catalytic coating for reduced coke formation in ethylene- producing steam crackers: experimental and model validation

Transcription

1 Water Technologies & Solutions technical paper catalytic coating for reduced coke formation in ethylene- producing steam crackers: experimental and model validation Schietekat, C.M.; Reyniers, P.A.; Sarris, S.A.; Van Geem, K.M.; Marin, G.B. Laboratory for Chemical Technology (LCT), University of Ghent Kool, L.; Peng, W.; Lucas, P.; SUEZ Global Research, Niskayuna, NY and Shanghai, China abstract In this paper we describe our four-year research effort that has resulted in the discovery of YieldUp, a robust, novel coating technology for the inner wall of furnace tubes that prevent the deposition of coke. YieldUp is based upon a family of ceramic catalysts having a unique structure that is designed to convert coke to carbon oxides on contact. Thus, when coke forms during cracking, it is instantaneously gasified on contact with the wall. The performance of the coating was tested in a Jet Stirred Reactor (JSR) set-up and at the UGent pilot plant for steam cracking. The JSR is used to assess the effect of three different coating formulations on both coke formation and product yields, including CO and CO 2. The JSR apparatus measures the quantity of coke deposited over time by means of continuous thermogravimetric analysis. The measured rates are compared to those of a reference uncoated alloy. These experiments allowed optimization of the catalyst activity to reduce coke formation, while minimizing the production of carbon oxides. The coating activity after several coking/decoking cycles remained stable. addition, presulfidization and dilution) and feeds (both ethane and naphtha) was evaluated. The quantities of coke produced were drastically reduced as compared to an uncoated reference coil. These pilot experiments showed that the catalyst is robust and maintains anti-coking activity over multiple crackingdecoking cycles. 1. introduction Coke formation on the inner wall of the tubular cracking reactors of steam cracking units has a major influence on the energy efficiency and economic viability of the steam cracking process. Hence, many efforts have continued to be carried out in recent years towards the development of technologies to reduce coke formation. One such technology is the application of a coating on the reactor inner wall. Distinction can be made between the coatings that passivate the inner coil wall 1, 2 and catalytic coatings 3 that convert coke to carbon oxides. Figure 1 shows the effects of different coatings on coke formation. A so-called barrier coating passivates the catalytically-active sites of the reactor alloy, eliminating catalytic coke formation. However, the non-catalytic coke formation, often termed pyrolytic coke, is not prevented. In contrast, catalytic coatings convert deposited coke to carbon oxides and hydrogen, by reaction with steam, through gasification reactions. The best-performing coating was also studied on a larger scale in a pilot plant. This experiment allowed the coating s performance to be evaluated under typical industrial conditions in a well-controlled and monitored environment. The influence of several process conditions (coil-outlet-temperature, continuous sulfur Find a contact near you by visiting and clicking on Contact Us. *Trademark of SUEZ; may be registered in one or more countries SUEZ. All rights reserved. AICHE_Catalytic_Coating_Apr2014.docx Apr-14

2 Reactor tube Barrier Coating Reactor tube Catalytic Coating Reactor tube Figure 1: Effect of different coating types; from left to right: bare tube, barrier coating and catalytic coating. We have developed YieldUp, a new catalytic coating based on upon a family of ceramic catalysts having unique chemical structures that are capable of converting coke to carbon oxides on contact. The performance of different formulations of our coating was probed in a Jet Stirred Reactor (JSR) set-up. The optimal coating was further tested in the UGent pilot plant 4. In addition, the predicted effects of scale-up of the coating in an industrial ethane cracker were simulated. In the following sections, the coating, experimental set-ups, procedures and results are discussed. Finally, the modeling results are summarized. Figure 2: Incolloy 800 HT JSR coupons coated with BCZ slurry catalyst. The same slurry formulation was used to coat the 9mm ID of the furnace tubes used in the pilot experiments described above. Typical coating thickness ranged from 20 to 50 µm, as shown in Figure description of the coating The coatings that were studied in the experimental program were developed at the SUEZ Global Research centers in Shanghai, China and Niskayuna, New York. YieldUp consists of an engineered synthetic ceramic. When exposed to high temperature steam, water molecules are chemisorbed and deprotonated, resulting in the formation of highly reactive oxygen atoms that instantaneously react to form CO and CO 2 upon contact with coke. We have developed a family of YieldUp catalyst materials that provides a range of anticoking activities, as described below. The ceramic coating was applied to the JSR coupons by forming an aqueous slurry of catalyst microparticles and other additives and dip-coat the coupons, followed by a high-temperature sintering step. Figure 2 shows typical JSR coupons prepared in this way. Figure 3: Coating thickness uniformity in pilot plant tubing ID. 3. description of the experimental set-ups 3.1 Jet stirred reactor A schematic representation of the jet stirred reactor (JSR) is shown in Figure 4. It has four nozzles oriented in symmetrically opposite directions to assure good mixing of the gas in the reactor, in an attempt to obtain as close to ideal mixing as possible, and emulating the concept of a CSTR reactor 5,6. The reactor is entirely made of quartz, including the nozzles and walls, to limit coking to the coupon surface. To study the coke deposition, a small flat coupon (shown in Figure 4) is inserted into the JSR and is suspended Page 2 AICHE_Catalytic_Coating_Apr2014.docx

