Scribd Logo
Unggah

Selamat datang di Scribd!

  • Carlson Properties Article

    Diunggah oleh

    Hyder Murtaza
    8 tayangan12 halaman
    Carlson Properties Article

    Hak Cipta

    © © All Rights Reserved

    Format Tersedia

    PDF atau baca online dari Scribd

    Apakah menurut Anda dokumen ini bermanfaat?

    Apakah konten ini tidak pantas?

    Laporkan Dokumen Ini
    Carlson Properties Article

    Hak Cipta:

    © All Rights Reserved
    Unduh sebagai PDF atau baca online dari Scribd
    Tandai sebagai konten tidak pantas
    8 tayangan12 halaman

    Carlson Properties Article

    Diunggah oleh

    Hyder Murtaza
    Carlson Properties Article

    Hak Cipta:

    © All Rights Reserved
    Unduh sebagai PDF atau baca online dari Scribd
    Tandai sebagai konten tidak pantas
    dari 12
    Pe Don’t Gamble With Physical Properties For Simulations Finding good values ‘for inadequate or missing physical property parameters is the key toa successful simulation. And this depends upon choosing the right estimation methods. Eric €. Cartson, Aspen Technalogy, fe hhemical engineers use process simulation to perform a variety ‘of important work. This work ranges from calculations of ‘mass- and energy balances of large flow. sheets to prediction of the performance of process alternatives that can save millions of dollars. An engineer very quickly can define a complex flowsheet and all the process conditions. Desktop computers ‘now allow rating, sizing, optimization, and dynamic calculations that previously required large mainframe computers. In the past, these simulations were often built by a geoup of experts, including @ physical property expert. Now, simulators such as ASPEN PLUS, ChemCAD Ill, HYSIM, PRO Il, and SPEEDUP are easi- er to use and more powerful than the standalone programs of the past. Today, a single engineer can set up the basic simu: lation specifications, including the physi: cal properties, in very litle time. Missing of inadequate physical prop- erties, however, can undermine the accu racy of a model or even prevent you from performing the simulation. That some re quired information is missing is not an ht in the simulator. After all, for ‘most compounds, physical property pa rameters ate not known for every thermo- dynamic model or for all temperature or pressure ranges. Models have built-in as- ssurmptions and practical limits that should apply In this article, we will provide practi cal tips and techniques to help you accu- rately describe the physical properties needed in a simulation. AS an engineer, you. always will have to make assump: tions in terms of physical properties, however. The goal ofthis article isto out line the appropriate assumptions and 10 provide techniques when properties are The five important tasks Successfully describing the physical properties to be used in a simulation in volves five tasks: L.selecting the appropriate physical property methods: 2. validating the physical properties; 3.describing nondatabank compo- nents (chemical species or compound) and missing parameters 4, obtaining and using phy prop: missing property eters. Tc can be argued that these tasks are rot sequential and, to some degree, they are concurrent, During simulation devel ‘opment, however, you will need to visit each area to be confident that your simu lation is a3 accurate as possible — so that important decisions can be made based on the results of your simulations. Selecting the appropriate physical property methods This essential first step will affect all subsequent tasks in developing accurate physical properties in your simulation Indeed, the choice of the physical proper: ty models for a simulation can be one of the most important decisions for an engi: net. Several factors need to be consid: CHEMICAL ENGINEERING PROGRESS + OCTOBER 195 + 36 | SUCCEEDING AT SiMULAT nd no si dle all systems. Tabl method can han. 1 tists. some thermodynamic models available in simulators The four factors that you should consider when choosing property methods are: ‘the nature of the properties of interest; + the composition of the mixture; pressure and temperature the availability of parameters, To ease the selection of the right physical property methods, we sug. ‘est using the decision tees shown in Figures 1-3. These trees are based on the four factors for selecting property methods, and can be used when the chemical components and approxi- ‘mate temperature and pressure rang; are known, While these diagrams are simplifications, they do show the basic steps of the decision-making process, while the notes in the sidebar amplify some of the Key points The nature of the properties of ix terest. A question that you may ask ‘yourself when starting a simulation is, “Does the choice of physical property methods matter?” The answer is an emphatic YES. The choice can strongly affect the prediction of the simulation. You should be select collection of methods that will best predict the properties or results of in terest to you Because many chemical process simulations include distillation, strip- ping, or evaporation, one important potential consideration for the choice of physical property models is vepor/iquid equilibrium (VLE). This is the area in which the most physical property work is focused in chemical engineering. Liquid/liquid equilibri um (LLE) also becomes important in processes such as solvent extraction and extractive distillation, Another critical consideration is pure-component and mixture en- thalpy. Enthalpies and heat eapaciti are important for unit operations such as heat exchangers, condensers, dis- tillstion columns, and reactors, Pe Pusat entre Equation-of State Models ‘Activity Coeticient Models Bene We-RubilB\WR)L09 Stating lecvayte NATL Hayden-0\Connel™ Flory Messing yeoger-ivride equation af stat for aT hexamerzetion® Scatchard Hida Idea! gas aw" | untguac Lee-Kesler (UK) Uunirac Lee Kesler Plocker Van tae ong-Robinson (PR) Wison Penurbed-Hard-Chain Pragictve SAK Special Models Redich-Kwang (RK) BPI sourwator method Redich-Kwang-Sosve RKS} Braun K-10 KS or PR ith Wong Sander mixing ele Chao-Seader KS or PR wth mosies-HuronNidal-2 mic GraysonStreed ingle Kent Eisenberg Sanchez Lacombe for polymers Steam Ties * Not use forthe iui pase See Figure? LL eecanee Eley NATL er piner Real Lee Kesler Poo I Bit reg ation | Radich Kong Soave, Grayson Streed or | | Pseudo & [Creal HY Polarity <> Pn UW Figure 1. The fiest steps for selecting physical property methods. 