Korean J. Chem. Eng., 18(2), 184-189 (2001) Heat Tranfer Effect of Inert Ga on Multi-Tubular Reactor for Partial Oxidation Reaction Kwang Ho Song, Sang Eon Han and Kwang-Ho Park LG Chem Reearch Park, LG Chemical Ltd., Yu-ong P.O. Box 61, Taejon 305-380, Korea (Received 1 July 2000 accepted 26 December 2000) Abtract The heat tranfer effect of an inert ga on a multi-tubular reactor for a partial oxidation reaction ha been determined. The model reaction ytem in the tudy wa partial oxidation of propylene to acrolein. Both theoretical modeling and experimental tudie have been performed to determine the heat tranfer effect of inert ga on the ytem. Among many inert gae, CO 2 wa elected and teted a a diluent ga for the partial oxidation of propylene to acrolein ytem intead of conventionally ued N 2. The productivity increae through changing the inert ga from N 2 to CO 2 wa poible due to the heat tranfer capability of CO 2. In thi tudy, by replacing the inert ga from N 2 to CO 2, productivity increaed up to 14%. Key word: Multi-tubular Reactor, Inert Ga, Heat Tranfer, Acrolein, Carbon Dioxide INTRODUCTION Multi-tubular reactor are deigned for reaction that generate a large amount of heat [Satterfield, 1970]. Due to the unique characteritic of eay heat removal, multi-tubular reactor are widely ued for partial oxidation reaction procee. A multi-tubular reactor uually ha more than 10,000 reaction tube. The heat tranfer medium, which i made of molten alt, circulate between the reaction tube and remove heat from them. Heat tranfer of reaction heat from the reactor tube to the heat tranfer medium i crucial for partial oxidation reaction procee. Detailed decription of multi-tubular reactor can be found elewhere [Franzen et al., 1964; Jung et al., 1999]. Since the reactant i a flammable ga and ha a poibility of exploion, a large amount of inert ga i fed into the reactor along with reactant. Inert ga alo play a role of heat tranfer agent becaue it take up a large portion of reactant ga mixture. Among many partial oxidation reaction, partial oxidation of propylene to acrolein wa elected a a model ytem to invetigate the heat tranfer effect of inert gae. Uually N 2 i ued a a diluent ga in the cae of propylene partial oxidation to acrolein. N 2 ga conit of more than 50% of total input ga by volume in propylene partial oxidation to acrolein. Different inert gae can be applied to thi reaction ytem a a diluent ga. When an inert ga i mixed with reactant, there exit a pecific ratio of mixture that become flammable or exploive. Thi pecific ratio of mixture alo depend on temperature and preure of the reaction ytem. Uually, thi ratio of mixture ha a lower and upper boundary at contant temperature and preure. If the range of thi boundary for another inert ga i narrower than that of N 2, that inert ga i conidered to be a better diluent ga in term of afety. The productivity can alo be increaed while uing the inert ga that ha a narrower flammability or exploion range becaue more reactant can be fed into the multi-tubular reactor. Although To whom correpondence hould be addreed. E-mail: khong@lgchem.co.kr a elected inert ga may meet the flammability limitation, the heat tranfer ability of the ga i alo very important for partial oxidation reaction. If the reaction heat i not removed properly, the temperature will rie above normal operating condition, poibly reulting in a run-away ituation inide the reactor. When thi happen to the reactor, there will be evere damage to the catalyt. In an effort to find a better diluent ga for thi model ytem of propylene partial oxidation to acrolein, CO 2 ha been applied to the reaction ytem intead of N 2 ga and compared with the tandard cae of N 2. CO 2 ha a narrower flammability or exploion range than N 2 for the preent combination of inlet mixture gae [Song and Park, 1998]. Since CO 2 i better than N 2 in term of the flammability limitation, only the heat tranfer effect of inert gae will be conidered in thi tudy. Both experimental and theoretical tudie have been performed to analyze the effect of inert ga on partial oxidation of propylene to acrolein. EXPERIMENTAL The chematic diagram of the experimental apparatu i hown in Fig. 1. The reactor ued in the experiment i a jacketed ingle tubular reactor [Choi and Eigenberger, 1989]. Since it i very dif- Fig. 1. The chematic diagram of the experimental apparatu. 184
