Enabling High Efficiency Combustion through an Improved Understanding of Cyclic Dispersion Robert Wagner Fuels, Engines, and Emissions Research Center Energy and Transportation Science Division 2011 ERC Symposium Future Engines and Their Fuels June 8, 2011 University of Wisconsin
Outline Stability challenges Control opportunities Operation near the edge of stability Transitioning combustion modes Avoiding abnormal combustion events Forward directions 2 Managed by UT-Battelle
Outline Stability challenges Control opportunities Operation near the edge of stability Transitioning combustion modes Avoiding abnormal combustion events Forward directions 3 Managed by UT-Battelle
Motivation and Challenges of High Efficiency Clean Combustion Motivation EFFICIENCY More rapid energy release Reduction in heat transfer losses Reduction in NOx and soot emissions less aftertreatment to meet emissions regulations Local Equivalenc ce Ratio 6 5 4 3 2 1 0 600 1400 2200 3000 Temperature [K] Low Temperature Combustion processes avoid the high soot and NOx regions Challenges Precise control of boundary conditions Turbo-machinery, variable valve actuation, heat exchangers Higher in-cylinder peak pressures Materials Improved fuel injection technology Injector architecture, faster response Increase in HC, CO emissions Stability Cyclic dispersion, adaptive controls, advanced sensors, etc. 4 Managed by UT-Battelle
Stability and control are potential roadblocks to the most efficient implementation of many advanced combustion concepts Practical implementations operate well away from the edge of stability to avoid unintended excursions. Cyclic dispersion driven by stochastic (in-cylinder variations) and deterministic (cycle-to-cycle coupling) processes. Very nonlinear relationship for conditions consistent with many LTC concepts. Deterministic mechanisms act as nonlinear amplifier to stochastic variations. Further complicated by cylinder imbalances. Cyclic dispersion may amplify cylinder-to-cylinder imbalances. Improved control will require an improved understanding of instability mechanism. Stable Combustion Safe Operation Acceptable Cyclic Dispersion Increasing Charge Dilution Transition Region Unstable No Combustion Complete Misfire Unintended excursions to the unstable region may result in misfires and very strong rebound events which could damage the engine and/or catalyst system. Edge of Stability 5 Managed by UT-Battelle
Advances in sensor technology and onboard computer power are expanding the possibilities for high speed predictive control Measurement Analysis Control In-cylinder pressure Ionization Acoustic Other Pattern recognition Prediction Modeling Avoid certain states Short-and long-time scale feedback perturbations Pro-active 6 Managed by UT-Battelle
Adaptive controls must compensate for prior cycle memory, stochastic influences, and speed/load demands which impact the current combustion event Fuel / Air IN Exhaust OUT EGR 1. Intake 2. Compression 3. Power Combustion depends on temperature and composition. 4. Exhaust Residual gas influences beginning of next cycle. Cycle-to-Cycle Coupling 7 Managed by UT-Battelle
EXAMPLE Even simple systems can have very complicated dynamic behavior 1 Logistic Equation 1 0.8 k = 3.2 x x(i+1) 0.6 0.4 0.2 0 0 x(i) 1 x(i + 1) = f [x(i)] = k x(i) [1 x(i)] 0 2.8 3.0 3.2 3.4 3.6 3.8 k Development of higher order periods and deterministic chaos 8 Managed by UT-Battelle 4.0
Stochastic noise complicates the system response, but underlying map is still recoverable (and necessary for control) Logistic Equation with random perturbations on k 1 1 x(i+1) x(i+1) 0 0 1 x(i) 0 0 1 x(i) Underlying map reconstructed, even with high levels of parametric noise, assuming form of n th order polynomial 9 Managed by UT-Battelle
Outline Stability challenges Control opportunities Operation near the edge of stability Transitioning combustion modes Avoiding abnormal combustion events Forward directions 10 Managed by UT-Battelle
Spark-ignition combustion becomes unstable with lean operation (or high levels of dilution) φ= 1.0 More Lean (higher dilution) Heat Release (i+1) 0 900 Heat Release (i) φ= 0.72 Heat Release Heat Release (i+1) 0 900 Heat Release (i) Equivalence Ratio Experimental Data 11 Managed by UT-Battelle Wagner et al. International Journal of Engine Research, 1, No. 4, pp. 301-320, 2001.
Example of map reconstruction and identification of fixed points φ= 0.68 φ= 0.70 φ= 0.72 φ= 0.74 Heat Release (i+1) Heat Release (i) Heat Release (i+1) Heat Release (i) Map reconstructed assuming form as 2 hr(i + 1) = b + b hr(i) + b hr(i) +... + b 0 1 2 n n hr(i) 12 Managed by UT-Battelle Wagner et al. SAE 2001-01-3559
Same approach used to reconstruct the bifurcation diagram Experimental Data Reconstructed Map Understanding the dynamics may allow operation closer to the edge of stability. 13 Managed by UT-Battelle Wagner et al. SAE 2001-01-3559.