36 + ocTOBER 856 » cHENNCAL EWGIEERING PROGRESS Braunk-10 Braun K-10 or Iden! eve ae pear as Fe What: solar mole fang. the own coe property tating po peeudecor laf Wy 2 tompanen smlators phase oat any a9 ction, a ingoutot iy of gase os at reel of Wh paar com Stared fiends than 108 tity coet ideat iu otstate wh tem pri. or editve rnethods. speciale wierse forts in cess Gist ccetcen Rt sin ‘vuauac or fe bine cen feerttr though parame Wher tity coe stoupe te Aout of uae festa vue: a Soave kor oe Navigating the decision trees Hove are some pointers to lp you navigate the decision tees that ap pear as Figuras 1-3 lWhatare pseudocamponents?ia many appicatans where ony nan polar molecules are present [such asin hysrocarbon processing andr: Fring). the mistur isso complex thatinstead of rpcesenting by ane navn constituents, tis easier to grove the constituents by some use propery such 3s boing point. In this way, amature of hundreds of eon: Sttuens can be reduced to 30 or fener. The properties of hese grouped constitu, called psoudacomponents, ae representa by an average iing point, spectic gray, and molecular weigh. Ifyou do not use ssudocomponents, the constiuents shouldbe dasenbed oy a moles Tarformul and are eletes tas ea components Why are elecroe macures dierent? Becohte micues include camponents that are charged molecules an) or tat form sats Some simulators allow calculation of eleckotrte reaction aquirium wih phase equilibrium. This is avery powerful metho and is usage covers ‘many applications such 9s caustic serbing, eutalaation, acd pro faction, and sat precipitation. The nondealty of elecvolte solutions ‘sally cortaing water, canbe observe in boing pon elvation, sl ing ou of gases thats, ing sls othe saluton to change te elu iyo gases), and sat preciptation, The most common elecroye met fds are the Pier model, and the modied-MATL actigy costicent Inadel of Chen and coworkers. Som electolas, tke formic acis and este acid, are very weak and an alata method nat egured Wich type of mathod should be chosen for mitures containing pala components bu no electaltes? There are tw0 groups of methods ‘— based on activity coefficients ar equations af state Use actiuty-cot ficient based methods when presses are low to medium pial less than 10 bar oF 150 silane ina components ara near critical pint. Ac tty costcient models aso ofen ae used fo accurately predict non ‘goa uid behavior such a2 for VLE and for LE. In const, equation state methods excel in thir ability 10 repraset data and axtraplae With temperature and pressure up to and above the mixure eres! sit Now, however, methods relying on cubic equations of stato wth predictive mixing cules effectively combine the svenghs of the two methods (Sea Table 2) For higher pressures (and temperatures, hese special equations of state are beter a thay wore developed to aay 0 wider range of temperatures These methods incorporate acity cost Fens in the calculation of component interactions represented by ax 855 Gis roe anergy. Mos af the ater use a UNIFAC-based acta toeficient model a the default, but you can use any activity cotfiient, [AL simulation pressures less than 10 atm and whare there are no ear citeal eamonents, fr the est eels use the Wilson, NATL, | UwQUAc binary parameters that may be avaible in buitn databanks, tr fit brary parameters to expermentl data | avalable) using actaty Cceticent node. These parameter may have bean determines at fecat temperatures, pressures and compesons than you are siulaing though, so you may aot obi the Best posible accuracy I ineractan parameters ae not sual, however you can use the UNIFAC method Wen should UNIEAC be used? UNIERC and ahar UNIFAC: base ae tty coeticient models ae precicive apgoaches that use stuctral 049s a estimate component interacvons, Frm stuctral information tout organic components usually avalable in the buitin databank, LUMEAC sable to pedir the actity cootcionts asa function of com Bastion and tmperatre You ean make ue of UNIFAC sehen you do aot have experimental data or binary parameters or when an approximate ‘ale is acceptable fr instance, fra campanant wth ow priarty.n| recem years, thee have been improvements to UNIFAC (see Table 3) thatcanbete pradet VLE, eat ot ming, and LLE ove a wider tamper sure range. Recent extensions ta UNIFAC propose for molecules such 35 reigerants and sugars may be ust and you can add the groups {and parameters to your simulation. Simulators may Rave the ay to generate binary interaction parameters for Wilson, UNIQUAC, of NATL from UMFAC. Not ai components can be described using UNIFAG, however, and nat allgoug interactions ae avaale Earpes of components that not have UNFAC groups incase metals, orgasametas, and pharphates So, we tight recommend abvays doing a search fo avaiable data on ‘nary o teary systems of interest. Hon should th vapor phase be treated? The choice ofthe VLE rmothod using an actiaty eoaticent modal alsa requires a chose of ‘model for the vapor ahase properties vapor phate association is ob served (asin the case of acetic acid), then the vapor phase model should be Hayden-O'Connell or Nothnage.