Towards a comprehensive DME/propane blended combustion kinetic model

Transcription

1 Sub Topic: Reaction Kinetics 9 th U. S. National Combustion Meeting Organized by the Central States Section of the Combustion Institute May 17-20, 2015 Cincinnati, Ohio Towards a comprehensive DME/propane blended combustion kinetic model Enoch E. Dames 1, Bryan W. Weber 2, Andrew Rosen 1, Connie W. Gao 1, Chih-Jen Sung 2, William H. Green 1 1 Department of Chemical Engineering, MIT, Cambridge, MA Department of Mechanical Engineering, University of Connecticut, Storrs, CT Abstract: Tomorrow's transportation engines will utilize dual fuel concepts and operate at high pressures to achieve up to 60% thermal efficiency, outperforming current diesel and gasoline engine efficiencies. One proposed fuel blend for these advanced engines is dimethyl ether (DME) and propane. However, the cross-reactions of DME and propane intermediates at higher pressures relevant to advanced engines have not yet been explored. Knowledge of this chemistry is useful for determining fuel/air mixture fractions to ensure optimal engine performance, including heat release rates and emissions outputs. In this work, the Reaction Mechanism Generator (RMG) is used for detailed kinetic model development with focus on engine conditions (20-30 atm and K). The ignition-promoting effect of DME on propane is discussed, and it is shown that even relatively small quantities (5% by mole) of DME can significantly increase propane autoignition. In addition, new Rapid Compression Machine data at engine relevant pressure and temperature conditions are obtained for the purpose of discovering where the coupling of DME and propane kinetics considerably affects autoignition. 1. Introduction Dual fuel combustion embodies a conceptually novel approach to internal combustion (IC) engine technology, relying on mixtures of high-octane fuels and high-cetane fuels, as well as their synergistic combustion characteristics in order to achieve high thermal efficiencies. Commonly referred to as Reactivity Controlled Compression Ignition (RCCI), these engines are typically made from retrofitted diesel engines with modified intake manifolds to allow for high reactivity vapor fuel mixing into the engine piston. For example, recent work [1] has identified a mixture of 20% dimethyl ether (DME) with 30% propane in concert with diesel fuel combustion, resulting in over 50% brake thermal efficiency (BTE) in a modified rail diesel engine, compared to a baseline BTE of 37%. The authors found that propane substitution delayed DME s early autoignition and shifted the combustion process closer to top dead center (TDC) and that autoignition of the fumigated fuels led to the diesel fuel igniting much earlier compared to the baseline diesel condition due to an increase in the bulk cylinder temperature. Within the United States, DME has received recent attention in part due to the current natural gas boom. As a result, studies of DME combustion abound in the literature. Recent work includes that of Li et al. [2], who have performed DME autoignition studies in a shock tube at 23 bar and various fuel equivalence ratios, capturing the negative-temperature-coefficient region (NTC) of DME combustion. The authors compared their data with predictions from a previous model [3], finding excellent agreement. In addition, Liu et al. [2] investigated the effect of CO dilution on elevated-

