Engineering Practice ‘Sunil Kumar Worley Engineering then handling gas, we usu- ally encounter two types of flow. The most common one is non-expanding flow. This is the normal condition in most oper- ating process lines. Generally, the flow is isenthalpic. The impact ofa ve- locity change is minimal and assumed to bo constant throughout tho length of the pipeline. Any pressure decrease is attributed to frictional losses in the line, If the total pressure drop is less than about 10% of inlet pressure, the unmodified Darey equation (1), ean be applied to calculate the flowrate with reasonable accuracy. By comparison, expanding flow is the engineer's headache. This. worrisome ‘occurrence staris from an abnormal flow condition, which manifests itself by a very high pressure drop. Flowrate caleulations have to account for the ef- fects of adiabatie expansion and frietion losses at high velocities. Hore is a sam- pling of abnormal situations: * At the tailpipe of high-pressure lowdown systems during shut- down or depressurization; soon par- ticularly on offshore petroleum and gas production platforms * Accidental gas-blowby as a result of inadvertent opening a drain valve of a pressurized vessel, or stuck-open failure of level-control valves ‘+ Rupture of heat exchanger tubes + Bursting of a rupture dise * Opening ofa pressure safety valve Ifyou are going to plan mitigation pro- cedures for abnormal events, or to de- sign appropriate discharge systems, a pre-check of flowrates is needed for the design record. It is not easy to fig- ure this out in a realistic frictional system, when the flow approaches the choked (critical) condition at an end- point. The equations are incredibly complex. However, we have comput- crs and sit will bo easy to differenti This method encompasses a wide range of isentropic and frictional factors, and subcritical flow ate between choking and non-choking flows and to estimate their magni- ‘tude. This article provides an Excel spreadsheet to put on tho computer, address expandingflow conditions, and determine flowrates without hours of resoarch, Specific constraints We focus upon expanding flow through constant diameter pipes. If you have an oritic, see Reference (2. The frst item of business is to ascertain the ‘thermodynamic condition of flow. Its either isothermal or adiabatic. In general, for short pipes (less than 10 m), the assumption of adiabatic conditions is reasonable, Gases have a poor heat-transfer coeffisient. We ean thus, farthermore, picture a massive flow (approaching sonic velocity) in 9 relatively short pipe, For longer pipes, the assumption of ‘an isothermal condition is reasonable. However, with pipes short than 100 nm, tho realistic flow condition loane more toward adiabatic, Therefore, for the purpose of deriving critical corre lations within this article, adiabatic conditions have been assumed. The flowrate can be determined using appropriate flow equations, ei ther the Darey Equation [1] or the Robert Kern Bquation (3). If eritical flow is the status, one way to deter- ‘mine flowrate is from the Crane cor- relation plot [d]. The Crane method simply uses the Darey formula modi- fied with an expansion factor You Ui), This expansion factor takes int acount all modifications in flow re- quired as a result of variations in density and velocity. The Crane plot correlates Yo, AP/P and K. (soo Nomenclature ‘Table on p. 63 for symbol definitions). However, the Crane method, being plot-based correlation, poses a fow limitations 62 CHEVICAL ENGINEERING WwW.CHE.COM OCTOBER 2002 that have not been widely publicized, Hore are some: © Twoinputshave diserote and limited ranges: K from 5 to 90 and y from 1.0, to 14, This raisos the need