AME 514 Applications of Combustion Lecture 5: Microcombustion science II
2 AME Spring Lecture 5 Microscale reacting flows and power generation Micropower generation: what and why (Lecture 4) “Microcombustion science” (Lectures 4 - 5) Scaling considerations - flame quenching, friction, speed of sound, … Flameless & catalytic combustion Effects of heat recirculation Devices (Lecture 6) Thermoelectrics Fuel cells Microscale internal combustion engines Microscale propulsion »Gas turbine »Thermal transpiration
3 AME Spring Lecture 5 Heat recirculating combustor - minimizes heat losses - can be used as heat source for thermoelectric or other power generator Toroidal 3D geometry: further reduces losses - minimizes external T on all surfaces Heat recirculating combustors 1D counterflow heat exchanger and combustor 2D “Swiss roll” combustor (Lloyd & Weinberg, 1974, 1975) Cold reactants Hot Products Combustion zone Heat exchange
4 AME Spring Lecture 5 “Swiss roll” combustors - methods Use experiments to calibrate/verify CFD simulations at various Reynolds number (Re) Re Ud/ ; U = inlet velocity, d = channel width, = viscosity Key issues Extinction limits, especially at low Re Catalytic vs. gas-phase combustion Control of temperature, mixture & residence time for thermoelectric or solid oxide fuel cell generator (Lecture 6) Implementation of experiments 3.5 turn 2-D rectangular Swiss rolls PC control and data acquisition using LabView Mass flow controllers for fuel (propane) & air Thermocouples - 1 in each inlet & outlet turn (7 total) Bare metal Pt catalyst in center of burner
5 AME Spring Lecture 5 Mass Flow Controllers Air PC with LabView FuelO 2 or N 2 Flashback arrestor NI-DAQ board Gas ChromatographPC with PeakSimple Thermocouples Outgoing products Incoming reactants Swiss roll experiments
6 AME Spring Lecture 5 Swiss roll experiments 3.5 mm channel width, 0.5 mm wall thickness Top & bottom sealed with ceramic blanket insulation
7 AME Spring Lecture 5 Swiss roll experiments (Ahn et al., 2005)
8 AME Spring Lecture 5 Quenching limits Gas-phase extinction limits ≈ symmetrical about = 1 Minimum Re ≈ 40 Catalytic Low Re »Very low Re (≈ 1) possible »Lean limit rich of stoichiometric (!), limits very asymmetrical about = 1 - due to need for excess fuel to scrub O 2 from catalyst surface (consistent with computations - Lecture 4) »Conditioning Pt catalyst by burning NH 3 very beneficial, »Rearranging catalyst or 4x increase in area: practically no effect! - not transport limited Intermediate Re: only slight improvement with catalyst Still higher Re: no effect of catalyst Near stoichiometric, higher Re: strong combustion, heat recirculation not needed, reaction zone not centered, not stable (same result with or without catalyst)
9 AME Spring Lecture 5 Thermal characteristics - limit temps.
10 AME Spring Lecture 5 Thermal characteristics - limit temps. Much lower limit T with catalyst but only slightly leaner mixtures For a given mixture and Re supporting gas-phase combustion, catalyst actually hurts slightly - only helps when gas-phase fails Limit temperatures ≈ same lean & rich Limit temperatures down to 650˚C (non-cat), 125˚C (cat), 75˚C (!) (cat, with NH 3 treatment) Limit temperatures follow Arrhenius law Ln(Re limit ) ~ -Ln(residence time) ~ 1/T Activation energies ≈ 19 kcal/mole (gas-phase), 6.4 kcal/mole (catalytic) Mechanism At limit, heat loss ~ heat generation Heat loss ~ T max -T ∞ Heat generation ~ exp(-E/RT max ) ~ ∞ U ∞ AY f Q R Limit temperatures approx. ~ ln(U ∞ ) ~ ln(Re)
11 AME Spring Lecture 5 Thermal characteristics - limit temps. Temperatures across central region of combustor very uniform - measured maximum T is indicative of true maximum
12 AME Spring Lecture Thermocouple placements Out-of-center regime Lean or rich Maximum possible heat recirculation needed to obtain high enough T for reaction Flame centered Near-stoichiometric Heat recirculation not needed - flame self-sustaining Reaction zone moves toward inlet Center cool due to heat losses
13 AME Spring Lecture 5 Exhaust gas composition All cases: > 80% conversion of scarce reactant Low Re No CO or non-propane hydrocarbons found, even for ultra-rich mixtures! Only combustion products are CO 2 and (probably) H 2 O Additional catalyst has almost no effect NH 3 catalyst treatment increases fuel conversion substantially for very low Re cases Moderate Re Some CO formed in rich mixtures, less with catalyst High Re Catalyst ineffective, products same with or without catalyst