2 pressure laminar burner stabilized premixed flames [4]; in comparing their experimental data with previously published models, the authors found that no model was able to reliably predict the highpressure flame speeds. Regarding the remaining uncertainties in low temperature DME chemistry, Tomlin et al. [5] have clearly illustrated the need for further work by showing that several literature models [4,6,7] fail to accurately reproduce earlier Rapid Compression Machine (RCM) ignition delay time measurements of DME [8]. In particular, Tomlin et al. [5] highlight the need for accurate pressure dependent descriptions for DME relevant QOOH + O 2 = OOQOOH and subsequent OH forming pathways. Fundamental combustion studies on DME/propane mixtures have also been recently performed, albeit to a lesser extent than for each individual component. Hu et al. [9] have studied DME/propane autoignition at engine-relevant conditions (20 bar) by measuring ignition delay times behind reflected shock waves at high temperatures ( K). In a related study, Burke et al. [10] performed a combined experimental and modeling study of DME/methane mixtures between K and 7-41 atm for varying fuel equivalence ratios. The authors performed rate theory calculations to elucidate the pressure dependent nature of CH 3 OCH 2 unimolecular decomposition, finding that this chemistry, in addition to pressure dependent low temperature DME chemistry adopted from Yamada et al. [11] resulted in improvements in model agreement with various experimental data compared to a modeling using only pressure independent expressions. The authors also point out the strong promoting effect small amounts of DME can have on overall ignition delays of 20:80 blends of DME:CH 4 [11]. This promoting effect is attributed to the overall weaker C-H bonds of DME compared to that of methane. Although reactions between dual fuel derivatives/intermediates are expected under engine conditions, no current model incorporating both DME and propane reaction kinetics includes these fuel intermediate cross reactions. The purpose of this preliminary work is to investigate the capability of the Reaction Mechanism Generator (RMG)[12] to identify - if any - reactions important to blended fuel combustion chemistry that do not exist in a pre-constructed based model which is a concatenation of DME and propane mechanisms. This capability can be especially valuable in the context of dual- or multi-component fuel mixtures given that many potentially important crossreactions between fuel components may be readily missed or over-looked during manual (i.e., human) mechanism construction. 2. Computational and Experimental Methodologies 2.1 Computational Methodology A base model (hereafter described as Model I) of 136 species and 748 reactions was assembled from a variety of sources. The low temperature chemistry of DME was adopted from recent work by Eskola et al. [13]. The corresponding low temperature chemistry of propane was adopted from Goldsmith et al. [14]. Last, a base mechanism for H 2 /CO/C 1 -C 4 fuels was adopted from Dames at al. [15]. The remaining DME chemistry was adopted from Burke et al. [10]. The development Python version 3.0 (see of the MIT Reaction Mechanism Generator [12] was used for automated kinetic model construction. Details of RMG s algorithm and methods can be found elsewhere [16] and are described briefly here. The 2

3 model is grown using the rate-based algorithm developed by Susnow et al. [17] rate-based algorithm, where RMG simulates isothermal, isobaric reactor conditions with a user-given set of initial core species, finding all possible reactions between them using a set of known reaction family templates. At each integration step, RMG evaluates whether a new edge species should be added to the model based on its flux and the user-specified flux tolerance. If the species is added to the model, it is reacted with all the existing species in the model to generate a new set of edge species and reactions, and the process is repeated until the model satisfies the user-given end conditions. Two approaches were considered. In one approach, Model I was designated as reaction library, or kinetics depository for RMG, overriding RMG s native kinetic estimation scheme. If the RMG model growth algorithm detects species and reactions satisfying the user-defined flux tolerance from the DME reaction library, those reactions will be included in the final mechanism. Thus, not all reactions in Model I enter into the final RMG model, which contains 83 species and 1289 reactions, and is hereafter referred to as Model II. In a second approach, all of the reactions in Model I were instead forced to be in the final RMG model, referred to hereafter as Model III. This model is substantially larger in size, with 220 species and 5115 reactions. Aspects of the three models used throughout this work are summarized in Table 1. Reactor conditions similar to those in Table 2 - but also extended in temperature down to 550 K and up to 850 K - were used as RMG input reactor parameters to generate Models II and III. In addition, pressure-dependence in RMG model generation was not turned on. The Python 2.7 interface of Cantera [18] version was used for all [isochoric adiabatic batch reactor] simulations in this work. 2.2 Experimental Methodology The present experiments are conducted in a heated RCM at the University of Connecticut [19]. The RCM is a single-piston, pneumatically-driven, hydraulically-stopped arrangement with compression times near 30 ms. The end of compression (EOC) temperature and pressure conditions, and respectively, are independently variable by varying the compression ratio, initial pressure, and initial temperature of the experiments. The primary diagnostic on the RCM is the in-cylinder pressure measured during and after compression. The compression stroke of the RCM brings the homogenous fuel/oxidizer mixture to the EOC conditions, and for suitable values of and, the mixture will ignite after some delay. This time period, called the ignition delay, is measured as the time difference between the EOC ( ) and the maximum of the time derivative of the pressure trace, as shown in Figure 1. 3