3 from the arm of an electrobalance. The reference coupons used were cut by electro-erosion from the internal surface of the tubes to dimensions 10 mm 8 mm 1 mm. A small hole (diameter 0.6 mm) is drilled 2mm from the top of the sample so the coupon can be positioned inside the reactor from a Kanthal wire with a 0.25 mm diameter. This Kanthal wire is connected to the arm of a Cahn D-200 electrobalance, which continuously measures the weight of the coupon as a function of time. The balance can measure the weight with an accuracy of 0.01 mg. Outlet Nozzles Reactor tubes Thermowell a) Inlet Coupon Figure 4: a) Diagram of the JSR Reactor; b) Cut industrial tubes and a JSR coupon. The reference samples were polished. The surface roughness values (Ra) of each coupon was measured for the polished reference coupons in three directions (vertically, horizontal and diagonally), but for the coated coupons only in one direction to avoid damaging of the coating. The measured values oscillated between 0.06 m and 1.49 m. Before placing the reference samples into the reactor, they were thoroughly cleaned to remove any contaminants on their surface. First they underwent a cleaning procedure, which started by washing them in a small glass with deionized water for approximately 5 minutes. Then, they were washed with diisopropyl ether for 2 minutes, and finally with acetone in an ultrasonic bath for 2 hours. After that, an electrolytic scouring in diluted H2SO4 (1.5 wt%) took place for 15 min at a voltage of 8V, with the metallic sample connected to the cathode and a graphite stick connected to the anode. Finally, the samples were briefly rinsed with deionized water, and hung from the balance, to start a new cracking test. After cracking, the effluent is cooled immediately downstream of the reactor, in order to quench the effluent and impede any further cracking reactions from taking place after the effluent leaves the reactor. For the analysis of the effluent composition, two gas chromatographs are used: a Refinery Gas Analyzer (RGA), capable of detecting and quantifying permanent gases (up to C4), and a Trace ultra GC, measuring the effluent from methane up to 2-methyl naphthalene. Computational fluid dynamics (CFD) simulations have been performed to verify the uniformity of the JSR in terms of temperature and species concentration. An ethane steam cracking experiment was simulated with operating conditions close to the experimental process conditions mentioned in paragraph 4.1. The temperature profiles in several cross sections of the reactor are shown in Figure 5. In the x-plane and y-plane cross section, the temperature in the quartz is also shown, in the xy-plane cross section, only the temperature of the process gas is shown. The voids are the locations of the coupon and the thermocouple well. The temperature gradients in the bulk of the reactor are quite small with a difference between highest and lowest bulk temperature of about 20 K. The temperature of the process gas surrounding the thermocouple well is close to the temperature of the process gas near the coupon indicating that the measured temperature is a reliable value for the coupon temperature. AICHE_Catalytic_Coating_Apr2014.docx Page 3