for un- easy extrapolations and interpola tions. We will show that y and K are key variables because Yor and (AP/Polex are strongly influenced by them + Reading a plot is an imprecise exer- ‘The accuracy of the data is sub- Ject to the precision adopted by the user. Moreover, we find that the user tends to pick heavily conserva tive values, and accordingly the ro- sultant flow gets unnecessarily overestimated. ‘Thermodynamics overview ‘The approach used in this article is purely thermodynamic and rational (61. No oxporimental data or empirical correlations have heen used. The goals of the equations and the resultant spreadsheet are to: 1, Cheek the status of flow. It is either critical or sub-crtical 2, Determine (AP/Po)cp. 8, Determine the flowrate, ‘When a compressible fluid accelerates through a restriction (such as a pipe, orifice, venturi or nozzle) as a result of massive pressure reduction down- stream, the velocity and specific vol- lume inerease at the expense of en- thalpy. The relevant thermodynamic equations for the stato of the gas follow a reversible path, but with reduced isentropic efficiency. Reduced isen- ‘ropie officioney is approximately the ‘sum of frictional losses and heat losses. ‘We start with a basie energy equa- tion and develop from it an equation for ovorall eritical-pressuxe ratio, Since there are a plethora of subscripts, the entire flow system has been split inta nodes. They are shown in FigureACHEMA 2003 ¢ TM aia aa a {G35 DUETS iam A ZA ey Cea Nis eg Lic TN Toa a) Pea Fax +449 (0) 69 7564-201 anh Tel. +449 (0) 69 7564-290/-038/-434 Vea eer kc eee Perien/227 wwwéache aC y eee tung 19-24 May 2003 ¢ Frankfurt am Main DTCC Ley Research and Innovation Engineering, Plant Design and Construction Tem ects eRe NM? TOR ee ACCS PM a CaM ON Reece Ser ond one eae) MU eC oT Cesc OC eM OS Ue ee a ne ON eT Bi (Teese ONO MLA LEO Please send me furthe ion on ACHEMA 2003 {1 Documents for Exhibitors [_] Visitors Information ieee io Se Fe TTT inetInternal coss-ection area of pipe, m2 da, of pipe, mm Ipc rertl ope, imersionless Noe ion Foor, Frictional hood losses, m2/s2 Rexstonee colficien, dimensionless 1 = Lenghh of low pipe, m 1m = Slope depen Eqvton (191, [My = Moch Number oil p ofthe Pipe (sat of fowl, dimensionless My = Mach Number of cule ip ofthe pipe, dimensionless Po = Stagnation upstream presure, Peat we, Presi ote tip fhe fber eiewl eae Frese ot out tipo he pps, fe Pa_= Ambient pressure, kPa 1=@;/Poca= Overall critical merge Up = 0 Stagnation upsvecm velocity, m/s Oj = Vale nt ip fb " lNatctow)mfs ty = Vasc oat fle pi, i = Myf ding rain Vora oki pareatthe Q Beaming” es Vy © Spec ce nei oihe 1 peel mil Vp = pectic etme ct ul po ho W = Hes Rowan citcal ke/h Wy & obo of abel (3P/a, cimensionless Y= Volue of Y corresponding to x (see Equation 18), dimensionless Yow * Gael expansion fxr, Napeinle 4. = Fie change opera, Smonsionons (AP /Poce= Overall crea diferontal rosie roto, dimensionless +y = Roto of ipecticheat at consort Preaek Erol! velume at vosreom Sonoma Node 0 is within a stagnant up- stream system, Node 1 is the starting point of low Node 2 is the discharge tip of pipe into the ambient atmosphere, which isat a low pressure In real situations, any expanding flow ‘emanates from an inexhaustible source such as avessel, oralarge-bore pipe. AS- sume that the upstream pressure P is both constant and stagnant. At the source conditions (Node 0), the assump tion of no-flow is reasonably valid. ‘At the upstream tip of the pipe, the flow starts abruptly with some initial volocity uj. The pressure Pp has dropped to Py as a result of an in- crease in kinetic energy. As the flow progresses down the pipe, the thermo- ‘dynamic changos and frictional