4 Figure 1. Definition of ignition delay in the RCM experiments. The experiment shown is for a 50:50 (by mole) mixture of DME/propane in stoichiometric O 2 /N 2 air. In addition to the reactive experiments, non-reactive experiments are carried out to determine the influence of the machine-specific operating parameters on the experimental conditions. In these nonreactive experiments, O 2 in the oxidizer is replaced with N 2 to maintain a similar specific heat ratio but suppress oxidation reactions that lead to thermal runaway. If the pressure at the EOC of the nonreactive experiments matches that at the EOC of the reactive experiments, it is assumed that no substantial heat release has occurred during the compression stroke, and the temperature at the EOC can be estimated by applying the adiabatic core hypothesis and the isentropic relations between pressure and temperature during the compression stroke. The experiments in this work are carried out for an equivalence ratio of of a 50:50 mixture (by mole) of DME/propane in O 2 /N 2 air, at EOC pressure of bar, for EOC temperatures in the range of 660K to 720K. Further experiments are planned at other equivalence ratios and DME/propane ratios. 3. Results and Discussion Although most of focus of this work lies in the lower temperature (T < 1000 K) regime, selected shock tube simulations were conducted at high temperatures for DME, propane, and their mixtures to verify that the base Model I performs well under such conditions. Figure 2 illustrates comparisons of Model 1 with the recent work of Hu et al. [9]. Table 2 shows the ignition delays measured in this study in the RCM. It can be seen that the ignition delays increase monotonically with decreasing temperature, indicating that the NTC behavior of the ignition delay was not found for any conditions considered in these experiments. This is in contrast to the work of Mittal et al. [8], who found strong NTC behavior for pure DME ignition delays under slightly lower EOC pressure conditions. In addition, no instances of two-stage ignition were found in the present experiments for the stoichiometric 50:50 fuel blend. Table 1. Summary of kinetic models used in this work. N reactions,rmg refers to reactions 4

5 discovered by RMG and included in the final model that were not in the base model. Model N species N reactions N reactions,rmg Model I n/a Model II Model III Table 2: Experimental conditions measured in this study. The fuel considered was a 50:50 mixture of DME/propane in the stoichiometric ratio with O 2 /N 2 air. Initial Temperature (K) Initial Pressure (bar) Compression Time (ms) Compressed Pressure (bar) Compressed Temperature (K) 1000/Tc (1/K) Ignition Delay (ms) Ignition Delay Error (ms) The total number of species and reactions for each model are shown in Table 2. As mentioned above, Model II utilized the base model (Model I) as a reaction library. As a result, only 197 reactions of Model I were adopted into Model II. Many of the reactions that were not adopted by RMG in Model II are those in the C 1 -C 4 section of Model I. The majority of the reaction added into Model II by RMG are H-abstraction reactions (~400), disproportionation reactions (~300) and O- substitution reactions (~230). 5

6 Ignition Delay ( s) Ignition Delay ( s) Ignition Delay ( s) C 3 H 8, P 5 = 20 bar K / T DME, P 5 = 20 bar K / T 50:50 DME:C 3 H 8, P 5 = 20 bar K / T Figure 2. Comparison of engine-relevant experimental [9] DME/propane/oxygen/argon autoignition delay times with Model I for varying mixture fractions and equivalence ratios. Lines: Model I predictions; symbols: experimental data, defined in top panel. Ignition delay defined as maximum [OH] gradient. Figure 3 below illustrates relative differences in the three models used in this work by plotting select species profiles for DME/air mixtures under pressure and temperature conditions similar to the RCM experiments conducted here. Qualitatively, all three models predict similar first and second stage ignition, indicated by the OH traces. However, the first [OH] peak of Model I is an order of magnitude lower than that of Models II and III, indicating that RMG has introduced more OHgenerating reactions. The OH plateaus of Models II and III are lower however. 6