4 Figure 5: Temperature [K] in the x-plane (top), xy plane (middle) and y-plane (bottom). The uniformity near the coupon surface is illustrated by the distribution functions of temperature and concentration at the coupon surface shown in Figure 7. The temperature is narrowly distributed around 1124 K while the ethene mass fraction is also distributed around wt% with small absolute deviations of about wt%. 3.2 Pilot plant A schematic representation of the pilot plant is shown in Figure 8. Since the main parts of this unit, the analytical equipment and the calibration procedure have been described elsewhere, only a brief description, based on Wang et al. 7, Dhuyvetter et al. 8 and Van Geem et al. 9 will be given. The feed section controls the supply of the different feedstocks to the reactor coil. The flow is regulated by the pumping frequency of a CORI-FLOW pump. A similar analysis is made for the mass fraction of ethene using the same cross- sections of the reactor as shown in Figure 6. The x-plane cross-section illustrates the jets where lower ethene mass fractions are simulated as ethane conversion is still low here. In general, the concentration gradients in the bulk of the reactor are small, the regions affected by the jets nonincluded. Furthermore, the ethene mass fraction measured at the reactor outlet is close to the ethene mass fraction near the coupon. Figure 6: Page 4 Ethene weight fraction [wt % wet, i.e. steam included] in the x-plane (top), xy plane (middle) and y-plane (bottom). Figure 7: Area-weighted distribution of temperature [left] and mass fraction ethene [wt% wet, i.e. steam included] at coupon surface. The mass flow rate of all feeds is measured continuously instead of the volume flow in order to AICHE_Catalytic_Coating_Apr2014.docx

5 avoid inaccuracies due to volume dependency on temperature and pressure. The pumping frequency is automatically adjusted by the controller of the CORI- FLOW. When the fluid level in the intermediate barrel is too low, the barrel is automatically refilled with feed from the storage barrels. Gaseous hydrocarbons are fed to the reactor directly from the gas bottle by means of a separate CORI- FLOW. The furnace, built of silica/alumina brick (Li23), is about 4 m long, 0.7 m wide and 2.6 m high. It is fired by means of ninety premixed gas burners, mounted with automatic fire checks and arranged on the side walls in such a way as to provide a uniform heat distribution. The fuel supply system comprises a combustion controller, for the regulation of the fuel to air ratio, and the usual safety devices. The furnace is divided into seven separate cells, which can be fired independently to set any type of temperature profile. Twenty thermocouples and eight manometers are located along the reactor coil to measure the temperature and pressure of the process gas. The reaction section of the tube is about 12m long, made of Incoloy 800HT, and has an internal diameter of 9mm. These dimensions are chosen to achieve turbulent flow conditions in the coil with reasonable feed flow rates (Re>7000). At the reactor outlet, the injection of nitrogen provides an internal standard and contributes to a certain extent to the quenching. Before the cooling of the effluent, a sample is taken for the on-linec + analysis. The liquid and the tar are separated from the cooler exit flow by means of a knock-out vessel and a cyclone. The pressure at the exit of the reactor is controlled by a reduction valve. A fraction of the product gas is then withdrawn for online C - analysis, while the rest of the effluent is sent directly to the flare. 4. experimental procedures 4.1 Jet stirred reactor The experiments carried out in the JSR setup had three main steps. These are pre-oxidation, cracking and decoking. Each one is described in detail below. Figure 9 summarizes the timeline of the followed procedure, and indicates the main parameters of each stage. Figure 8: Schematic representation of the pilot plant unit. AICHE_Catalytic_Coating_Apr2014.docx Page 5