losses, will further tond to drop the pressure until the critical pressure Pp is reached at the downstroam tip of pipe (Node 2). The pressure P» ropresents the critical pressure at the tip, rather ‘than the actual downstream ambient pressure. We now have a criterion for evaluating critical flow. IEP < Critical Pp, choking exists + IfP4 > Critical Pe, there is no choking Energy balance ‘We next develop equations correlating P, and P3. In an adiabatic flow process, there is no heat transfer, and ‘the shaft work is generally zero. As ‘the flow expands in the direetion of de- ‘easing pressure, enthalpy (the total energy, the product VAP) decreases ‘and transforms simultaneously into Kinetic energy (Au2/2) as the flow tends to stay isentropic. However, the ‘entropy in fact increases because of vvolocity-induced frietion. Equation (1) is a fundamental energy balance [5] applicable over a fraction of pipe length, Al vap+ ar a Equation (2) is an expression from Reference [6]. nat aS) ® Plug Equation (2) into Equation (1). au? at )u’ vars s(mS)e=0 co) In differential form, as the deltas shrink to infinitesimals, we arrive at Equation (4). aya? @ vaPendve( yi) 0 Now, vide though Houten (0) by we, 20g 2 (4 © Substitute: w? = (W2V2/42) for flow 1h Sonatas cottons ptt et Equation 2AtdP , Adu av nee Hae) 0 © nigra cepted -) titen solr Mode 1) and he culeb de 2 conditions of pen Corn carte roca pressure Py FIQURE 1. Consider a small diameter pipe between an inexhaustible high-pres- Sure source and a large ambient receiver. {Ris possible thatthe flow from the pipe will approach the speed of sound 242 FP of du, ipa). elves ale o)° a When dealing with adiabatie oxpan- sions, tho fundamental PV relation- ship is Equation (8). Solve Equation (8) for V, and write that into Equation (7) to get Equation (9). = ® 07/2 limes tem) 0 wg ote led, Consistent with Figure 1, Py and V; are constants. They are the upstream conditions, Taking the constants out of integration, we got Bquation (10) alle ctelles)- 0) Intogration gives ws Equation (11). (1) an wera re 7 py eal t2) fra lta We ean factor ot P sed to the (rt?) power. fom the Bt sein ‘Fie enn: tc ding ono essary igera, Banton (1) tans torn nte Equation 2) ast, |(a)? s}sou(t)sft-0 WAC) LA, 4)" By ‘The Mach Nomberand tho resitane ccfiint dO have bmn denen Re. cree. Ue the deiitions ‘Equations (13) and (14). Substitute into ‘Equation (12) and get Equation (15). get WY (as) My, Pat ie = wep eels i‘ sae wats) foe) eg, CHEMICAL ENGINEERING WHW.CHECOM OCTOSER 2002 63‘TABLE 1. LIMITING CRITICAL VALUES K yas 1 a You | GP/Padca| Vor | @PIPdce| Yer | GPiPcn| Vor | GPIPoen I 052 | 064 [osi—| 0460.50 [068 | 050 | 070 2z 054 1 os3_| 0.49 | 053 | 077, 052| 073_| 081 3 058 | 0.70 | 086 | 073_| 068 | 075 | 084 | 0.78 | 089 ‘4—|~o71_|—060| 0.74 | 058 | 076 | 087 | 078 | oss | 0.80 | 8a ‘3 | 074081 | 0.76] 0.59] 0.78 | 0.88 | 081 | 086 | 0.82 | 084 ‘6 | 076 | 0.62 | 0.78 | ~0.40_| 0.80 [0.88 | 042 | 0.56 | 0.84 | 0.88 7_|_ 07] 062 [0.79 | 0.40 | 0.81_| 088 | 04 | 086 | 0.85 | 0.84 ‘| 076_|~063_|o.80_| 0.1 043 | 089 | 04s | 056 | 0.86 | ona ‘9|~079| 063081 | —~0.st_|084_|089| 045 | 086 | 0.88 | 084 1o|0.80—|“063 | o182_| 0.61 [085 | 059 | 0.86 | 0.56 | 0.88 | ~0.8a ‘20_[ 0.86 [0.44 [0.88 | 0.61] 0.89 | 0.68 | 091 | 0.85 | 0.93 082 ‘30_|088_| 06a] —0.90_| 0.40 | 092 | 0.86 | 093 | 0.53 | 0.94 | 00. ‘40_[ 090 | 0.43 | 0.92 | 059 | —0.94|—085_| 094 | 0.52 | 0.96 | 04a 60_| 092_|0.43| 0.98 | 0.59 | 094 085 | 095 | 0.52 | 0.96 | aa7 ‘60_| 092 | 0.42 | 0.98 | 0.58 | 0.95 | 0.84. O51 097 | 046 70_|092_| 0.62 | os | 0.58 | 096 16a ‘080__[ 097 | 0.46 a0 | 093 | 0.62 | 0.98 | 057 | 0.96 | 0.83. 