14 AME Spring Lecture 5 Exhaust gas composition
15 AME Spring Lecture 5 Scale-down experiments Wire-EDM fabrication, Pt igniter wire / catalyst Can’t reach as low Re as macroscale burner! Wall thick and has high thermal conductivity - loss mechanism! 2D mini Swiss Roll
16 AME Spring Lecture 5 Polymer combustors Theoretical study showed importance of wall thermal conductivity on combustor performance - counterintuitive: lower is better - heat transfer across thin wall is easy, but need to minimize streamwise conduction Low T max demonstrated in metal burners with catalytic combustion - no need for high-temperature metals (high k) or ceramics (k = W/m˚C but fragile, hard to fabricate) Use polymers??? Low k (DuPont Vespel SP-1 polyimide, k = 0.29 W/m˚C), rated to T > 400˚C, even in oxidizing atmosphere Easy to fabricate, not brittle
17 AME Spring Lecture 5 Plastic combustor - implementation World’s first all polymer combustors? (Sanford et al., 2008) CNC milling: 3.5 turn Swiss roll, 3 mm channel width, 0.5 mm wall thickness, 2.5 cm tall NH 3 -treated bare metal Pt catalyst in central region General performance No damage even at T > 400˚C (high enough for SOFCs) Thermal expansion coefficient of Vespel ≈ 4x inconel, but no warping Sustained combustion at 2.9 W thermal (birthday candle ≈ 50 W) Catalystregion 5.5 cm
18 AME Spring Lecture 5 Results - polymer burner - extinction limits Extinction limit behavior similar to metal burner at larger Re Improved “lean” and “rich” limit performance compared to macroscale burner at 2.5 < Re < 20 Sudden, as yet unexplained cutoff at Re ≈ 2.5 in polymer burner Sanford et al., 2008
19 AME Spring Lecture 5 Numerical model Kuo and Ronney, 2007 FLUENT, 2D, 2nd order upwind 32,000 cells, grid independence verified Conduction (solid & gas), convection (gas), radiation (solid-solid only, DO method, = 0.35) k- turbulence model - useful for qualitative evaluations but not quantitatively accurate for low Re 1-step chemistry, pre-exponential adjusted for agreement between model & expt. at Re = 1000 All gas & solid properties chosen to simulate inconel burner experiments Boundary conditions: Inlet: 300K, plug flow Outlet: pressure outlet Heat loss at boundaries + volumetric term to simulate heat loss in 3rd dimension
20 AME Spring Lecture 5 inletoutlet Numerical model d Thermocouple locations
21 AME Spring Lecture 5 User-Defined Function to simulate heat loss in 3rd dimension (includes radiation to ambient) Numerical model Intake Exhaust h = 10 W/m 2 K = 0.35 T1 Heat loss in 3 rd dimension blanket T_gas T_blanket T_plate T_wall T_ambient T_gas T_blanket T_plate T_outside T_ambient
22 AME Spring Lecture 5 Model results - comparison to experiment Temperatures too high to conduct experiments above this Re!
23 AME Spring Lecture 5 Reasonable agreement between model & experiment for all Re when turbulence included High-Re “blow-off” limit - insufficient residence time compared to chemical time scale At high Re, wider limits with turbulence - increases heat transfer (gas wall), thus heat recirculation At low Re, limits same with or without turbulence (reality check) Low-Re limit due to heat loss Heat generation ~ mass flow ~ U ~ Re Heat loss ~ (T max - T ambient ) ≈ const Heat loss / heat generation at low Re - need more fuel to avoid extinction Model & experiment show low-U limit at Re ≈ 40, even for stoichiometric mixture (nothing adjusted to get this agreement at low Re!) Model results - comparison to experiment
24 AME Spring Lecture 5 Model results - turbulence effects Extinction limit with laminar flow deviates from turbulent flow at higher Re Higher heat transfer coefficient (h ~ u’ ~ U) for turbulent flow vs. h = constant for laminar flow Adiabatic reactor temperature (homework…): If h ~ U ~, T reactor (thus limit Y fuel ) ≈ independent of U (thus independent of Re) Vital to include turbulence effects in macroscale model to obtain correct pre-exponential factor
25 AME Spring Lecture 5 Model results – temperatures at extinction T max T ad