Heat Tranfer Effect of Inert Ga on Multi-Tubular Reactor for Partial Oxidation Reaction 185 ficult to make a lab-cale multi-tubular reactor, a ingle tubular reactor that exhibit the ame characteritic of a multi-tubular reactor wa made intead. The reactor conit of a 2.54 cm outer diameter tainle teel tube and an outer jacket made of 7.62 cm outer diameter tainle teel pipe. The wall thickne of the reactor tube wa 1.651 mm. The heater wa placed outide of the jacket. The length and diameter of the heater were 90 cm and 30 cm, repectively. The reactant and inert ga are fed into the reactor by uing a Brook ma flow controller. Water wa pumped into the heated tainle teel tube by uing a Hitachi L-7110 contant ma yringe pump. Water wa then evaporated into team and mixed with both reactant and inert ga. The molten alt, a mixture of NaNO 2, Na- NO 3, and KNO 3, wa ued a the heat tranfer medium, which upplie and remove heat from the fixed bed reactor. A thermowell tube of 0.398 cm outer diameter wa mounted coaxially at the center of the reactor tube and alo inide of the jacket, repectively. A mobile thermocouple wa placed inide of the thermowell tube located at the center of the reactor o that the temperature could be meaured along the length of the reactor. Additionally, three fixed thermocouple were placed inide of the jacket pipe in order to meaure the temperature of the heat tranfer medium at three different location. The product were analyzed by a Hewlett-Packard ga chromatograph HP6800. Ga chromatography provided on-line analyi of a ample through a computer. Two different type of column were ued imultaneouly to analyze the product ga: an Altech AT-1000 capillary column and a Hewlett-Packard chromoorb 102 packed column. Two different ize of cylindrical hape catalyt were ued for the experiment. A detailed decription of the catalyt ued in thi invetigation can be found elewhere [Byun and Lee, 1999; Ko et al., 1998]. Alumina, a an inert material, large and mall ize catalyt were charged into the reactor in erie. The height of the Alumina and catalyt bed were 5.2 cm, 10.9 cm, and 30.4 cm, repectively. The amount of catalyt charged in the reactor wa 32 g for the large ize catalyt and 91 g for the mall ize catalyt. The olid mixture of NaNO 2, NaNO 3, and KNO 3, wa the heat tranfer medium. It wa melted by uing the heater placed outide of the jacket. After temperature of the heat tranfer medium reached the pre-et value, the reactant were fed into the reactor. It took at leat 72 hour to reach the teady tate. The experimental condition are hown in Table 1. MODELING Since partial oxidation of propylene i a highly exothermic reaction, there remain the poibility that the temperature of the fluid phae may be much different from that of the olid phae. Thi lead to the neceity for introducing a heterogeneou model. The conideration of intraparticle gradient in the heterogeneou model i appropriate for the preent reaction ytem becaue it i important to predict preciely the maximum temperature of the olid phae. In general, the radial gradient can be neglected for all practical purpoe if the radial apect ratio m(=r/d p ) doe not exceed 4 [Carberry, 1976]. In the preent experiment, the value of m wa 2.7 and 3.2 for the two different ize of catalyt bed, repectively. All of thee value jutify the uage of a one-dimenional heterogeneou model accounting for intraparticle gradient. The reaction kinetic model and parameter of Tan et al. [Tan et al., 1989] were adopted in thi invetigation. The model equation wa derived from teady tate ma and energy balance for fluid and olid phae a follow [Froment and Bichoff, 1979]: Fluid Phae: d( u C i ) d 2 C ---------------- + εd i ea --------- = k 2 gi a v ( C i C i ) d( ρ g u c p TM w ) ---------------------------------- d 2 T λ ea ------- = h 2 f a v ( T T) 4 U --- ( T T d r ) t Solid Phae: 1 -- ---- d D r dr ei r dc i -------- ρ dr r i = 0 1 -- ---- d λ r dr e r dt ------- + dr ρ [( H 1 )k 1 + ( H 2 )k 2 + ( H 3 )k 3 ]C 4 θ x = 0 where i=1, 2, 3,, 7. 1: acrolein; 2: CO 2 ; 3: acetaldehyde; 4: propylene; 5: H 2 O; 6: O 2 ; 7: N 2 The rate coefficient k i (i=1, 2, 3) in Eq. (4) correpond to k 12, k 13, k 14 of Tan et al. [Tan et al., 1989]. Boundary condition Fluid Phae: u dc ( C i0 C i ) = εd i ea ------- u ρ g c ------------- p ( T 0 T) = dt λ ea ----- M w dc i dt ------- = ----- = 0 Solid Phae: dc i dt -------- = ------- = 0 dr dr dc D i ei -------- = k dr gi ( C i C i ) (1) (2) (3) (4) at z=0 (5) at z=0 (6) at z=l (7) at r=0 (8) at r=r (9) Table 1. Experimental condition Run Inert ga C 3 H 6 flow rate O 2 flow rate H 2 O flow rate Total flow rate Inlet temp. (K) Jacket temp. (K) Reactor preure (atm) 1 N 2 0.014 0.026 0.015 0.2 423 581 1 2 CO 2 0.014 0.026 0.015 0.2 423 581 1 3 CO 2 0.016 0.030 0.015 0.2 423 581 1 Korean J. Chem. Eng.(Vol. 18, No. 2)