6 Figure 9: Timeline of the coking decoking experiments in the JSR setup: the blue line represents a typical weight signal. In the pre-oxidation step, the samples were first oxidized in-situ prior to the cracking runs to mimic the surface state of an industrial cracking coil. For that purpose, the reactor temperature was raised to 1023 K with a heating ramp of 300 K/h and a constant N2 flow ( Nl/s). Once this temperature was reached, the feed to the reactor was switched to a constant flow of air only ( Nl/s). This pre-oxidation lasted hours, after which, keeping the temperature constant at 1023 K, N 2 was fed again to the reactor ( Nl/s). To start a cracking run, the temperature of the reactor was raised to 1173 K, with the same N 2 flow as before ( Nl/s). This in order to have a reference state in which the weight of the sample could be measured before and after every cracking run. After the weight of the sample was recorded, the reactor was further heated to 1283 K. As the heating proceeded, water with DMDS ( kg/s) and ethane ( Nl/s) are being fed to the evaporators (dilution δ = 0.33 kgh 2 O/kgC 2 H 6 ) and sent to the vent, in order to get a steady evaporation and mixing before sending this cracking mixture to the reactor. Once the reactor temperature was stable at 1283 K, the cracking mixture is send to the reactor. The nitrogen flow is used as internal standard for the chromatographic analyses. The cracking runs lasted for 6 hours, throughout which the reactor gas temperature was controlled to 1159 K yielding an ethane conversion around 70%. During the cracking runs, several online injections on the gas chromatographs were performed to analyze the reactor effluent gas. For quantification, the internal standard method was used. 10 When the 6 hours of cracking were completed, the flow rate of ethane and steam was stopped, leaving only nitrogen to enter the reactor. At the same time, the reactor temperature was set to 1173 K. Once the set temperature was reached, the weight of the sample was registered, to calculate the weight difference between the start and the end of the cracking run, thus verifying that the obtained coking curve provided reliable data. Afterwards, the reactor was cooled down to 1023 K with a steam flow rate of kg/s, and Page 6 once that temperature was reached, a mix of air ( Nl/s) and nitrogen ( Nl/s) was fed to the reactor. At the same time that this mix started flowing to the reactor, the temperature of the reactor was set to 1173K again, using a heating ramp of 300 K/h. As soon as the reactor reached 1173 K, the air flow was maintained, but the nitrogen was switched off to also mimic the industrial decoking practice. These conditions were kept for 15 minutes, and then the feed to the reactor was switched back to only N 2 ( Nl/s). Finally, and as an overnight mode, the reactor was cooled down to 1023 K with N 2 flowing through, and kept like that until the next cracking run would start. Once the third cycle was completed, the reactor was cooled down to room temperature instead of going to the overnight mode. The samples were rarely decoked after 3 coking cycles, to have coupons with coke on their surface for further SEM and EDX analyses. 4.2 Pilot plant The experiments carried out in the pilot plant also consisted of the same three main steps; preoxidation, cracking and decoking. Prior to a cracking experiment, the reactor was pre-oxidized with a steam/air mixture. The process conditions of preoxidation are summarized in Table 1. Table 1: Process conditions during pre-oxidation. FH 2O Fair (kg/h) (kg/h) ( C) Cell 1 Cell 2 Cell 3 Cell 4 ( C) ( C) ( C) Cell 5 Cell 6 Cell 7 ( C) ( C) ( C) Pre-ox Prior to the introduction of the feed, the reactor was heated under a flow of steam of 2.0 kg/h until the desired temperature profile is reached. Table 2 summarizes the main process conditions of the various ethane and naphtha experiments. The effect of a higher was evaluated in one experiment using the Ethane conditions. AICHE_Catalytic_Coating_Apr2014.docx

7 Table 2: Pyrolysis conditions. Ethane Ethane Naphtha HC flow rate (g/h) Steam flow rate (g/h) COP (bar) Duration (h) CIT cell 3 ( C) cell 3 ( C) cell 4 ( C) cell 5 ( C) cell 6 ( C) cell 7 ( C) After heating up, the flow rate of steam is set to the desired value for cracking and the hydrocarbon feedstock is introduced. Upon the introduction of hydrocarbons the temperature of the cracking coil decreases by about 20 C due to the endothermic nature of the cracking reactions. After about 20 minutes the temperature of the cracking coil reaches the set values. The cracking runs of all experiments lasted for 6 hours starting from hydrocarbon introduction in the reactor. The process conditions used in all tests are specified in Table 2. In the experiments where continuous DMDS addition was applied, the DMDS was added in the water feed barrel to provide the desired concentration. For the ethane experiments, the influence of (steam) dilution was investigated. In DIL1, the effect of lower steam partial pressure was investigated. The water mass flow rate of steam was halved to 578 g/h. However to provide the same molar dilution and space time, nitrogen was added. To maintain the total molar inlet flow, a nitrogen flow rate of 1270 g/h was necessary. In DIL2, the effect of lower total dilution was investigated. The dilution was halved to kg/kg. To maintain the same total molar inlet flow, the ethane and steam flow rate were scaled to 3729 and 718 g/h respectively. After 6 hours of cracking, the temperature of cell 4 to 7 is set to 800 C under a N 2 flow 1 kg/h. To start decoking, a mixture of steam/air is introduced to the coil. The conditions for decoking are given in Table 3. The CO and CO 2 are continuously measured by means of an infrared meter. When the concentration of CO 2 in the effluent gas is lower than 1 mol %, the temperature from cell 4 to cell 7 is increased to 850 C. When the concentration of CO 2 in the effluent gas is lower than 0.1 mol %, steam is stopped, only air is used. Furthermore a filter is installed in the condensers after the reactor (in Figure 8 depicted as (7)), where entrained coke is collected. After each experiment, this collected coke is dried and weighed. Table 3: Decoking conditions. FH 2 O Fair FN 2 Cell 3 Cell 4 Cell 5 Cell 6 Cell 7 Pre-start Start CO 2 <1vol% CO 2 <0.1vol% experimental results 5.1 Jet stirred reactor Three different coating formulations were tested and compared to an uncoated reference alloy, i.e. Incoloy 800HT. Table 4 summarizes the coke and yield data of the experiments performed. The BCZ1 coating shows an increase in ethane conversion and CO 2 and CO yields compared to the reference alloy, Incoloy 800HT. The other coatings show similar yields compared to the reference experiment. The coking rate is significantly reduced for all coating formulations compared to the reference alloy. For the first cycle the coking rate is reduced by a factor of 7.8, 3.9 and 3.8 compared to the reference alloy for BCZ1, BCZ2 and 5BCZ respectively. This improvement of performance was obtained by gasification of the cokes by the coating. Consequently, the effluent during cracking over BCZ1 contained more CO and CO 2 than during the reference experiment. AICHE_Catalytic_Coating_Apr2014.docx Page 7