26 AME Spring Lecture 5 Model results – temperatures at extinction “Virtual thermocouples” - 1 mm x 1 mm region at same locations at thermocouples in experiments Maximum temperatures at limit higher for 1-step model than experiments - typical result for 1-step model without chain branching steps Low Re: T max < T ad due to heat loss - even with heat recirculation Higher Re: heat loss less important, T max > T ad due to heat recirculation T max at extinction nearly same with or without turbulence even though limit mixtures (thus T ad ) are different At high Re, extinction is caused by insufficient residence time compared to reaction time - determined by flow velocity (Re) Reaction time far more sensitive to temperature than mixture Re determines T required to avoid extinction, regardless of transport environment required to obtain this temperature
27 Temperatures too high to conduct experiments above this Re! AME Spring Lecture 5 Model results - extinction limits
28 AME Spring Lecture 5 Model results - heat loss & radiation Radiation: effect similar to heat loss Causes heat to be conducted along the walls and subsequently lost to ambient Less important at smaller scales »Conduction ~ k( T/ x) »Radiation ~ (T 4 -T 4 ) »Radiation/Conduction ~ x … but unless you include radiation, you get the wrong answer when you calibrate a macroscale model then apply it to microscales! High Re: convection dominates heat transfer, finite residence time dominates extinction, all models yield almost same predictions
29 AME Spring Lecture 5 Model results - out of center limit Model shows that when fuel mole % increases, reaction zone moves out of center - consistent with experiments Semi-quantitative agreement between simulations & experiments - NO ADJUSTABLE PARAMETERS Again need to include turbulence at high Re
30 AME Spring Lecture 5 Model results - wall conductivity Heat recirculation requires spanwise conduction across wall from products to reactants … but conduction to wall also causes streamwise heat conduction - removes thermal energy from reaction zone which can be lost to ambient, narrows extinction limits (Ronney, 2003; Chen & Buckmaster, 2004) BUT if wall k = 0, no heat recirculation THERE MUST BE AN OPTIMUM WALL THERMAL CONDUCTIVTY Computational predictions High Re: convection >> conduction, wall k doesn’t matter unless it’s too small Lower Re: convection ≈ conduction, heat loss dominant; optimal k exists, but is less than air! Optimal k roughly where thermal resistance across wall ≈ thermal resistance air wall
31 AME Spring Lecture 5 Model results – wall conductivity
32 AME Spring Lecture 5 Model results - 3D effects Q: Does 2D model properly account for heat loss in 3rd dimension? A: (Chen & Ronney, 2011) Generally yes, but new effects arise - Dean vortices in flow in curved channels - additional heat transport - heat recirculation (thus extinction limits) similar with or without turbulence (RSM = Reynolds Stress model) included, whereas 2D model (no Dean vortices possible) shows very different results! Equivalence ratio at ext. limit
33 AME Spring Lecture 5 Model results - 3D effects No turbulence With turbulence
34 AME Spring Lecture 5 Model results - chemistry effects Q: One-step model: pre-exponential term (Z) adjusted to match experiments – can Swiss- roll combustors be modeled without adjustable parameters and/or complex chemistry? A: Yes – 4-step model (Hautmann et al., 1981) designed to model flow reactor experiments (not flames) works well with no adjustable parameters Equivalence ratio at ext. limit Reaction rate map: Re = 55 Reaction rate map: Re = step 4-step 1-step 1-step
35 AME Spring Lecture 5 Scale effects - revisited Simplified analysis (Chen and Ronney, 2013) Adiabatic energy balance across heat exchanger: equate heat transfer Q T to enthalpy increase of reactants due to Q T yields excess enthalpy (E) U T = overall heat transfer coefficient, A T = exchanger area N = number of transfer units from heat exchanger literature Non-adiabatic analysis using “mixing cup” (average) temperatures
36 AME Spring Lecture 5 Scale effects - revisited Heat transfer Laminar flow: U T ~ h ~ (k/d)Nu ~ (k/d)Re 0 h = heat transfer coefficient, Nu = Nusselt number N ~ U T A T / C P ~ (k/d)d 2 /( Ud 2 )C P ~ Re -1 ~ 1/d Turbulent flow: U T ~ (k/d)Nu ~ (k/d)Re 0.8, N ~ Re -0.2 Either way, Re (which is known a priori) is uniquely related to N, so can use Re as a scaling parameter instead place of N (which depends on h and isn’t known a priori) Heat loss U L generally independent of scale (for buoyant convection or radiation), A L ~ A T, thus for laminar flow with U T ~ 1/d, ~ d Thus, at low Re, for the same Re performance is poorer for large scale combustors