049097 | 0.45 90_|093_| 0.42 | 094 | 0.57 | 0.96 | 083 ‘049097 —|—0.5 yoo] 094 [0.61096 | 087 | 0.96 | 052 048098 | 04a ‘Gatamne 4, 6,8 end TO sow the variation between (AP/PQ ce and K {Galarmna J; 7,9 end 11 show variation betweon Ya and ‘We can remove pressures and veloc- using thermodynamie ratios. ‘The math is rather involved, so it is condensed in a box on p. 65. Substi- critical ratios from Equations (44) into Equation (15) to ar- rive at Equation (16). ities by tute the (43) and 2 Mi) saa a pal treme fy IME } =} a6) +K=0 Useful eritical ratio Equati gineers n (16) represents the flow sta- ‘ius during endpoint choking between Node 1 and Node 2, and as such, itis able to provide a solution for part of the eritical flow problem, This equa- tion is a complex function of two vari- ables, M; and K. For application’, en- correlating botwoon Nodo 0 and Node 2. We need ‘more manipulation to have a solution are happier that estimates (P2/ Polen. ‘There is a way to estimate the pres- sure ratio botwoon Nodo 1 and Node 2. Multiply Equation (32) by Equation (44) and thus come up with a correla tion for (P3/Po}ey: Equation (17). () ‘The simultaneous solution of Equa- tions (16) and (17) eliminates M and yields a value for r, the critical pres- ‘sure ratio, We can now determine criti cal low in ether ofthe following ways: * Determine the velocity of sound at [ost-+naca]* : i) = freosty-aimt]®-) ny the pipe's endpoint. This is equiva: Tent to the lino’s flow velocity. Now estimate flowrate based on pipe di- ameter. This approach assumes adi abatie expansion. Use Equation (45) Or, preferably, use the isothermal correlation [6], which is conserva- | | volume, Vo tive, and usually gives a flowrate | [entroplo coomiclent, 7] 1 20-80% higher than the adiabatic | [inner dia ef pipe, D-— [49.2 7am calculation Ifthe flow happens to be subcritical, | | Number of ows 1 (as determined by the preceding two | Resisfanee coeffcion? K| steps) use the modified Darcy for- | | (a8 estimated for above | 45 ‘mula with appropriate Yon. Skip to the last part of this article for fur- ther diseus ‘TABLE 2, PROBLEM SETUP Tnput tem Value Upstream pressure, Pp_| 6,600 KPag| ‘Downsfroam discharge Eecaraen 200 kPaa Upsiream specilic 11/58 kg/m Tength of pipe. TOO m im pipe data) ginally higher. The reason for the de- Equations (16) and (17) are computer- | viation in results is attributed to the ized in the spreadsheot that was | limitation of maximum ‘promised at the beginning ofthis article. | Crane method, whereas, for the ac- It is provided as a download at | tual gas y= 1.55. ‘www.che.com(CEEXTRA, .4-in the In addition to the spreadsheet pro- ‘The derivation of ¥en is shown in | vided, we present Table 1 for a large Node 0 ssheot method, as described determine the value of Yor Putting the results to work We present an example for doing | —Hereistheexample.Bstimatethema- choked-‘low calculations. Equation (17) | terial loss for a tube-rupture incident in the box on p. 68. The ostimation of | range of conditions so that readers ean ‘Yon ean be accomplished with the fol- | verify the accuracy ofthe program after lowing caleulation steps. ‘* Estimate K for the pipes, and + at | variation in the eritical values with re- downloading. Table 1 indicates a wide spect to- and K. Crane publishes a sim- ‘+ Determiner using the Excel spread- | ilar table in Crane A-22 [4]. Figure 2 shows the variation between * Substitute r into Equation (62) to | (AP/Pglop, and K. Figure 3 shows variation between Yo, y and K. Both of the charis indicate a reasonable varia- tion with respect to isentropic factor used to determine the maximum | adouble-tube gasheater. The gas heater critical (choked) flow as well as to esti- | is receiving gas from a 102-mm main rate the critical expansion factor. gas-line