8 The initial coking rate is taken as an indication of the deposition of catalytic coke and is calculated as the rate between cracking times t1 = 15 min and t2 = 30 min. The asymptotic rate, on the other hand, is calculated between t1 = 5 hours and t2 = 6 hours, when the mass increase reaches a stable linear regime. As can be seen from Figure 10 both the initial and asymptotic coking rates for all experiments are seen to increase over the number of cycles, which is attributed to an increase in surface roughness. Table 4: Summary of cokes and yield data of jet stirred reactor experiments. Coupons Cycles Incoloy 800HT BCZ1 BCZ2 5BCZ Coke gain [mg] over 6 hours of cracking 1st nd rd Species Yield* (wt %) H CO CH CO C 2 H C 2 H C 3 H C 3 H C 2 H ,3-C 4 H Benzene * Average over all cycles with analyses per cycle. To be performed Figure 10: Comparison of the initial and asymptotic coking rates 5.2 Pilot plant As the BCZ1 coating showed the lowest coking rate of all tested coating formulations, this coating was tested in the pilot plant unit. The reactor inner surface of all tubing used in cell 3 to cell 7 was coated with the BCZ1 coating and compared to a reference reactor made out of Incoloy 800HT Ethane experiments The results of the experiments with ethane as feedstock are summarized in Table 5. Seven experiments were performed evaluating the effect of sulfur addition, coil-outlet temperature and dilution on coking rate and product yields. The total coke has two contributions: the coke burnt off from the reactor during decoke measured by the infrared CO/CO 2 meter and the weighed amount collected in the filter downstream of the reactor after cracking. Due to the high CO 2 and CO yields, normalization was performed based on the carbon balance instead of the total mass balance. Comparing the reference reactor (INC) and the coated reactor (BCZ1), it is seen that the amount of coke deposited was reduced by a factor of 4.4 by application of the BCZ1 coating compared to the Incoloy 800HT reactor. Consequently, the effluent contained more hydrogen, CO and CO 2 than during the INC experiment as cokes are gasified to carbon oxides and hydrogen. The increase in hydrogen and carbon oxide yields by application of the BCZ1 coating Page 8 AICHE_Catalytic_Coating_Apr2014.docx

9 is much higher in the pilot plant experiments than in the JSR experiments due to the larger surface-to- volume ratio of the pilot plant reactor. The surface-to-volume ratio is 5.1 m -1 and m -1 for the JSR and pilot reactor respectively. Although operated at a similar conversion, the ethylene and 1,3-butadiene selectivity are slightly reduced by application of the BCZ1 coating as was also seen during the JSR experiments. Table 5: Summary of cokes and yield data of pilot plant ethane experiments. Experiment INC BCZ1 DMDS PRES DIL1 DIL2 Process conditions Reactor Incoloy Coated Coated Coated Coated Coated Coated Feed [g/h] H 2 O flow rate [g/h] N 2 flow rate [g/h] S addition [ppm S/g HC] H 2 O/HC ratio [g/g] (H 2 O+N 2 )/HC ratio [g/g] (H 2 O+N 2 )/HC ratio [mol/mol] [ C] COP [bar abs] Yields* [wt%] C Ethylene selectivity Ethane conversion C0-C4 species H CH C 2 H C 2 H ,3-C 4 H CO@ CO CO Coke formation From reactor [g coke/6h] In filter [g coke/6h] N.D N.D. Total coke [g coke/6h] * Average over analyses per experiment. AICHE_Catalytic_Coating_Apr2014.docx Page 9