& system containing hyde igen fuonde may require a special modelo reresor tho igh degree ot {ssaiation due to hydrogen bonding. Assaciatan in the vapor phase an havea strong effect on phase equi and enthalpy. When shoud defauts be overrtéen for otter physica property methods? Friction of éeasiy, enthalpy, and viscosty alsa are impo. ‘ant in simlatrs, and you shold automatically accept te detaut iethods Check the simulator documentation far te deta mettod and ising rts, Vapor density is calculated by an equation of state o the ideal gas law Mitre liquid densities ean be calculated by an etuation asta, temperate dapendent model suchas that of Racker, or by 8 tener ture: and pessure-¢ependent model such a the COSTALD. For psuedo ‘components, an American Patou lst (API seth typically is ‘employed. The Racket model is ecarmmanded for general ue. ‘Vapor enthalpy usualy is ealuled vie an ideal gas assumption or se equation of state The equation otstata metads calculate 9 cepa: ‘ure from ideality eae the vapor enthalpy depetare For companents such ae acetic acd the Hayden-0'Connell modes best and al ale Inte a lacgeethan-noumal vapor enthalpy departure {Uguid enthalpies are caleulated bya variety of methods. the sna lato uses tho ieal gos asthe reference stat, then te pure-conponent gui enthalpy is calalated fom the ideal g3s enthalpy and lui an thalpy departure. Tis cane writen os Hele Ws efit Hr) a lahore Hts the pure-camonent lid enthalpy, Wve tha ideal gas a thay, and A) Ws the guid enthaly deparure. Ths depamue ‘loaes me heat of vapszaton the vapor enthalpy departure frm ne ices pressure tothe saturation pressure, andthe quid pressure ca: rection from the saturation pressure tothe seal grossure, Simulators also alow separate calculations fora igud enthalpy dee tom the liqus-neat-eapacitypolima. For some components, the method in qt wll not be accurate enough for guid heat capacity peedictane Tis canbe very important you are exporting your progeny internation to another goatam such 9s oe lor rigorous hast exchanger ign You an use the later iqus-nest capacity (CpL) method to improve the ae: uracy of hui heat capacies. \svosty is another important propery for sang of siping. pugs, heat exchangers, and dictlavon columns, There are vanaus vapor and wu methods for calculating viscasty and ganeraly, he parameter uramente fr thse matods are subetanbal ‘CHEMICAL ENGINEERING PROGRESS + ocTORER =95 + 37 a aaa aS SUCCEEDING AT SIMULATION | Yes. nan. umauae, Hie Table 2/Exaniples of special | iam | and therrverances | ER CET Nec [ ver | Prestive SAK SK) Be pf PR wth modiied Huror-Vidal-2 mixing ele 0 Phe Pape ng a | | uo __ WLSON, NATL UNIOUR PR ith Wong Soler mixing ola 8 | renew A no Undies Vomneae "| Rts tnmotied Reon [ ibeao a viracue | ining le » Boe’ o | sxS win uooupls wn se | No KS wth Worg Sandler mung rule | @ | 8 sla | yg Uma ane as Rees ® ves Exensons usc Seer PROTA ca a —— : tose orm sted | we. Wea Vapor Phase Association | the vapor phase also can be significant pegreesof Palmerston if the components form complexes > The ort memotes es] | | ate eleciostatic, induction, attraction and repulsion between nonpolar com ponenis, and chemical forces such ® |g Figg ae hydrogen bonding. A good overvie™ | rig, 1 Figure 3. Options or vapor-phase calculations with acivin-coeficien models. Of these forces i given in Ref. 1 Source (7 | 38 + ocroser 195 + chencaL ENGINEERING PROGRESS | | Vapor Maletracion © tiqua Moletretion “Temperature, C a Molelaction Actonitile b Sou 8 Figure 4 (above). VLE of acetonirile/water system at 1 atm. 1M Figure 5 (right). VLE of toluene/phenol system at atm. ‘The magnitude of the electrostatic and induction forces is related to the mx polarity of the components. Compo- nents such as water, acetone, formallehyde, and methyl chloride are strong dipoles. Many polar com- pounds are associative, and form complexes or dissociate into ions. Components like ethane and n-hep- tane are nonpolar. You can use your polar components, ‘Vapor Molefraction Toluene simulator to report the dipole mo- 5 of databank components as one measurement of polarity. In g of nonpolar will exhibit less nonideal behavior Figures 4-7 illustrate the effect of polarity on binary vaporliquid equi libria. Figure 4 shows the predicted and experimental VLE of to hi acetonitrile and ‘components 1 | | — | water, at I atm, The azeotrope is ac- curately predicted at approximately sneral, 0.7 mole fraction of acetonitrile. Fig ure 5 presents VLE for a mixture of two slightly polar compounds, toluene and phenol, at 1 atm, The de- viation from ideality is shown by compari edited curve from an ideal liguid assumption to that ‘od predicting nonideality OOS Liquid Moletraction Toluene € é i oor 02 03 04 05 06 07 08.084 Motetraction s Source: (19 Figure 6 (above). VLE of exclohexanelbenzene system at I atm W Figure 7 (right). VEE of ethanepropylene system at 40°F. Tota Pressure, Atm (CHEMICAL ENGINEERING PROGRESS + OCTOBER 1095 + 39 SUCCEEDING AT SIMULATION © Liquid? Moetaction Ct It << "Qos. 01 015 02 025 03 035 04 045 05 05508 (the nonrandom two-liquid activity coefficient. model (NRTL) and Redlich-Kwong equation of state for the vapor phase), Figure 6 depicts th VLE of a mixture of cyclohexane and benzene at 1 atm. Here, the interac tion of seemingly similar molecules with a difference in boiling point of less than 1°C causes an azeotrope at a composition of about 0.54 mole frac tion of benzene, A mixture such as ethane and propylene (Figure 7) is an almost ideal one, and does nat