186 K. H. Song et al. dt λ e ------- = h dr f ( T T) at r=r (10) Since denity change in the axial direction, the velocity alo change to atify the continuity equation: u ρ g =contant (11) Many correlation for ma and heat tranport are available in the literature, but there are trong deviation among them epecially for heat tranport. The temperature profile wa hown from the numerical experiment to be dependent more upon the heat tranfer coefficient than the ma tranfer coefficient. Becaue of the lack of ufficient certainty in ma and heat tranfer coefficient, it eemed reaonable to ue a correlation which bet fit the experimental data even though the catalyt ued in the preent experiment wa different from that ued in the work of Tan et al. [Tan et al., 1989]. The Petrovic and Thodo [Petrovic and Thodo, 1968] correlation for ma tranport and the Rowe et al. [Rowe et al., 1965] correlation for heat tranport were put to ue in the modeling. The wall heat tranfer coefficient wa found to have much more effect on the temperature profile than the heat tranfer coefficient between the fluid and the olid phae in the modeling. Many empirical correlation uch a Leva and Grummer, Hanratty, Yagi and Kunii, De Wach and Froment [Doraiwamy and Sharma, 1984], Calderbank and Pogorki [Calderbank and Pogorki, 1957], and Beek [Szekely et al., 1976] were teted. The equation of Calderbank and Pogorki [Calderbank and Pogorki, 1957] among many empirical correlation wa found to fit bet to the preent experimental reult. The model equation were olved by uing finite difference method with an upwind cheme [Roach, 1972]. Since large et of coupled nonlinear equation are difficult to converge unle an appropriate initial gue i given, equivalent unteady tate equation were contructed and made to march until the teady tate wa reached. It wa ufficient to apply the peudo-tranient cheme only to the fluid phae equation. Since the tranient olution wa not of interet, the time tep were choen to minimize computation time. The nonlinear equation were effectively olved with the ucceive over relaxation method. The experimental condition are decribed in Table 1. The firt experimental condition hown in Table 1 i a tandard operating condition of propylene partial oxidation to acrolein. All the reult obtained from uing CO 2 a a diluent ga are compared with thi tandard cae. The experimental and calculated temperature profile of the fluid phae are compared in Fig. 2 for both diluent gae N 2 and CO 2, where imilar hape are oberved. The dicrepancie between the experimental and modeling reult may be mainly attributed to the difference in reaction kinetic and the poor validity of the wall heat tranfer correlation equation. In both cae of modeling and experimental reult, the fluid phae temperature profile wa lowered by the adoption of CO 2 a a diluent ga intead of N 2. The modeling reult of propylene concentration and temperature profile of the olid phae, in the cae of N 2 a a diluent ga, are given in Fig. 3 and 4, repectively. The concentration of propylene decreae with axial ditance a the reaction proceed. The Fig. 3. Three-dimenional calculated propylene concentration profile of the olid phae in the cae of experimental condition 1 hown in Table 1. RESULTS AND DISCUSSION Fig. 2. Experimental and calculated axial temperature profile of the fluid phae in the cae of N 2 and CO 2. Fig. 4. Three-dimenional calculated temperature profile of the olid phae in the cae of experimental condition 1 hown in Table 1. March, 2001
Heat Tranfer Effect of Inert Ga on Multi-Tubular Reactor for Partial Oxidation Reaction 187 Fig. 5. Calculated maximum temperature of the olid phae along axial ditance for experimental condition hown in Table 1. Fig. 7. Heat capacity of N 2 and CO 2. location of maximum temperature in Fig. 4 indicate the hot pot of olid catalyt. When thee figure are compared with Fig. 2, it can be hown that the dominant tranport reitance fall on the ma diffuion in the catalyt pellet and on the heat tranfer between the fluid phae and external catalyt urface. Since the maximum temperature in the olid phae i a major concern from the viewpoint of catalyt lifetime, it profile alo need to be examined. The utilization of CO 2 a a diluent ga intead of N 2 i found to be ucceful in diminihing the maximum olid phae temperature, a hown in Fig. 5. The following apect can