10 The absence of continuous addition of DMDS was evaluated in experiment DMDS. The influence on hydrocarbon species yields is minor. The increase in CO and CO 2 yield is much more apparent; CO increases from 1.7 to 3.7 wt% and CO2 from 2.2 to 3.6 wt%. Hence continuous DMDS addition can mitigate CO and CO2 production when the coating is applied. Presulfidization of the coil with a steam/dmds solution prior to the continuous DMDS addition was evaluated in experiment PRES and shows similar influence on yields compared to solely continuous DMDS addition in experiment BCZ1. As expected, higher amounts of cokes were measured by continuous DMDS addition. Presulfidization before continuous DMDS addition (3.3 g/6h) shows a coke tendency lower than to solely continuous DMDS addition (4.4 g/6h). The increase of the coil-outlet-temperature to 870 C with continuous DMDS addition (experiment ) shows the expected yield results: as ethane conversion increases, yields of methane, ethylene and 1,3- butadiene increase. The yields of CO and CO 2 increase compared to the BCZ1 experiment; CO from 1.75 to 3.3 wt% and CO 2 from 2.2 to 3.4 wt%. This can be attributed to higher coke formation and more CO and CO 2 production by the coating. Surprisingly the higher experiment showed lower coke formation than the BCZ1 experiment (from 4.4 to 1.4 g/6h). This means that the coating converts more coke at higher temperatures as higher coke formation is to be expected at a higher. The higher CO and CO 2 yield in this experiment confirm this argument. This means that longer run lengths are possible at higher in an industrial cracker. In experiment DIL2 the steam dilution was also halved to g steam/g ethane. No nitrogen was added. To keep the same space time, the mass flow rate of ethane and steam were scaled to keep the same molar flow rate. Lower ethane conversion and resulting lower olefin yields were measured. The CO and CO 2 yields decreased to 0.99 and 0.98 wt% respectively. In experiment DIL2 a higher coking rate was measured (6.1 g/h) compared to BCZ1. Higher coke formation is expected due to higher hydrocarbons partial pressure and the coating converting less coke to CO and CO 2 due to the lower steam partial pressure Naphtha experiments The results of the experiments with naphtha as feedstock are summarized in Table 6. It is seen that the amount of coke deposited was reduced by a factor of 2 by application of the BCZ1 coating compared to the Incoloy 800HT reactor. Consequently, the effluent contained more CO and CO 2 than during the INC experiment as coke is gasified. The cracking severity is lower in experiment BCZ1, although this can mainly be attributed to a slightly different process gas temperature profile. Finally, the effect of lower dilution was evaluated. The effect of dilution was evaluated in two experiments: DIL1 and DIL2. In experiment DIL1 the steam mass flow rate was halved and nitrogen was added to remain at the same total mole flow to cancel out the effect of reduced space time on product yields. As DMDS was continuously added, comparison is made to the BCZ1 experiment. The yields of CO and CO 2 were lower; CO from 1.75 to 1.65 wt% and CO 2 from 2.4 to 1.0 wt%. Hence, the reduction of steam seems mainly to affect the conversion of CO to CO 2. The low dilution experiment DIL1 shows a coking rate similar to BCZ1. Hence, it seems enough water is present to convert coke to carbon oxides as in experiment BCZ1. Page 10 AICHE_Catalytic_Coating_Apr2014.docx

11 Table 6: Summary of cokes and yield data of pilot plant naphtha experiments. Experiment INC BCZ1 Process conditions Reactor Incoloy Coated Feed [g/h] H 2O flow rate [g/h] S addition [ppm S/g HC] H 2O/HC ratio [g/g] [ C] COP [bar abs] Yields [wt%] C P/E ratio C0-C4 species H CH C 2H C 3H ,3-C 4H CO CO coils are suspended side by side. Each coil makes eight passes. The coils have a larger diameter in the two last passes, compared to the first six. The process gas enters the coils at both end sides and in the middle of the furnace and flows downwards. As the furnaces is symmetrical, only one half is simulated. The maximum allowable tube metal temperature for the reactor alloy is around 1070 C. If this temperature is exceeded, production is halted and decoking is started. The characteristics of the furnace, reactor coils and material properties are summarized in Table 7. C5+ species Benzene Toluene C5-C9 (BTX excluded) Coke formation From reactor [g coke/6h] In filter [g coke/6h] N.D. N.D. Total coke [g coke/6h] simulation of an industrial ethane cracker To evaluate the effect of application of the BCZ1 coating on runlength and product yields in an industrial unit, two coupled reactor-furnace runlength simulations were performed using the in-house developed COILSIM1D 11 and FURNACE codes; with and without application of the coating BCZ1 respectively. First the industrial unit is described. Afterwards the main results of both simulations are summarized. 6.1 Description of the industrial ethane cracker Figure 11 shows a top view of half of the furnace. The furnace is rectangular. In the center of the furnace four Figure 11: Top view of half of the furnace. In the side walls of the furnace 128 radiation burners are placed, 64 on each side. The total fuel gas flow for the 128 burners is adjusted over time to maintain an ethane conversion of 65% averaged over the four reactors. The air excess is 2% (i.e. excess to the stoichiometric combustion). The burner cup temperature and the temperature of the flue gas entering the furnace is calculated using the method developed by Plehiers 14. The composition of the flue gas is derived from the stoichiometry of the combustion reactions. The composition of the feed and the operating conditions of the reactor are listed in Table 8. AICHE_Catalytic_Coating_Apr2014.docx Page 11