7 HO 2 mole fraction OH mole fraction mole fraction DME OH HO Model I Model II Model III time, seconds Figure 3. Comparison of simulated OH and HO 2 species profiles using Models I, II, and III for an isochoric adiabatic reactor at 650 K and 30 atm, 50:50 DME:propane stoichiometric fuel/air blend. See Table 2 for model definitions. Solid lines: DME; dashed lines: OH; short-dashed lines: HO 2 (propane not shown). OH, HO 2, propane, and DME mole fractions are plotted as a function of simulation time and varying fuel blend ratios in Figure 4. Under the conditions simulated, propane alone is predicted to be unreactive. However, with addition of 25% DME, the mixture has similar behavior as the 100% DME mixture. With a DME:propane blending ratio of 25:75, the time of maximum [OH] gradient is nearly equal to that of the 100% DME mixture. However, the relative OH and HO 2 mole fraction peaks mimic the relative molar concentration of DME in the blend. OH profiles for fuel blended cases suggest subtle multi-stage ignition events are taking place: a first stage due to DME, a second stage again due to DME and a 3rd stage event due to propane's second stage ignition. In contrast, the 100% DME OH profile only exhibits the expected two stages of rapid OH rise. Also shown in Figure 4 are species mole fractions for a 05:95 DME:propane blend, where maximum OH and HO 2 mole fraction are orders of magnitude lower than other cases illustrated. Nonetheless, 90% of the propane is consumed after 0.1 seconds, compared to effectively none in the 100% propane simulation. The overall ignition delay time of the 05:95 DME:propane blend mixture is 0.1 s, much faster than the ignition delay time of the full propane mixture which is predicted to be on the order of seconds for these conditions (a) T = 650 K, P = 30 atm (b) DME:Propane 0:100 25:75 50:50 75:25 100:0 05: time, seconds time, seconds 7

8 mole fraction Propane mole fraction DME mole fraction (c) time, seconds (d) time, seconds Figure 4. Select species profiles plotted on a semi-log scale for varying DME/propane blending ratios (stoichiometric mixtures in air) at 650 K and 30 atm, simulated using Model II. Blend ratios are indicated in (a). HO 2 and OH mole fractions for 100% propane mixtures are too low to be seen in panels a and b. The promoting effect of DME on propane combustion is perhaps more apparent in Figure 5, which compares species mole fractions for a stoichiometric 05:95 DME:propane blend with a corresponding stoichiometric 100% propane/air mixture. Respective OH and HO 2 profiles differ by several or more orders of magnitude at later times, and multistage ignition events are evident in the 05:95 DME:propane blend HO 2 DME propane OH HO Black: 5% DME, 95% Propane Red: 100% Propane T 0 = 650 K, P 0 = 30 atm time, seconds Figure 5. Select species profiles plotted on a semi-log scale for a stoichiometric 100% propane/air mixture (red lines) and a 5:95 DME:propane blend in air (black lines) at 650 K and 30 atm, simulated using Model II. OH 8

9 As mentioned above, DME is used as a diesel surrogate, while propane can be used as a gasoline surrogate. Thus, DME auto ignites faster compared to propane for the same temperature and pressure. As a result, radical intermediates of DME can be expected to consume propane when the two fuels are mixed together. Although not considered here, a unique labeling of atoms in each fuel compound would allow for tracking the relative contribution of DME radical intermediates on propane consumption, something to be considered in future work. OH Sensitivity Spectrum 25:75 DME:propane in air = 1.0, T 0 = 650 K, P 0 = 30 atm DME+O2=CH3OCH2+HO2 OH+CO=H+CO2 ic3h7+ho2=ch3cho+oh+ch3 HO2CH2OCH2=CH3OCH2OO CH3OCH2OO=HO2CH2OCH2 O2+HO2CH2OCH2=OOCH2OCH2OOH C3H8+OH=iC3H7+H2O nc3h7+ho2=c2h5+oh+ch2o DME+OH=C2H5O+H2O OOCH2OCH2OOH=HO2CH2OCHO+OH OCH2OCHO+OH=HO2CH2OCHO Sensitivity Coefficient Figure 6. Sensitivity of OH mole fraction to reactions in Model II for a stoichiometric 25:75 DME:propane blend at 650 K and 30 atm. In looking at the sensitivity of OH mole fraction to the reactions in Model II for a stoichiometric 25:75 DME:propane blend initially at 650 K and 30 atm and at the ignition delay time (Figure 6), it is clear that the low temperature chemistry of DME is primarily responsible for OH generation. It can further be inferred that under the fuel blend conditions studied here, propane combustion is promoted as a result of low temperature DME chemistry through the following reactions: DME + O 2 = CH 3 OCH 2 + HO 2 CH 3 OCH 2 + O 2 = CH 3 OCH 2 OO (RO 2 ) CH 3 OCH 2 OO (RO 2 ) = HOOCH 2 OCH 2 (QOOH) O 2 + HOOCH 2 OCH 2 (QOOH) = OOCH 2 OCH 2 OOH (OOQOOH) OOCH 2 OCH 2 OOH = HOOCH2OCHO + OH HOOCH 2 OCHO = OCH 2 OCHO + OH C 3 H 8 + OH = (ic 3 H 7, nc 3 H 7 ) + H 2 O DME + OH = CH 3 OCH 2 + H 2 O 9