via a 49-mm branch line, ‘The We solved the example by two | length of branch line is approximately 64 CHEMICALENGINEERING WAW/CHE.COM OCTOBER 20c2 methods, the Crane plot-correlation, | 100 m. The upstream conditions of gas and the spreadsheet. The flowrates | and other relevant data are in Table 2 from Crane plot-correlation are mar- | Tho estimate of K'=45 comes out of theFIGURE 2. Tho critical pressure difference is influenced by ¥, ‘especially at modest resistance coefficients FIGURE 3. Tho critical expansion factor is influenced by ‘especially at low resistance coefficients, ts « bit complicated to move from Mach Nimber 1 the general enone Row rans But it is necessary, so here ore the sep. “The Mock Number in ow is dated os 4 ff between ctclv.ociy of gos othe velocity of sat at iretd conditions evling het portal point (31 feel of sound no gor mediums: 1.90 (20 ‘Thus, the Mach Number is; Meee Te en There is no motion at a point. Eliminate frelon fom Equation {I} od reve os Equsion (22). In hernedynamic. tat ‘books, this is the fundamental energy bal- ‘ance for frictionless isentropic Flow. VaP + 0.5au? = 0 (22) Integrate from general Point to generol Point 2: 2 2 dp 0.5} udo =0 (23) The isentropic process has a wellknown relationship [7 jvep = (ave FM) (24) Subsitule Equation (24) into. Equotion (23); integrate the second term, and write Equation (25) x (ef). (25) Failte-*M) ig 4)-0 ¢ Because of he zara, itis possible to convert Equation (25) inlo a general, fundementa, Ficonless, is correlation 2 ipyy+ (26) Son Nee eee on 26) he poet Figure SPtedhspeecced, (7o)F = Tavyash of 71 y There are two velocities: v= 0 and (uy? = yMj2°)¥) rom Equation ([20}). Peg ‘hase nto Equation (26) APY, Ftee)= (e+ on) Rebrenge ond sine PM oe Poo, (r- WMP (29) From Equation (8), the fundamental aia baie equation, P;V)7 = PV [5] ean be verilen in he form of Equation (30) v(t) vo UF, (20) Substituting the above into Equation (29): Ayr eee recan mena) Sole for Pi /Po ii 2 fe t-legaal ed Between Node 1 and Node 2 ‘ppl eaucon (26) oe pe en. The Dee eae a A Palm = OM) S (33) ape eae da Ae Bele a: a Replace tho veloctos inside the brackets with Mach Numbers as derived from Equa- fon 2 uf, way, (35) (36) page rane ”'| feces Equose (37), ome (38h ae te eae _ [eo ma ne 2 Yee IMF (38) Ober rots, sch os Vi/¥ end P/F can be worked out with some basic thermo: -dynomic knowledge [51 Tn pipe flow: (U/V)= constant (39) vy. Jeeta ao) Va f2-(r =F} In adiabatic ow: ai ares, ai oc l P [2+(y— Meme (42) If the flow is choked, then there is sonic flow atthe pipe outlet In short, Mo = 1. This Canals spel oof Epon (8 (40) rd (2) ol: [ (noe (aCe -yzra-me (43) 5 (44) Fee eiearelon tees CHEMICALENGINEERING WWHW.CHE.COM OCTOBER 2002nce the erica pressure ratio is known, 0: mass flowrate at choking con be esimated ‘Adiabatic approach: The gas veloc ofthe endpoint ofthe pipe isthe some os the velo of sound at end point conditions IF we know the inner di- ‘ameter ofthe pipe iti possible to trans form the end point conditions to upstream conditions This isthe adiabatic approach, nnd we vse appropriate units and he cxit= cal pressure ratio To davaop this corel Hon , Equation (45). W=0. 