12 Table 7: Characteristics of the furnace. Furnace Length m Height m Depth m Thickness refractory material m Thickness insulation material m Number of burners Reactor coil Number of reactors 4 - Type swaged coils - # passes 8 - Total length m Internal diameter pass m pass m External diameter pass m pass m Tube wall thickness m Table 8: Feedstock composition and operating conditions of the reactor coils. Composition of the feed Ethane wt% Reactor operating conditions Total hydrocarbon flow 14 ton hr-1 Inlet temperature 600 C COP 1.8 bar abs Steam dilution 0.35 kg steam/kg feed Ethane conversion 65 wt% 6.2 Results Table 9 compares the average values over all reactors for the uncoated and coated case at start-ofrun (SOR) and end-of-run (EOR) conditions. The most important differences are discussed here. The runlength increases from 47 days to 214 days, i.e. by a factor Assuming 24h for a decoking operation of the furnace, 7.6 and 1.7 cracking-decoking cycles are possible within 1 year of operation. Hence, by adoption of the coating = 5.9 extra days of production are available per year. Moreover the lower energy input need on a yearly basis due to a reduction of the number coking/decoking cycles will influence the cracker economics beneficially. Page 12 AICHE_Catalytic_Coating_Apr2014.docx

13 Table 9: Average results for the coated and uncoated reactors at SOR and EOR. Uncoated BCZ1 coated SOR EOR SOR EOR Coil-inlet pressure [bar abs] Coil-outlet-temperature [ C] Residence time [s] Maximum tube temperature [ C] Maximum coke thickness [mm] Ethane conversion [wt%] Runlength [days] Yields [wt%] H CH CO CO C 2 H C 2 H C 2 H C c H C 3 H ,3-C 4 H n-c 4 H Benzene On the down-side, the yield of carbon oxides increases by application of the coating. The carbon monoxide yield increases from wt% for the uncoated reactor to wt% for the coated reactor. The carbon dioxide yield increases from wt% for the uncoated reactor to wt% for the coated reactor. The yield of carbon oxides by application of the coating is significantly less than in the pilot plant experiments due to the lower surface-area-to-volume ratio in these industrial reactors. Furthermore, reduced yields of methane and ethylene are simulated by application of the coating, consistent with the pilot plant experiments. The loss of ethylene yield is 0.41 wt% and 0.65 wt% at SOR and EOR respectively. In the following more details are given that allow to compare the results of the coated and uncoated case. Comparison is made for Reactor 1. All conclusions for the other reactors are similar to those for Reactor 1, unless stated otherwise. Figure 12 shows the tube external wall temperature profile of Reactor 1 for the uncoated (top) and coated (bottom) case. As seen from the top figure, the maximum allowable temperature of 1070 C is reached after 1125 h, i.e. 47 days, of operation. For the coated reactor, the maximum tube metal temperature has then only reached C due to much lower coke formation. Only after 5125h, i.e. 213 days the TMT surpasses 1070 C. Figure 13 shows the coke thickness profile for the uncoated (top) and coated (bottom) case. Obviously, the coke layer grows much quicker for the uncoated reactor. At EOR for the uncoated reactor, a maximum coke thickness of 14 mm is reached. The maximum coke thickness in the coated reactor is then only 4 mm. AICHE_Catalytic_Coating_Apr2014.docx Page 13

