat transer coeff icient is determined separately for reactor contents with DZI reactor-design computer program. Jacket fluid in, at t,,w,C FIGURE. Thls sketch defines the parameters used in the table for an agitated batch reactor

X1 = e Jwfl
X2 = e %,C ^2 =
U

A,/

(14)

(9) (10)

lj'T-t wC IX-li'T-tx Integration over the appropriate in-terval (Time from 0 to fl, T from T to T2) (11) gives: wC -[3

da = -

[x-i.

I n

"21-i l

To get the heating temperature-ad-justment time, do a similar calculation. Equate parts (d) and (e) in Equation 2. Use Equation 8 for t2. Equation 12 is the result. (12) n= MrC L^pL wC IX
i-T, i-3.

Calculation example No. 1 The table lists vales for the process parameters related to the agitated reactor with a partial-pipe coil jacket shown in the figure. These are used in the following example for completing a cooling temperature adjustment step in that reactor. Calclate A = Ag + Ag Calclate X as eUA'wC = 1.961. Apply Equation 8 and wC(t2 -1) for conditions at the start and at the end of the cycle:

Heat Removal t2,F Btu/h Tons Start of step 161.3 1,650,000 137.5 End of step 130.5 763,000 63.6 Use Equation 11 for the entire cooling step and, -l = 0.85 h The results show quantitatively the significance of considering both bound-aries of the unsteady state operation. The load needed at the end of the step is less than one half the load needed at the start of the step. Reporting the valu at only one boundary could be misleading. For example, the total heat load (provided by chillers or heaters) from mltiple reactors with overlap-ping temperature-adjustment cycles could be skewed if the changes within those cycles are not considered. Adjust the time for scale-up 1. General Considerations: For a reactor at any scale X=e /wfi (i3) Where X, A and u>i are specific vales for the scale being considered. For in-stance, when scaling-up from WV to WV2:

(15) If the larger reactor has a larger size half-pipe-jacket, the jacket flow in-crease is linear: w2 = kuij Use the defi-nition for the heat transfer rea scale-up factor from Part 1, (SUFA = A2/Ai). We can readily determine that m ^-. ^%p (16) Scale-up is thought of in terms of re actor volume. If the reactor is scaled-up with the same aspect ratio (L/D) as the pilot model, Equation 12 from Part 1 applies. Repeated here as Equation 17. SUFA = lSUFvpl3 (17) Substitute Equation 17 into Equation 16. Apply the algebraic law of powers. A simple expression for X2 emerges. X2=(Hb (18) where a/ = UAJ/WJC and b = (SUFv)2'3/k 2. Jacket Outlet Temperature Substitute X2 into Equations 8, 11 and 12 to determine the effect of scale-up on the jacket outlet temperature and on the time required to complete the temperature adjustment step. To pre-dict directly the expected valu for the outlet temperature (t^ from the jacket of a reactor scaled-up by some volume

CHEMICAL ENGINEERING/DECEMBER 1997

101

ENGINEERING PRACTICE

Part2

scale-up factor (SUFy), plug Equation 18 into form (h) of Equation 8. h-T

2 indicates the scaled-up reactor.) ML2CpL _SUFV]ML1CpL To predict directly the expected valu for the time required to complete a cooling step in a reactor scaled-up by some volume scale-up factor {SUFy), Equations 18 and 20 can be
Plus you get all these tools n one package:

i,=T +

substituted into Equation 11. 'SUFvjML1CpL~


Q=

(19)

3. Final Adjustment Time Equation 20 results because M = SUFy x Mi2 and u>2 = kwp (Subscript
Coade has raised the bar another notch with CAESAR II Versin 4.0, the new, native Windows 95/NT versin of the world's most

*J

ln

Tizh
T2-h

(21)

rf-i

popular program for pipe stress analysis and

CAESAR II
4.0

Versin
Pipe stress analysis

Pipe Stress Analysis for Windows


5^51 .SaiPlatelri.riiit -.."-i Jien-;Wi.ma
r7

^r. I :~l;.S:Frv. :::

design. CAESAR II 4.0 provides: Wireframe plots & rendered graphics Simultaneous display of input spreadsheet and plot Cut & paste capabilities

ENGINEERING

SOFTW ARE

HOUSTON, Phone: Fax: WWW:

TEXAS

USA

API, NEMA, WRC, AISC component evaluation Automatic spring sizing Component databases Structural steel modeling (now with spreadsheet input) Automatic underground pipe modeling Complete material datbase with allowable stress data 2 codes for fiberglass reinforced plstic pip API, WRC & BS nozzle flexibilities Bi-directional link to CADWorx/PIPE drafting package CAESAR II is the world's standard because it is technically comprehensive, flexible, easy to use, reliable, and is supported by the industry's best technical support team.

Calculation example No. 2 The objective is to scale up the reactor to twice the volume. i.e., SUFy = 2. As-sume U2 = Uj. Assume that the size of the half-pipe jacket is identical at both scales. In other words, k = 1, therefore w2 = lXw}. Use Equation 18 and calclate a and b a, = 1.961 b = (SUFyp/3/k = 22/3/1 = 1.588 Apply Equation 19 and wC ( - j) for conditions at the start and at the endofthecycle. Heat Removal Btu/h Tons t2,F Start of step 180.8 2,212,000 184.3 Endofstep 139.5 1,022,000 85.2 There are increased heat loads at the larger reactor scale because the heat transfer rea is 1.59 times larger. The coolant rate has remained constant. Use Equation 21 for the entire cooling step. Result: O = 1.26h. The time required is also larger. It computes to be approximately 48% more than the 0.85 h predicted at the smaller scale. Mltiple temperature adjustment steps usually occur in a chemical reac tor production cycle. The sum of such cycle time increases could be signficant. The plant capacity at the larger scale will be adversely impacted if the cycle times are not corrected. Edited by Peter M. Silverberg Author
Edward H. Steve is a director of technology for the Process & Industrial Divisin, Day & Zimmermann International, Inc. (1818 Market St, Philadelphia, PA 19103; Phone: 215-299-8795, Fax: 215-299-2236) Formerly, he was a lead process engineer in the Life Sciences Divisin. He prepares the conceptual engi-neering for specialty chemical and petrochemical plant design prqjects and contributes to technical development, technical supervisin and business development. He has worked in design, operations and research for more than thirty years and is the author of sev-eral technical articles. He is a gradate of Cornell University (B.Ch.E.), a Member of AIChE and a Professional Engineer in six states.

281-890-4566 281-890-3301 http://www.coade.com

CALL FOR FREE BROCHURE 1 -800-899-8787


For More Information, Circle 71
102 1997 CHEMICAL ENGINEERING / DECEMBER

You might also like