37 AME Spring Lecture 5 Scale effects - revisited Chemical reaction Reaction_rate/volume ~ Y f,∞ Z gas exp(–E gas /RT) ~ 1/(Reaction time) Residence time ~ V/(mdot/ ) ~ V/(( UA)/ ) ~ (V/A)/U (V = volume, U = velocity) V/A ~ d 3 /d 2 = d 1 Residence time ~ d/U Residence time / reaction time ~ Y f,∞ Z gas d/U exp(–E gas /RT)] ~ Da/(exp(–E gas /RT)])Re d -1 ; Da = Y f,∞ Z gas d 2 / Blowoff at high u occurs more readily for small d (small residence time / chemical time); at same Re d, need Z ~ 1/d 2 to maintain same extinction limit Radiation Convective transfer per unit area between walls i and j ~ U T (T i – T j ) Radiative heat transfer ~ [ /(2- )] (T i 4 – T j 4 ) Radiation / convection Surface radiation effects more important at larger scale; as previously discussed, hurts performance in a manner similar to streamwise wall heat conduction
38 AME Spring Lecture 5 Scale effects - revisited Simulations in 3D, 3.5 turn Swiss roll, without and with property values adjusted to obtain constant , Da and R Without adjustments, at small Re heat loss effects result in worse performance for large combustor whereas at large Re, residence time (Da effects) results in worse performance for small combustor; with adjustments, all scales similar Property HalfFull Double h L (W/m 2 K) ε L (external wall) ε L (insulation) Z (m-sec-kmole units) 1.44 x x x 10 9 ε i (internal wall) Without property adjustment With property adjustment
39 AME Spring Lecture 5 Linear exchanger vs. spiral Swiss roll Create pseudo-3-turn spiral exchanger from linear exchanger cut into 3 pieces, again use mixing-cup temperatures
40 AME Spring Lecture 5 Linear exchanger vs. spiral Swiss roll Adiabatic linear exchanger performance much better than spiral exchanger at large N (low Re) With increasing heat loss ( ), linear exchanger performance deteriorates substantially compared to spiral exchanger (homework problem!) … but this is all just heat transfer, what about with chemical reaction? Linear Simulated spiral
41 AME Spring Lecture 5 Linear exchanger vs. spiral Swiss roll Consistent with detailed calculations (Chen & Ronney, 2013) Adiabatic »Linear better (leaner extinction limit) at low Re (large N) »Same performance at high Re (small N) (Swiss roll has 2x larger A T than linear device, so 2x lower equivalence ratio at limit) Non-adiabatic »Swiss roll MUCH better at low Re (need to reduce for linear device heat loss coefficients by 4x just to get plots on the same scale!)
42 AME Spring Lecture 5 Model results - number of turns Fair comparison – same overall dimension and wall thickness (fabrication limitation) Ronney, 2015: More turns means larger N but more material, thus more thermal conduction (and heat loss) in 3 rd dimension – optimum exists, but relatively flat; optimal n larger at higher Re (lower N, more “starved” for additional heat recirculation)
43 AME Spring Lecture 5 References Ahn, J., Eastwood, C., Sitzki, L., Ronney, P. D. (2005). “Gas-phase and catalytic combustion in heat- recirculating burners,” Proceedings of the Combustion Institute, Vol. 30, pp Chen, C.-H., Ronney, P. D. (2013), “Scale and geometry effects on heat-recirculating combustors,” Combustion Theory and Modelling, Vol. 17, pp (2013) Chen, C.-H., Ronney, P. D. (2011) “Three-dimensional Effects in Counterflow Heat-Recirculating Combustors,” Proceedings of the Combustion Institute, Vol. 33, pp Hautman, D. J., Dryer, F. L., Schug, K. P., Glassman, I. (1981). “A Multiple-step Overall Kinetic Mechanism for the Oxidation of Hydrocarbons,” Combustion Science and Technology Vol. 25, pp Kuo, C.-H., Ronney, P. D. (2007). Numerical Modeling of Heat Recirculating Combustors, Proceedings of the Combustion Institute, Vol. 31, pp Lloyd, S.A., Weinberg, F.J., Nature 251:47-49 (1974). Lloyd, S.A., Weinberg, F.J., Nature 257: (1975). Maruta, K., Muso, K., Takeda, K., Niioka, T., Proc. Combust. Inst. 28: (2000). Ronney, P. D. (2015). “Heat-Recirculating Combustors,” Chapter 8 in Microscale Combustion and Power Generation (Y. Ju, C. Cadou and K. Maruta, Eds.), Momentum Press LLC, New York. Sanford, L. L., Huang, S. Y. J., Lin, C. S., Lee, J. M., Ahn, J. M., Ronney, P. D. (2008). “Plastic mesoscale combustors/heat exchangers,” Proceedings of the ASME International Mechanical Engineering Congress and Exposition, Nov. 11 – 15, 2007, Seattle, WA, pp. 141 – 145. Targett, M., Retallick, W., Churchill, S. (1992). “Solutions in closed form for a double-spiral heat exchanger,” Industrial and Engineering Chemical Research 31,