10 In the above sequence of reactions, DME reacts with molecular oxygen to produce the CH 3 OCH 2 radical, which then engages the low temperature autoignition sequence of DME. Unsurprisingly, a net gain of OH radicals results in H-abstraction from propane, generating iso-propyl and n-propyl radicals, and therefore engaging the low temperature autoignition chemistry of propane. As shown in Figure 5, even a 5:95 DME:propane blend ratio contains enough DME promote comparatively early propane autoignition. The strong promoting effect of DME on overall autoignition even in low quantities illustrates why this compound has been proposed as a fuel additive in automotive engines, in contrast to a designing new dual-fuel tank systems with high techno-economic barriers. Propane Sensitivity Spectrum 25:75 DME:propane in air = 1.0, T 0 = 650 K, P 0 = 30 atm OH+H2=H+H2O OH+HO2=O2+H2O HO2+HO2=O2+H2O2 OH+CO=H+CO2 O2+nC3H7=C3H6+HO2 O2+H(+M)=HO2(+M) O2+iC3H7=C3H6+HO2 ic3h7+ho2=c3h6+h2o2 HO2+HO2=O2+H2O2 nc3h7+ho2=c3h6+h2o2 C3H8+OH=nC3H7+H2O C3H8+OH=iC3H7+H2O ic3h7+ho2=ch3cho+oh+ch3 C2H5+CH3(+M)=C3H8(+M) nc3h7+ho2=c2h5+oh+ch2o Sensitivity Coefficient Figure 7. Sensitivity of propane mole fraction to reactions in Model II for a stoichiometric 25:75 DME:propane blend initially at 650 K and 30 atm and 50% propane consumption. On the other hand, the sensitivity spectra shown in Figure 7 illustrates the reactions important to propane consumption for a stoichiometric 25:75 DME:propane blend initially at 650 K and 30 atm - at this point in the reactor (50% propane consumption) where the temperature exceeds 1400 K. All of the DME has been consumed, and therefore no low temperature DME relevant species show up in the list; here, propane consumption is predicted to be solely dictated by high temperature chemistry. 4. Conclusions RMG was used in two different approaches to generate a model from a base model for the combustion of DME/propane blends. A combination of simulations and sensitivity analyses illustrated the promoting effect of DME kinetics on propane ignition, even from small relative molar quantities of DME. In addition, new data for the autoignition delays of 50:50 mixtures of 10

11 DME/propane by mole were collected in a heated RCM. For EOC conditions near P c = 30 bar and T c in the range of 650K to 720 K, and for the stoichiometric equivalence ratio in O 2 /N 2 air, no evidence of NTC behavior of the ignition delay and two-stage ignition behavior was found. Further experiments are required at other equivalence ratios and DME/propane mixing ratios to determine the cause of this result. Finally, cross reactions between low temperature intermediates of DME and propane were not found in any of the sensitivity analyses conducted in this work, suggesting models for low- and high- reactivity fuels may be constructed from decoupled models of their pure components and still accurately describe fuel-blended combustion characteristics under real engine conditions. 5. Acknowledgements The work at UConn was supported by the National Science Foundation under Grant No. CBET and as part of the Combustion Energy Frontier Research Center, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Award Number DE-SC