deviate ‘much from Raoult law. Mixtures of nonpolar and polar compounds, such as Water and hydro- carbons, often will form two liquid phases that are very immiscible. Fig- tures 8 and 9 show examples of misci- ble and immiscible systems of lig- Luidiquid equilibria, respectively, at 1 atm. In Figure 8, cyclohexanol is im- miscible in the water phase but the organic phase contains up to 0.50 mole fraction water (0.10 mass frac tion water), Figure 9 shows the high degree of immiscibility in both the organic and water phases for a mix- ture of benzene and water where there is less than 0.06% by mole ben zene (0.3% by mass). Because of this behavior, some simulators have a cial property method to treat the | water phase as organic-free (also called Free-Water) Most simulators offer collections of property methods in. predefined Souree: (16 Wi Figure 8 (ef). LLE of eyelohexanolwater system at I atm. W Figure 9. (above). LLE of benzenefwater system at 1 ai. sets based upon methods that fre- quently are used for certain types of mixtures. Usually the sets are identi fied by the method used for phase equilibria. When these sets use an equation-of-state model, the same model is used for many properties, in- cluding those for phase equilibria, The pressure and temperature range. This is especially important in choosing the method t perform phase equilibria calculations. Meth- ods that are based on Raoull’s law or that use activity coefficients are not accurate at high pressure or when th temperature is above the critical tem: perature of a component. You can use Henry's Taw when you have light gases in subcritical solvents, but it generally is not recommended for Concentrations of solute greater than 5%, In general, equations of state are better suited to predict VLE over a ‘wide temperature or pressure range, especially at high temperature and pressure The availability of parameters Without sufficient pure-component and binary parameters, you will ‘unable to calculate pure-component or mixture properties, You must choose among obtaining and using experi- mental or literature data, estimati jeters, ot choosing a less 1igo ‘ous method. This should be investiga ed forall physical property methods in cluding those shown in Figures 1-3. 40 + octosER 136 + cHeMiCAL ENGINEERING PROGRESS Validating the physical properties A necessary step in any simulation project is validation of the physical properties. This involves reporting, tab- ‘lating, or plotting pare-component and ‘mixture properties and comparing the results to known data or expected be- havior. This is an important step in any simulation and should be performed for databank as well as nondatabank com- ponents, Simulators can provide these caleulated properties in tabular and plot format, This is a useful tool for under- standing bow pure-component and mix ture properties, such as density, heat ca pacity, and excess properties, vary with temperature, pressure, and composition, and how they behave when extrapolat- ed Similarly, such results can be used '0 ‘generate plots of VLE and LLE to com- pare to diagrams in the literature and 2 tual field data. Some simulators have the capability to generate residue curves for distillation of ternary mixtures. The residue plot capability also is a powerful ‘ool for distillation analysis, Use the tabulation and plotting tools to determine the cause of dis- crepancies in properties, If a mixture property is incorrect, investigate if @ single component is the cause by 1 porting pure-component properties ‘Another useful technique is 0 com ppare the same flowsheet table results while using different physical property methods pos ey ide ed pre ni mm ha Ph da toy da pr ju ai > By default, most phase equilibria create a list of parameters that are check for different ordering of atoms. calculations are performed assuming missing. You should detail this infor- For instance, ammonia can be de > ‘por and liquid phases. If your pro- mation when communicating the as- scribed as HN instead of NH, Ret. 2 fess involves two liquid phases sumptions of the simulation to other contains a formula index of organic (WLLE), be sure to specify three- users or your management. compounds and is a good resource for phase calculations. If not, you will Certain property parameters always alternative names. fet incorrect results. As a part of the are requited for a simulation. These Once you have determined the px Salidation, you also should check that can include molecular weight, vapor rameter requirements that are not sat Your propeaty methods do not falsely pressure, and ideal-gas heat capacity isied, the next stage should be ob- predict two liquid phases Constants. The need for other parame- taining and using physical property Simulators let you specify that only ters depends upon your choice of data one phase is present in a steam or a physical property methods. The simu Unit operation. If vapor and liquid lator manuals should include the infor. Obtaining and using phases. are possible, however, you mation about the parameter require- physical property data Should use the twosphase specification. ments (7). There also are parameters Sources of data. To provide pi- { that willbe required for calculating the rameters. for_nondatabank_compo- Nondatabank components hheat of reactions or the reaction equi- nents or t do regression for pure- ! end missing parameters librium constants. This includes the component and binary parameters, When you want to simulate non- heat of formation and the Gibbs free you will need to search for available databank components or have compo- energy of formation of all components data. Such data may be found in a va nents for which parameters are