14 This slower growth of the coke layer makes that the pressure drop over the coated reactor increases slower as seen from Figure 14. Figure 12: Heat flux [W/m²] profile to the uncoated Reactor 1 (top) and the coated Reactor 1 (bottom). Figure 14: Process gas pressure [bar abs] profile for the uncoated Reactor 1 (top) and the coated Reactor 1 (bottom). Figure 13: Coke thickness [m] profile for the uncoated Reactor 1 (top) and the coated Reactor 1 (bottom). Figure 15 shows the ethylene yield for the uncoated (green) and coated (blue) case. Interestingly, the ethylene yield for the uncoated reactor decreases faster due to higher pressure drop because of the higher coke formation. After about 500h, the ethylene yield of the uncoated reactor is below the ethylene yield of the coated reactor. Over the entire runlength, the average ethylene yield is wt% and wt% for the uncoated and coated Reactor 1 respectively. However, if operation would be halted in the coated case after 1875h, an average ethylene yield of wt% is obtained. This allows an increase in runlength of a factor 1.67 while keeping the ethylene yield constant. This analysis shows that although the coating results in lower ethylene yields at SOR, the decrease of ethylene yield over time is different for a coated reactor. Hence, in some cases, a longer runlength can be obtained while keeping the timeaveraged ethylene yield the same as for an uncoated reactor. Page 14 AICHE_Catalytic_Coating_Apr2014.docx

15 8. references 1. Broutin, P.; Ropital, F.; Reyniers, M. F.; Froment, G. F., Anticoking coatings for high temperature petrochemical reactors. Oil & Gas Science and Technology - Revue d'ifp Energies Nouvelles 1999, 54, (3), Ganser, A.; Wynns, K. A.; Kurlekar, A., Operational experience with diffusion coatings on steam cracker tubes. Materials and Corrosion- Werkstoffe Und Korrosion 1999, 50, (12), Figure 15: Ethylene yield [wt%] as a function of run time for Reactor 1) for the uncoated (green) and coated (blue) case. 7. conclusions YieldUp, a robust, novel coating technology for the inner wall of the furnace tube that prevents coke deposition was developed. The performance of the coating was tested in a Jet Stirred Reactor (JSR) set-up and at the UGent pilot plant for steam cracking. Three different coating formulations were tested in the JSR set-up. For all of the coated coupons, a decrease in coke formation during cracking was observed compared to Incoloy 800HT. It was shown that by tuning the coating formulation, the activity and corresponding CO and CO2 yields can be optimized. The coating that showed the lowest coking rate during the JSR experiments was tested in a pilot plant unit using ethane and naphtha as feedstock. Application of the coating resulted in a coke reduction by a factor 2 to 4 compared to an uncoated reference reactor. Consequently, the effluent contained more hydrogen, CO and CO2 than during the reference experiments as coke is gasified to carbon oxides and hydrogen. The effects of continuous sulfur addition and presulfidation, coil-outlet-temperature and dilution on product yields and coking tendency were also tested. These pilot experiments showed that the catalyst is robust and maintains anti-coking activity even after >10 coking/decoking cycles. 3. Petrone, S.; Chen, Y.; Deuis, R.; Benum, L.; Gent, D.; Saunders, R.; Wong, C. In Catalyzed-assisted Manufacture of Olefins (CAMOL): Realizing Novel Operational Benefits from Furnace Coil Surfaces, AIChE 2008 Spring National Meeting, New Orleans, Louisiana, 2008; New Orleans, Louisiana, Pyl, S. P.; Schietekat, C. M.; Reyniers, M.-F.; Abhari, R.; Marin, G. B.; Van Geem, K. M., Biomass to olefins: Cracking of renewable naphtha. Chemical Engineering Journal 2011, 176, D. Matras, V. J., Continuous reactor perfectly agitated by gas jets for kinetic study on rapid chemical reactions. Chem. Eng. Sci 1973, 28, Plehiers, P. M.; Froment, G. F., Firebox simulation of olefin units. 7. Chemical Engineering Communications 1989, 80, Rao, M. V. R.; Plehiers, P. M.; Froment, G. F., The coupled simulation of heat-transfer and reaction in a pyrolysis furnace. Chemical Engineering Science 1988, 43, (6), Plehiers, P. M.; Reyniers, G. C.; Froment, G. F., Simulation of the run length of an ethane cracking furnace. Industrial & Engineering Chemistry Research 1990, 29, (4), Banded gas and nongray surface radiation models for high-emissivity coatings. Aiche J. 2005, 51, (10), AICHE_Catalytic_Coating_Apr2014.docx Page 15