2004 er (45) Isothermal approach: However, the ue offothermal ow equa fin is more conservative, and yields cbout 30% more fw than does the osicbatic ined con cx rsedinborin body ofthe are, the ow says somewhere between isctermal and isentropic. There ton fenton ons fre The mew fowrlefr expanding fow fs Sen lig el 3 ich ve vation (4) The og ol hs expan areas areata eenaoaaet onsontross secon met pes, Ao ae) \ K+2H{"%]h o Converting this equation to convenient (rretrie) units, subsisting fr area and den- Sily (as defined in the nomenciature), and we OVERALL CRITICAL EXPANSION FACTOR, Ycr USING THE SPREADSHEET rearranging the terms yields Equation (47). Wwe od2esD"» a7) jot Sa OIE ok conan wih Eguaton (Fined os Ppa afi wore Equation (49) is Darey formula [1], i tote w orz650% V0 ‘Substitute P2/Po = rand cearrange Equa on (49) (49) |Polt=r) | KV (50) Equation (48) equols Equation (50). Yor [~ x(-7) W = 0,1265D%eq (51) frm (1-1) from the no rmerator and denominator. We got Equa tion (52 fr critical expansion factor. (52) NOTE: For flows through an orifice, the ex: pansion fotos provided in [2). Crane book [6]. The Moody Friction Fae- tor is 0.02 for 100,000 mm of pipe to- ‘gether with d standard elhows and exit ‘and entrance losses, Solution via Excel spreadsheet Solve Equations (16) and (17) simulta- neously: r= (Po/Pp)op = 0.048 Hand calculation: (AP/ Palo = 0.952 From Part 6 of spreadshoct: XP, = 0,018, X 6,600 = 317 kPaa Since P» > Pa, the pipe will choke. ‘Therefore, for flow estimation, AP will be (6,600 ~ 817) = 6,288 kPa, Using Equation (49), critical mass flow, W= 18,724 kg/h Note that we alternatively ean use Equation (48) to estimate the critical mass flow, by applying the caleu- lated value of r directly. The for- mula-based correlation provides an opportunity to input actual values of ‘yand K. We achieve a less-consorva- tive and more-preeise flowrate. Also, the Excel model is alot quicker. Solution by Crane method: By using the plot-correlation as given in Rofarence (f] (y= 14, and = 45) we get tho following critical values after interpolation betwoon K = 40 and K = 100: (APP = 0.885 Yo = 0.72 ‘The critical prossure, Pp = 759 kPaa Since Py > P, the pipe will choke ‘Therefore, for flow estimation, AP will be (6,600 ~ 759) = 5,841 kPa Use Equation (48) with these data Critical mass flow, W = 19,130 kg/h ‘The plot-based correlation presents limitations, as the data wore outside the graphed range. K had to be ex- trapolated and only y= 1.3 and 1.4 are plotted. ‘The extrapolation process, and referring the values from plots, are both time consuming. 66 CHENICAL ENGINEERING ‘WWW.CHECOM OCTOBER 2002 /n general, the Excel spreadsheet solves comple equation whore ikon edi oie. toe (16) and (17) are non-linear and simul- Yaneous equations wih two unknown variables: M; and r. Therefore, in order to determine the value of r, both equa tions must be solved simultaneously, to. eliminate My Here are tho sleps inside the Excel readshee! method. We have appended [program in www.che.com/CEEXTRA with three notational changes. While working in Excel, fe tes been ro- ‘ced with My wih al P/Po th 4 Dafne Nones Kg ory) end for M)) in separate cells that refer to their respoctve volves, Nol: Use Nome/De- fine command on Insert menu. 2.Print the following formula in a cell which refers to the Left hond side of Equation (14) and returns the value to ‘zero via “Goal seek” command as ex: ploined below. (WA 2*(ge)ae)* MA2/(244g~1)"MA2})((g# 1/21) “+2°UNISQRTI(2Hg-1)"_ Me2yiget MOSH 13. Print the following formula in a cell that ‘refers to the right side of Equation (17) ond returns the value of Po/Pp, slgetMA2/2/%/2V 1g MA2/2INg"g1)/(2*4a-1)) ‘4.Use "Goal seek’ command {from Tools menu) to calculate M, by having the M-value to change so as fo set the value ofthe lef hand side of Equation (1) zero 5. Print the solved values for rand M. 3 nee a ‘can be doter- mined by simple frm inserted in ‘spreadsheet (or done on a desk calcu lator). The spreadsheet formula for Yee or the right -hand side of Equa: tion (52) is as follows: =SQRTIK(141}/(2*(Ke2°INT/i Suberitical flow Graphical methods can give a eue how to deal with subcritical flow. The sub- critieal values for ¥ean be determined by nearly interpolating