miss- that participate inthe reactions riety of sources, including data-com- : ing, ask yourself he fol. pete cecum Psion references, hand lowing ‘books, journals, and inter i +15 this a major compo- Ani move inimi: nal data collections. rent in the mixture? IF itis Techniques to remove or minimize While most streams in rina, cn I take oot of the impact of specific parameters simulations comain x the simulation’ tures, accurate property cal~ + Does the component should be used with caution. calations are not possible take part in VLE? without accurate pure-com- + Is the component non- ponent properties. The im- volatile? portance of pure compo- ‘Is it polar or nonpolar? You can use your judgment about nent data should not be underestimat: + Will reaction (includi the importance of a parameter to set ~ ed as they are the basis for both pure- position) cause this component fo be nominal values for unimportant prop- component and mixture properties, depleted? erties. For example, if you know that For instance, pure component proper- = What properties need to be accu a component is very nonvolatile and ies such as vapor pressure will be rate for the chosen property methods? are using Antoine’s equation for used in phase equilibria calculations These questions will help you 10 vapor pressure (In P =A + B/(T+C)), Table 4 contains common sources for ‘dently the parameters that are ned: you can set the valuc of parameters A, pure component properties, hile ta based on yout choice of physical Band C to -100, 0, and 0, respective: Table 5 lists common sources for property methods Ii these parameters ly. (Tis temperature) Tis will assign minture properties, Ie not available or cannot be deter” the vapor pressure used in Raoul’s The recommended order of data mined through Titerature search, re- Law a very small value, almost ero search is gression, of estimation, then you will (3.7 x 1041). This and similar tech---eiicall evaluated data sources, have to teevaluate your choice of niques to remove or minimize the im- 2. onevaluated sources mi Physical property methods or oblain pact of specific parameters should be 3. experimental measurements; and tis | lata by measurement {sed with caution, however 4 extmatin techniques ee "You should determine what the pa- If you cant find a component in Binary parameters for phase equ tw | rameters will defait to f the simls- the simulstor's databanks, make sure ibria, Because of the lege number #2 1 tordoes not find any available. Its you check for synonyms. For exam of binary pits in even 2 simulation (7 | angerous to assume thatthe physical ple, methoxybenzene may be listed as of only ten components, we recom ‘tes |property parameters were avaiable methyl phenyl ether of anisole. A mend ranking the components 50 2s i just because the simulator did not good approach is to search for the to prioritize the pairs and focus the ett ive you an ertor message. Use the component using its formula. When literature search and measurement ef: simulator manuals and on-line help to selecting the component by formula, forts on the most important parame: CHEMICAL ENGINEERING PROGRESS + OCTOBER 1995 + 41 ‘A ait a i ag SUCCEEDING AT SIMULATION ‘ileal Data of Pure Substances, DEOHEMAT DIPPR Data Complton™ Encyclopedia of Chemicals, Drugs, and Biologic Encyclpadia ct Polymer Scionce and Engineering ESDU Validated Engineering Data Index Handbook of Thermophysial Properties of Gases and iuids LJANAE Thermochamicl Tables Lange's Handbook of Chemistry Perrys Chemical Engineer's Handbook Properties of Gases and Liquids Selactod Values of Proper of Chemical ‘Compounds (TAC Vapor Pressure of Pura Substances * Pars ofthese souces ate avable onine tram ‘ional or Technical Databases Services, Inc. (TDS) 1 ters, Fits, divide the components into three groups: high, mediam, and Jove priority. Base the priority on teria such as composition, and the purity specifications of the process | Tif component purty is specified, that component is important even i it appears only in low concentrations Second, pair the components into high/high, highimedium, high/low, medium/mediam, medium/low, and low/low groups. Search the avaiable sources, including in-house ones, for any data for all groups. If cevain component pairs are known to be- have ideally. they can be exchuded from the search, Next, use the UNI FAC method forthe missing pais in the mediommedium, medianviow, and lowrlow categories. UNIFAC is not recommended, however, for any pairs that include the components of high priority. secondary literature Search can be used to find binary data for similar compounds and those pa- rameters then substituted, Propose experimental work if any binary pa- {rameter data are still missing of if | data regression exposes data as inad. equate (3 Gems eerie ils Ragan aa me ‘af pure component data: [CRC Hencbook of Chemisty and Physis (Beis Citcaly Noneteaby ves Cries ves vee Nonettcaly ves Nonertically Yes Noncically Yes Cricaly ves Noneteally Yes Noneicay ‘es Nonertically Yes coeay Yes Noneitcaly Yes DIALOG lnformaton Senices, STN Interna Regressing data Data regression is a powerful cool for engineers not just to make the best of available data, but also to ana- lyze the goodness of fit of a physical property model to the data, Most sim- blators include a data regression fea- ture, Examples of commonly. re- agressed data include binary VLE and ELE, vapor pressure, heat of vapor- ization, density, and heat capacity. Data regression finds the best fit of parameter estimates to the experimen: tal data. The best fi is represented by finding the lowest value of an objec tive function while matching the phase equilibrium or other constraints One common regression