12 References [1] Prabhakar B, Jayaraman S, Vander Wal R, Boehman A. Experimental Studies of High Efficiency Combustion With Fumigation of Dimethyl Ether and Propane Into Diesel Engine Intake Air. J Eng Gas Turbines Power 2015;137: [2] Li Z, Wang W, Huang Z, Oehlschlaeger MA. Dimethyl Ether Autoignition at Engine-Relevant Conditions. Energy Fuels 2013;27: doi: /ef400293z. [3] Zhao Z, Chaos M, Kazakov A, Dryer FL. Thermal decomposition reaction and a comprehensive kinetic model of dimethyl ether. Int J Chem Kinet 2008;40:1 18. [4] Liu D, Santner J, Togbé C, Felsmann D, Koppmann J, Lackner A, et al. Flame structure and kinetic studies of carbon dioxide-diluted dimethyl ether flames at reduced and elevated pressures. Combust Flame 2013;160: [5] Tomlin AS, Agbro E, Nevrlý V, Dlabka J, Vašinek M. Evaluation of Combustion Mechanisms Using Global Uncertainty and Sensitivity Analyses: A Case Study for Low Temperature Dimethyl Ether Oxidation. Int J Chem Kinet 2014;46: [6] Zheng XL, Lu TF, Law CK, Westbrook CK, Curran HJ. Experimental and computational study of nonpremixed ignition of dimethyl ether in counterflow. Proc Combust Inst 2005;30: doi: /j.proci [7] Metcalfe WK, Burke SM, Ahmed SS, Curran HJ. A Hierarchical and Comparative Kinetic Modeling Study of C1 C2 Hydrocarbon and Oxygenated Fuels. Int J Chem Kinet 2013;45: doi: /kin [8] Mittal G, Chaos M, Sung C-J, Dryer FL. Dimethyl ether autoignition in a rapid compression machine: Experiments and chemical kinetic modeling. Fuel Process Technol 2008;89: [9] Hu E, Zhang Z, Pan L, Zhang J, Huang Z. Experimental and Modeling Study on Ignition Delay Times of Dimethyl Ether/Propane/Oxygen/Argon Mixtures at 20 bar. Energy Fuels 2013;27: [10] Burke U, Somers KP, O Toole P, Zinner CM, Marquet N, Bourque G, et al. An ignition delay and kinetic modeling study of methane, dimethyl ether, and their mixtures at high pressures. Combust Flame 2015;162: [11] Yamada T, Bozzelli JW, Lay TH. Comparisons of CBS-q and G2 calculations on thermodynamic properties, transition states, and kinetics of dimethyl-ether + O2 reaction system. Int J Chem Kinet 2000;32: doi: /(sici) (2000)32:7<435::aid-kin6>3.0.co;2-4. [12] Allen JW, Goldsmith CF, Green WH. Automatic estimation of pressure-dependent rate coefficients. Phys Chem Chem Phys 2012;14: doi: /c1cp22765c. [13] Eskola AJ, Carr SA, Shannon RJ, Wang B, Blitz MA, Pilling MJ, et al. Analysis of the Kinetics and Yields of OH Radical Production from the CH3OCH2+ O2 Reaction in the Temperature Range K: an Experimental and Computational study. J Phys Chem A 2014;118: [14] Goldsmith CF, Green WH, Klippenstein SJ. Role of O2+ QOOH in Low-Temperature Ignition of Propane. 1. Temperature and Pressure Dependent Rate Coefficients. J Phys Chem A 2012;116: [15] Dames EE, Lam K-Y, Davidson DF, Hanson RK. An improved kinetic mechanism for 3-pentanone pyrolysis and oxidation developed using multispecies time histories in shock-tubes. Combust Flame [16] Green Jr. WH. Predictive Kinetics: A New Approach for the 21st Century. In: Marin GB, editor. Adv. Chem. Eng., vol. 32, Academic Press; 2007, p [17] Susnow RG, Dean AM, Green WH, Peczak P, Broadbelt LJ. Rate-Based Construction of Kinetic Models for Complex Systems. J Phys Chem A 1997;101: doi: /jp [18] Goodwin DG. An open-source, extensible software suite for CVD process simulation. Chem Vap Depos XVI EUROCVD 2003;14: [19] Mittal G, Sung C-J. A Rapid Compression Machine for Chemical Kinetics Studies at Elevated Pressures and Temperatures. Combust Sci Technol 2007;179: doi: /