between (Y= 1, AP/Pp =0) and (Yop, {4P/Polog). If ¥and (AP/Polop ae potted ony and.z axis respectively, a linear equation to determine intermediate values for Y can be written as Equation (18). ral as) cas)Erne ees Solved via spreadsheet. The prob: Jem is the same one as above with one change, The downstream pressure is 8,000 kPaa and not 200 kPaa, which is a significant differenee, r= (PelPolcy (aP/Polex Yor ‘Tho critical pressure, P = 317 kPaa Sinco P < Pa, the pipe will not choke. Therefore, AP will be (6,600 — 3,000) kPaa. = 3,600 kPa Sub-crtieal (AP/Po)su1s = 8,600/6,600 = 0.545 Interpolate linearly between (¥=1, (AP/Pa)ep = 0.952). Suberitieal ¥= 0.816 ‘Use Equation (49) for flow, W= 17,028 kg’ ‘Solved by Crane method: By using the plot-correlation in Refer- ence [dl, (y= 1-4, and K= 45) we get the following critical values after interpo- lation between K= 40 and K= (AP/Pplog = 0.885 Yop = 0.72 ‘The critical pressure, P; = 759 kPaa Since Pp < P4, the pipe will not choke. Therefore, AP will be (6,600 — 8,000) kPaa = 3,600 kPa Suberitical “ (AP/Po)srm = AP/Po = 0) and (Yon = 0.68, | 9,606,600 = 0.545 References & Shap 1 Grane Cop “Plow of Fide” Publication | 5 Smith JM. ond Van Nets HC. nto. ‘Ee, Eqeation Iti, Jol action Cemial nina hon SOM ft,” Bazan Toth dai Mh, | ier aah a See Band Mi hades 2. Mil, RW, “lon Mensarementngnger dpa nti apo Euston 15 Me kN ert = 3: Kem, RF o Sze Piping as Gas Expands Fl Gants, er Bie OS BIBI Fa Se 8 Gann pc Stn Eine 3.4.98. Ganon die set Rebar, de Fy ‘Gra gang OE 2 {rary Buttorwt Halnemans Interpolating linearly between (et, AP/Po = 0) and Yon = 0.72, AP/Poon = 0.545), or read diveetly Reference (4, Suberitical ¥ = 0.82 Using Equation (49) for flowrate ‘The mass flow, W = 17,104 kg/h. This is a close check. Obviously, the graph- feal correlation deviates less in sub- critical applications. . Eaited by Peter M, Silverberg Author Sani Kumar is goajor loss ginger Worley TTerrce, Porth, Wastarn 10; Phone. BL AD secre smiey ‘a and gas production pla frm. pettloum refineries, cecal glne dg pes jchial. He has « Mfeah. In chanel engl ‘neering fom Horeort Butler Technol tater He ea fewrs (Avera petrochemical AG aS New Filtration Technology Keeps Cooling Tower Water Clean Typical Full Flow fitration using existing pump. Typical Side Stream filtration using a booster pump. Typical Side Stream filtration of basin using a recirculating pump. Typical Side Stream filtration using existing pump. www.orival.com » [email protected] very cooling tower bas its unique dit E crrciies soca conan wl er special characteristics. Filtration sjstems must take these into account in onder to ‘maintain particle foe water rival filters are versatile enough 0 meet these criteria. They remove a wide range of dirt down to micron size, of any specific gravity, even lighter than water ‘Typical examples include: airbome parti- cles, microbiological growth, bugs, sand, scale, alga, est, tc Line pressure powered, they permit use of the existing cooling tower pump, lowering intial sytem cost. The fully a tomatic self cleaning cycle takes seconds and does not interrupt flow: Reclaim units recireulae rinse water mak: ing filtration a zero discharge operation. For ‘more information and system design assistance, contact: Orval, Aucmatic SeifCleming Filters, 40 N. Van Brune St., Bglewood, NT (07631. (201) 568-3311, (800) 567-9767, Fax: (201) 568-1916. CCHEVICAL ENGINEERING WW”W.CHE.COM OCTOBER 2000 67Engineering Practice Unraveling the Myriad of Models Used in Modern Automation Projects Keeping track of the models and their I/Os will avoid confusion and ensure proper implementation Greg Martin, Mark Funderburk and Michael Sugars Pavilion Technologies, Inc. ontinuous proeess