technique is called Maximum Likelihood Estima: tion. The objective function for this ‘method is: E, wy(EAiCr - Goo, Q where j isa data group, C7 and C¢ are ‘measured and estimated Variables, re spectively, such as temperature, pres: sure, composition, or heat capacity, is the standard deviation or the errr in the measurement of the variable, and w, is the weighting of the data group, | | 42 + ocToBER 196 + chenost ENGINEERING PROGRESS ‘When fiting phase equilibria data, the regression algorithm attempts 10 re duce the objective function while the physical property method is being used {o check that the components meet the constraints of phase equilibria. The work of a successful regres: sion involves selecting the right phys. ical property model and parameters, representing the data properly, choos ing appropriate standard deviations of the data, and starting with suitable initial estimates of the parameter ‘The following are general guidelines for data regression. + Make sure that you are regress: ing the right parameters. Use the same physical property method and builtin databank that you will be using in the simulation. Choose pa- rameters that have impact on the data being used, For example, when using an equation-of-state method such as Peng-Robinson or Redlich-Kwong- Soave, you should determine the acentric factor, ©. But, if you are using an activity coefficient method, ‘you should determine wo or more “constants for the Antoine model + Estimate as few parameters as possible. There is a tendency to use a large number of parameters when fit ‘ing a model to data such as tempera: ture-dependent properties. or binary phase equilibria. Try to regress the data with as few parameters as possi ble. If the regression results report that the standard deviation of the est ‘mated parameters is of the same order ‘of magnitude as the values of the pa rameters, you may be estimating 100 ‘many parameters for your given data The larger the temperature range of your data, the more parameters that you can estimate + Watch out for incomplete data. regression may yield poor results if there are missing data points, partic larly composition data. For example some authors do not report all cot positions in VLLE or immiscible LLE. You may need to estimate the missing compositions so that phase equilibrium can be calculated for all components. Find aut how your sit cor ea bar lator handles missing data tw best deal with incomplete data | + Specify the right number of phas: “ea | 4s A regression will yield incorrect Sources 1 the | felts ifthe number of phases is not Asti asics tne Dison GEOHEMIA Chery Saris specified correctly. This 1s a common Bary VLE Ox DIFPR | SBoplem in VELE systems, For some Dour abank spre of OCHEMA da cclecton” Fe aene rth aoa ce phases Hest Mang Oxa Colston, OREM Chemsy Sees co, |__Mawtte dia, he numberof eseS iat ga Euan ta lac, OECHEE Chery Sens SE | shad nerpre because of ie B= ae Euan Eanes of Eero Suton DEKEMA Shit Sees | 25, | __ seatation of the data or lack of de- vapor igi Equi Data Colecton, DECHEMA Chemisty Series Sof | sergton, Oren in VLE dat, only 8 Yiprtg nur sta or ect Soon, OECHEMA hry Sees le | oul liquid composition is reported Yayoi xara fr res fo Boing SaeancesDECHENA remy Sees es even though two liquid phases were ‘Selected Values of Chemical Compounds, Texas A&M University. ines} etemt The author may be reporting SsélguaEqaitramataColetan, OHENA remy Sees | heterogeneous azeotrope — an azeotrope where the vapor compo tion equals the total liquid composi- *Onsine databanks tion but two liquid phases are present. is aot outside the range of your simu- phase, you can remove them from the i When doing the regression of a het- lation. For example, vapor pressure phase equilibria constraints. ‘This is erogeneous azeotrope, divide the data parameters may not have been deter- useful in VLE. Hl into two groups, the VLE data and the mined at temperatures below the nor- “Generate equilibrium data, If ' VLLE data. This will ensure that the mal boiling point. Most physical prop- you have binary parameters for an ac- comect phase equilibria is considered. erty models extrapolate outside the tivity coefficient or equation-of state | In regressions such as this, itis im- temperature bounds reasonably well model, your simulator may be able to portant to use the property tabulation — but at some compromise in accu te VLE or LLE data for regres fd plotting features ofthe simulator cy. ‘The parameter valves also. may parameters. You can \ to check thatthe parameter estimates apply toavery wide range of tempera- repress these "data" with another sore cortectly reproduce the original data, ture and thus not provide as good a fit physical property model, This allows Use amodel’ full functionality A iF you only need & narow eange in the consolidation of known parameters 4 | physical property model may be used simulation, For phase equilibria caleu- into a single property methos ea to calculate several properties. For ex- lations, to improve the accuracy of “Fit other data. Your simulator fit ample, you can use binary excessien- VLE or LLE predictions, you may may have a data fitting feature that can ora thalpy (H,) data and binary VLE or want (o use ternary or quaternary data __be used for plant data. This method sary | LEE data to determine binary parame- 0 fineune binary parameters that may not be as useful for predictive the tts for activity coefficient models. For may be available in the simulator. simulation, though, if the data are not oe ‘equation-of-state models, you simulta- * Check that the parameters repro- _ from a wide variety of conditions. or neously can use liquid- and vapor heat duce the data. The simulator