explain the role of CO 2 in decreaing the temperature of both fluid and olid phae: thermodynamic and kinematic perpective. The thermodynamic propertie include heat capacity and the kinematic propertie encompa vicoity, diffuivity, thermal conductivity, etc. To get an inight into the mechanim of the temperature decreae in the cae of CO 2, the effect on temperature of all the differing propertie of CO 2 from N 2 wa examined through a parametric enitivity tudy. The variation of vicoity, diffuivity and thermal conductivity had little effect on the temperature profile. The factor which had large influence on temperature were heat capacity and molecular weight (denity changed in accordance with molecular weight). Effect of both factor on the fluid phae temperature profile are hown in Fig. 6. Similar trend were oberved in olid phae temperature profile. The effect of heat capacity can be accounted for eaily. Becaue of it larger heat capacity, CO 2 aborb more heat than N 2 to reduce the temperature of both fluid and olid phae. The variation of heat capacitie with temperature for both diluent gae i hown in Fig. 7. A for the effect of molecular weight, it i appropriate to find it relationhip to kinematic propertie to undertand the mechanim. The only kinematic parameter that had a relation to molecular weight and had a ignificant effect on temperature wa the wall heat tranfer coefficient. Since the wall heat tranfer coefficient i generally dependent upon Reynold number [Calderbank and Pogorki, 1957], it i reaonable to examine what effect molecular weight of CO 2 ha on Reynold number. The Reynold number and correponding wall heat tranfer coefficient change due to the adoption of molecular weight of CO 2 are hown in Fig. 8 and 9. Therefore, kinematic vicoity i an important parameter, which can be ued to predict temperature change, ince Reynold number i entirely affected by kinematic vicoity under certain operating condition. Although other propertie uch a thermal conductivity and Prandtl number Fig. 6. Calculated axial temperature profile of fluid phae in the cae of N 2, N 2 with heat capacity of CO 2, N 2 with molecular weight of CO 2, and CO 2. Fig. 8. Effect of molecular weight on Reynold number in the cae of N 2 and CO 2. Korean J. Chem. Eng.(Vol. 18, No. 2)
188 K. H. Song et al. CONCLUSIONS Fig. 9. Effect of molecular weight on heat tranfer coefficient in the cae of N 2 and CO 2. alo affect heat tranfer coefficient, they vary inignificantly. It ha been hown that the temperature of both fluid and olid phae can be lowered via increaed heat tranport and heat capacity by uing CO 2 intead of N 2 a a diluent ga. From the above reult, it can be expected that the productivity can be increaed through increaing inlet propylene concentration without exceeding the afe operation range if CO 2 i ued a a diluent ga. The calculated temperature profile of fluid and olid phae are hown in Fig. 2 and 5 where inlet propylene concentration i 8 vol% with CO 2 a a diluent ga. The reult how that the maximum temperature of the fluid phae i within that of the tandard cae; therefore, it will be conidered a a afe operating condition, ince the maximum temperature of fluid phae i a meaure of afety. Fig. 5 clearly how that catalyt lifetime will not be affected by increaing inlet propylene concentration to 8 vol% with CO 2 a a diluent ga. Fig. 2 how that the modeling reult i in good agreement with the experimental reult. Although it wa in the afe operation range, the propylene converion wa lightly lower in the cae of 8 vol% propylene concentration with CO 2 a a diluent ga than that of 7 vol% propylene concentration with N 2 a a diluent ga. The lower converion i due to a decreae in reaction temperature. To compenate for the decreae in propylene converion, the inlet coolant temperature increaed in the numerical imulation. The reult of increaing the inlet coolant temperature are hown in Table 2. The propylene converion increaed and the fluid phae maximum temperature wa within the afe operation range. The reult how that the utilization of CO 2 a a diluent ga can improve productivity up to 14% within the afe operation condition by increaing inlet propylene concentration from 7 to 8 vol% and coolant temperature. A heterogeneou model, which account for intraparticle gradient, wa ued to imulate the partial oxidation reaction of propylene in a ingle tubular reactor, repreenting one of many tube in a multi-tubular reactor. It wa made clear that the temperature of the fluid phae and olid phae wa lowered