automation has evolved in the process indus- tries from single, model-based property predictors that support single-loop control applications, to rulti-unit optimization methods that provides targets to multiple linoar and nonlinear model-predictive control (MPC) applications, Models are the foundation of these control and opti- ization solutions. ‘There may be as many as six differ- cent types of models in one solution, in luding models in the optimizer, the controller, the supporting property predictors, and the forward chaining logic. This multiplicity of models ean cause confusion in discussion and ‘even occasionally in application, ‘This article describes the different types of models and their purposes, and diseusses how they are imple- ‘mented, and how they relate to the ‘other models in tho overall applica- tion, References are provided that pro- vvide details for the particular models Aeseribed, Introduction ‘An automation project in the process industries may include several types of applications, each containing a modet of the process or a part of it, The mod- cls in single project may include: * Property predictors * MPC * Nonlinear MPC gains * Forward chaining * Process optimization * Unit optimization ‘ Multi-unit optimization ach model is characterized by its in- puts and outputs (VO), and the YOs predictor | —>[ PO predicier | —>[NPG dt | | po |[Pe [Fo presale Process Process FIGURE 4. Schematic ofa property predictor witha PID controler for one model are noceccarily related to the 0s of another model. There is often confusion about these various ‘models in terms of what they do and ‘what are the proper /Os for a particu lar model. Furthermore, since these ‘models are implemented as parts of a cascade in the overall application, some inputs may be outputs of a super- visory application. Tt is important to ‘understand where each model fits into the overall heirarchy of the eascade. Property predictors Online correlations for laboratory ‘analyses, based on neural network models, wore introduced in 1996 as “soft sensors” {Z]. Since then, these types of models have beeome Iknown a8 virtual online enalyzors or property predictors. Applications range from petroleum refining to food processing [2], Property predictors have also been extended to include physical models for distillation and the switching of different crude feedstocks [3,41 A property-predictor signal-flow dia- sgramisshownin Figure | Several mea- surements from the process are used to caleulate one output, usually a product G8 CHEMICALENGINEERING WWW.CHECOM OCTOBER 2002 FIGURE 2. Introducing model-predictive ‘control (MPC) toa property predictor analysis, Ifa neural notwork model is ‘used, a fow months of process and labo- ratory values are required to train the ‘model. The model is implemented in an online runtime packago whereby it ean provide periodie output tothe program- mable-logic controller (PLC) or distrib uted control system (DCS). For the ex- ample shown in Figure 1, the online prediction provides feedback to a con: ventional proportionalintegral-deriv- ative (PID) controller, Model predictive control A property predictor providing feed- back to an MPC is shown in Figure 2. ‘This is a multivariable MPC applica- tion, which has three manipulated variables (MVs). Each MV is the set- point (SP) of a PID controller. "The property predictorhas.a model in it, and the MPC has a model init. Those models are quite different, however, as the property predictor model is a stoady-stato relationship (used to pre- dicta property) and the MPC model isa matrix of step-response-dynamic mod. els (used to predict the CV response). ‘The UOs of these models are re lated. Inputs to the property predictor