will re- sti ‘capacity, vapor pressure, and heat of port qualitative resuits of the regres- Estimating missing der | Yaporzaion date If date a available Sion, including the residuals (expen- property parameters Tor these properties, use these data to- mental minus estimated variables) Use Property estimation usualy is done ft gether to estimate the parameters. Data the property tabulation or plotting fea- after a data search is performed, to ata. ‘roups of different types can be used tures to reproduce the data at the speci- supply missing property parameters of together in the same regression. fied conditions. This can be performed You can use built-in estimation meth- hat ‘if necessary, regress parameters in the same regression run. Check that ods to fil in some gaps in your physi ven if vaiues are available in the the comet number of phases is pre- cal-property-parameter requirements ca databank. The physical property pa- dicted by allowing tworliquid-phase Simulators include one or more est 3 if rameters found in the built-in pure- calculations for the property able or mation methods for each of the most cu ‘component and binary databanks gen- plot, In addition, your simulator may common parameters. There are two ale, erally are very reliable, You may find, have an option where you can evaluate types of estimation methods for pure mm however, that you need to determine the ft using the exiting parameters component parameters: structural ble new parameters to replace the data- and model_with experimental data group, and corresponding states. oe bank values for your application, without doing a regression. ‘Structural group methods are based Check the built-in parameters to en “= Remove components not in phase on the idea that contributions of the i | ste tate recommended temper equa. f components that ae pas or sacra rope of the compo lure, pressure, and composition range solids or ions da not appear in a nent are additive for properties such as 22 Qe AN a eS] SUCCEEDING AT SIMULATION enna j the Property Name [unis | nas | estate | sotror_| Estimation Method = cncauibeians a a ae] ola bai f Cheat temperature K | ans ssane im Uydersen of] Cencaltenperatre K ons sam os Ambrose 2 Ctrl pressre Nine 2420000 asno 43 deback : Cee! ressire Nine 2m asm 28 iydersen suk Creal press Nine 3az0000 aaaamo 1 Ambrose ' ical voine | mame a2 os om doback the Crical lane mimo ox oan om iysersen the Cia vote | crnmot os tomy | oan bo nee 5 Standard ent omatont “simole siovsmon | 150000 an Benson re Standard eat oatont “akmote =101g00000 104 4000 20 deback es ata, 25° for eal gs. cor sp } aa (189.9°C), density at 25°C 0.944 | aif i gett), and molecular structure iy Estimated ale | Estimation Method oe or Es ileclr wah may | Forma Citic empertce senso | Ambrose ee ee Citealorersre Nine 2ms.350 Anise Beer eee \ Cito inate asta | Fetes tes of PPE is | Stardardeatotfomatont | Inno | 1253%000 Benson Seep 1 Detecmine dhe best eto ia + at an 25° frie gos tion methods for a similar phenyl | no: ether Select other compound(s) chem- | me ically similar to PPE for which you | — cor normal boiling pont, criical tempera. Corresponding states methods are have experimental property data. (Of | om ture, critical pressure, idea-gas heat ca- based on empirical mathematical re- course, the tore similar compounds | un pacity, and standard heat of formation, Iationships among properties. For ex- you ean use, the stealer your conf | ple Some methods, such a5 that of Benson, ample, the LetsouStiel method re- dence that you are selecting the most alk contain additional comections for next- lates liquid viscosity to critical tem- appropriate methods ) In this case, for | lar | nearest-neighbor afoms of for rings. perature, eitical pressure, and acen- simpli let's choose only phenetol Fo i Structural group contributions ae deter. trie factor, These methods most likely oc pk | mined by aking an average contibution will be inaccurate when used for oO me | tased on known physical constants of compounds unlike those upon which ja } many organic compounds, Because te the corelation was based Data for phenctlis available from | by \ Benson, Jobsck (10), and other sruc- A. good approach for both group the DIPPR data collection (5). ae | tura-group methods are based mainly contribution and corresponding sates Use the simultor'sbultin methods vs } on data for organic compounds, they methods is to check the accuracy of to estimate properties for phenetol | yo | cannot be used for inorganies, including as many methods as. possible for ‘Then, compare the results ofthe various te |) set oF fos. In adliton, sacral compounds for which properties are mths wth the experimentally dt soup methods do not accurately repre- known and which ae structurally mined values to identify which methods « | Sent very large organic molecules (that similar to the compound you are esti--__ give the best estimates for this clas of oy | is. ones with a molecular weight > 200) mating. The following example Compounds. Table 6 lists the results for a | suchas poten, New peupeourbn. Show he ane of ti come Oediteen metiisfareeol | |L—Sonsmediods ike tof ConsantneuEstmarng the papers ofpropy! "You ae that ie ambos | of and Gani (4) potentially may provide phenyl ether. Let's say that you ‘method gives the best overall predic> = | kn + esis for canes Ott odelng process cowaning op)! om foetal empere sal pes) poss wef metas ae poposedin phenyl eter (PPE), ao caled swede Felon metal frcatealvor | the erate bt ma apps only cet propyoxy benzene. ‘The ony dia ume, and he Joback meted foe sa” | tin fans of eompore Jou have aes ‘boling” point daventofformaton fa phe $0, | 70 44 + ocToser iss » cHemicaL ENGINEERING PROGRESS