when CO 2 wa ued intead of N 2 a a diluent ga through the calculation with thi model. It wa proven that large heat capacity and low kinematic vicoity were the main caue of diminihing temperature in the preent operating condition. The other parameter did not how any ignificant effect on temperature. Thu the reult can provide a guideline for electing an inert ga ued in partial oxidation reaction. The model wa then applied to the condition in which the inlet propylene concentration wa higher than normal operating condition with CO 2 a a diluent ga. The temperature of both fluid and olid phae wa ufficiently within the limit, but the propylene converion wa le than that of N 2 a a diluent ga. The problem of low converion wa effectively olved by increaing the inlet coolant temperature. It wa revealed that increaing productivity wa poible while maintaining afety by uing CO 2 a a diluent ga in thi model reaction ytem. a v C i C i0 C i C i c p D ea D ei d p d t h f NOMENCLATURE : external particle urface area per unit reactor volume : molar concentration of component i in the bulk fluid : molar concentration of component i at the inlet : molar concentration of component i inide olid : molar concentration of component i at olid urface : heat capacity : effective diffuivity in axial direction : effective diffuivity of component i : particle diameter : internal tube diameter : heat tranfer coefficient H : heat of reaction k gi : ma tranfer coefficient of component i k j : rate contant for the reaction j (j=1, 2, 3) M w : molecular weight r : radial coordinate inide olid R : radiu of reactor tube r i : rate of reaction of component i T : temperature in fluid T 0 : temperature at the inlet T r : temperature of urrounding : temperature inide olid T Table 2. Effect of coolant temperature on C 3 H 6 converion and maximum temperature of fluid and olid phae Inert ga [Propylene (vol%)] March, 2001 Inlet coolant temperature (K) Converion of propylene Maximum temperature of olid phae (K) Maximum temperature of fluid phae (K) N 2 [7] 581 0.846 737 661 581 0.838 725 655 CO 2 [8] 586 0.839 725 655 591 0.855 732 661
Heat Tranfer Effect of Inert Ga on Multi-Tubular Reactor for Partial Oxidation Reaction 189 T U u z : temperature at olid urface : overall heat tranfer coefficient : uperficial velocity : axial coordinate Greek Letter ε : void fraction of packing λ e : effective thermal conductivity λ ea : effective thermal conductivity in axial direction θ x : fraction of oxidized ite ρ g : denity of fluid : denity of olid ρ REFERENCES Byun, Y. C. and Lee, W. H., Proce for Manufacturing Catalyt for the Acrolein and Methacrolein Production, KR Patent, 99-21570 (1999). Calderbank, J. J. and Pogorki, L. A., Heat Tranfer in Packed Bed, Tra. Int. Chem. Eng. (London), 35, 195 (1957). Carberry, J. J., Chemical and Catalytic Reaction Engineering, McGraw- Hill, New York (1976). Choi, J. and Eigenberger, G., Dampfeinparung Durch Peiodiche Katalyator-Reaktrivierung Bei Der Styrol-Synthee, Chem.-Ing.- Tech., 61(8), 641 (1989). Doraiwamy, L. K. and Sharma, M. M., Heterogeneou Reaction: Analyi, Example, and Reactor Deign, Wiley, New York (1984). Franzen, E. P., Maycock, R. L., Nelon, L. E. and Smith, W. C., Tubular Catalytic Reactor with Cooler, US Patent 3147084 (1964). Froment, G. F. and Bichoff, K. B., Chemical Reactor Analyi and Deign, Wiley, New York (1979). Jung, K. Y., So, J. H., Park, S. B. and Yang, S. M., Hydrogen Separation from the H 2 /N 2 Mixture by Uing a Single and Multi-Stage Inorganic Membrane, Korean J. Chem. Eng., 16, 193 (1999). Ko, D., Lee, W. H. and Paek, K. H., The Preparation Method for Partial Oxidation Catalyt, KR Patent, 177326 (1998). Petrovic, L. J. and Thodo, G., Ma Tranfer in the Flow of Gae Through Packed Bed, Ind. Eng. Chem., Fundamental, 7, 274 (1968). Roach, P. J., Computational Fluid Dynamic, Hermoa, Albuquerque, N. M. (1972). Rowe, P. N., Claxton, K. T. and Lewi, J. B., Heat and Ma Tranfer from a Single Sphere in an Extenive Flowing Fluid, Tran. Int. Chem. Eng., 43, T14 (1965). Satterfield, C. N., Ma Tranfer in Heterogeneou Catalyi, The Maachuett Intitute of Technology, 78 (1970). Song, K. H. and Park, K. H., Proce Development of Acrylate, LG Chem reearch report (1998). Szekely, J., Evan, J. W. and Sohn, H. Y., Ga-Solid Reaction, Academic Pre, New York (1976). Tan, H. S., Downie, J. and Bacon, D. W., The Reaction Network for the Oxidation of Propylene over a Bimuth Molybdate Catalyt, Can. J. Chem. Eng., 67, 412 (1989). Korean J. Chem. Eng.(Vol. 18, No. 2)