j ods tol ter ods of for So, we will use these methods to predict the comesponding properties for PPE. Step 2. Enter the available data and structure for PPE, Enter normal boiling point and molecular structure of PPE, and specify the methods that agave the best predictions for phenetol, ‘Step 3. Examine the estimation re- sults for PPE. These appear in Table 7 One area of property estimation that is more difficult is ditferentiating the properties of stereo isomers. Some _group-contribution methods have cor- rections for ortho, meta, and para configurations, fut few have builtin corrections for optical isomers. The separation of these isomers in a chem. ical process is based on their slight different properties — relative volatl ity in distillation is one example Employing simpler methods In addition to structural group and corresponding states methods, another ‘useful estimation approach is provided by series and family plots. Series plots look a the values ofa property such as normal boiling point with increasing molecular weight or carbon number for compounds in a series that differ by ‘one substituent group, such asthe CH, unit in malkanes. Figure 10 isa series plot for the normal boiling point of alkylbenzenes, Family plots are simi- lar, but the number of groups is larger For example, Figure 11 shows a family plot of the critical pressure of ‘methyl(hydrogen}ehlorosilanes, You can use these plots to predict properties by extending the curve or to check your data for errors (6). To create a useful series or family plot, however, ‘you must be careful about the compo: rents included ‘When accuracy isnot critical, consid er the simple but powerful technique of ‘component substitution. kn tis, you use the properties of another, similar compo- nent for all properties of the component of interest that you do not know. A si Jar component is one that has a compa: rable volatility (vapor pressure), density, and heat capacity, This is useful if the component is nonvolatile or is not in- ‘volved in phase equilibria. For example, you fave a small amount of a non ‘olatile component in 2 stream that is at 100°C and 1 atm, You can access the properties of a nontolatile component, say CgHy, (molecular weight = 282 and boiling point = 343,78°C), instead of estimating properties. This method is very efficient if you do not need accurate properties of the component. Take ca though, if you use this approach and one of the UNIFAC acuvity-ccefficent ‘methods, as you may change the as- suptions made about the Fiquid phase Another technique to simplify a mixture of similar components is to present them with a single compo- nent, This is a useful technique when components are not known exactly. For instance, Component C;+ could represent hydrocarbons of 5 carbon atoms and greater. 5 one eo Number of Carbon Atoms Si —9 sit ‘yom < Ost | noise, “er ee i ‘200m bys & en “Oo esi aaron |- mK Ome amas Le Estimating binary parameters You can estimate binary parame ters for Wilson, NRTL, and UNL QUAC. sctivity-cueficient models using 80 approaches: UNIFAC and infinte-dilution activity coef UNIFAC-estimated binary parameters usually do not provide enough accu racy and, so, only are recommended for early stages of physical property data invesugation and to “fill in the blanks” for components with medium or low priorities ewer binary parameters can be estimated using infnite-dilution ac tivity coefficient data, (Some simula tors may inelude this feature under their regression tools) This method is beter because it is based on the com. ponents of interest, unlike the group contribution method, which averages Figure 10 {left Series plot of normal boiling point for nealls- benzenes BW Figure 11 (below). Family plot of erical pressure of Imethylthydrogen) hlorosilanes 2 8 ‘CHEMICAL ENGINEERING PROGRESS + ocToBER 1385 + 45 a SUCCEEDING AT SIMULATION Literature Cited 1. Reld RC. M. Prausnite, and BLE Poling, “The Properties of Cases snd Liquids 4th ed, McGraw Hil, New York (1987), 2. °CRC Handbook of Chemistry and Physics” D. Lids, ed, CRC Press, Bocs Raton, FL (0998), 3. Dewan, A. K, and MA. Moore, ‘Methodology 19 Develop an ASPEN PLUS Model in Shell Development Com any presented at_ ASPEN WORLD, Cambridge, MA (1994), (Available fom [Aspen Technolog Ine) 4 Constantinou, Ly and R. Gani, “New Group Contibtion Method for Estat ing Properties of Pure Compounds AICHE 1, 40 (10), p. 1,697 (1998), Dauber, T By RB. Danner, He M. Sul, and C. C. Steins, "Phil sed Thermodynamic Properties of Pure Chem icals: Dua Compilation," Design Taine for Physical Property Daa (DIPPR, [AICHE, Nes Yer (1989-onwd). & Smith, AL, “Fam Plots for Evatt ing Paysical Propetes of Organs Compounds” AICHE J, 40°C), p. 3 9%) 7. “ASPEN PLUS User Guide” Vol. 1, Re lease 8, Aspen Technology, Ine, Cam Iaadge, MA (1995, 8, Gmehlng, J, J-Li,and ML Seller," Modified UNIFAC Model. 2. Present a rameter Matox and Rests fr Diferent Thesenodynamie Properties” MLEC Res, 32, p178 (1993, tions from the effect of group inter different components Estimation of physical properties can get you started in a simulation problem — bot you should do an ex haustive literature search to find missing pure-component and binary parameters. It is important to enter any known parameters before doing property es timation. First, experimental dat generally are more accurate than esti ‘mated values. Second, corresponding states estimation methods require ther physical constants as input Using an experimental value will im prove the prediction of these property parameters. Otherwise, the error in estimating parameters such as normal boiling point, critical temperature, 9, Hansen Hh. Ky P. Rasmussen, A. Fre:

    Anda mungkin juga menyukai