Simple model captures nature of lean combustion instabilities and assists with physical interpretation of data Reconstructed Map Model Low-order model used as basis for predictive control algorithm development. 14 Managed by UT-Battelle Model results from Daw et al., Physical Review E, 57:3, 2811-2819
Cylinder-to-cylinder differences become more of an issue for operation near a stability limit Small differences may result in different dynamic behavior for each cylinder. An abnormal combustion event in one cylinder may influence other cylinders. Ford V8, φ= 0.66 Heat Release (i+1), J Heat Release (i), J 15 Managed by UT-Battelle
Low-order maps reveal cylinder-to-cylinder differences in dynamic state GM Quad 4, φ= 0.6 700 Cylinder 1 Cylinder 2 Cylinder 3 Cylinder 4 HR (i+1) 0 0 700 HR (i) 700 HR (i+1) 0 0 700 HR (i) Air Flow 16 Managed by UT-Battelle 1 2 3 4
Cylinders communicate with each other and tend to synchronize under unstable conditions Symbol sequence analysis used to resolve dynamical relationships between cylinders. Original time series transformed into sequence of discretized symbols. Transformation aids in detecting and characterizing nonrandom patterns, even in the presence of high levels of noise. Data partitioned into discrete bins (binary example shown) 1 0 0 1 1 0 1 0 0 1 0 1 0 1 1 1 0 0 1 0 1 Binary Symbol Stream 17 Managed by UT-Battelle
Symbol analysis used to construct synchogramsto observe evolution of joint cylinder bifurcations and correlation in time GM Quad 4, φ= 0.59 60 60 50 50 Seq quence Index 40 30 20 Seq quence Index 40 30 20 10 10 0 0 100 200 300 Cycle 0 0 100 200 300 Cycle Uncorrelated (F1, F3) Anti-Correlated (F2, F3) 18 Managed by UT-Battelle
General observations on synchronization Synchronization occurs frequently with bifurcation of 2 or more cylinders. One cylinder can act as a driver for others. Synchronization occurs episodically (correlations persist for long times and suddenly shift). Probably associated with pressure waves in intake/exhaust manifolds, fueling interactions, common noise 19 Managed by UT-Battelle
Example application of adaptive control by ORNL and Ford Motor Company on an 8-cylinder SI engine Uncontrolled Controlled 1 1 Combustion Index, cycle i+1 0.5 0-0.5 Combustion Index, cycle i+1 0.5 0-0.5-1 -1-0.5 0 0.5 1 Combustion Index, cycle i -1-1 -0.5 0 0.5 1 Combustion Index, cycle i Demonstrated significant improvements in engine stability for idle and low-load conditions (U.S. Patent 5,921,221). 20 Managed by UT-Battelle
Outline Stability challenges Control opportunities Operation near the edge of stability Transitioning combustion modes Avoiding abnormal combustion events Forward directions 21 Managed by UT-Battelle
Spark assist is potential mechanism to control and transition HCCI for more widespread use Motivation Use of Homogeneous Charge Compression Ignition (HCCI) combustion has potential for significant reduction of NOx emissions and increased fuel efficiency. Unstable transition between SI and HCCI operation is challenge for HCCI implementation. Objective Improve understanding of instability mechanism during transition from SI to HCCI. Make use of engine-based, dynamic combustion measurements to quantify effective global kinetic rates. Develop simplified cyclic combustion models for rapid simulation, diagnostics, and controls. HCCI Description Pre-mixed, pre-heated fuel-air charge Ignition very dependent on temperature and species Uniform, spontaneous combustion without flame front 22 Managed by UT-Battelle Figure adapted from GM images by K. Dean Edwards, ORNL.
Distinct combustion modes observed during transition from SI to HCCI Nominal operating conditions 1600 rpm, 3.4 bar IMEP with no feedback control 13% Internal EGR 58% 15 2500 NOx (ppm) Conventional CO OV IMEP, % 12 9 6 COV IMEP (%) 2000 1500 1000 NOx, ppm Discontinuity between transition range and spark assisted HCCI HCCI 3 500 No spark required 0 0 360 340 320 300 280 260 240 220 Exhaust Valve Closing, ATDC Transition Requires spark and exhibits multiple combustion modes including hybrid modes Spark Assisted HCCI Requires spark to operate 23 Managed by UT-Battelle Wagner et al. SAE 2006-01-0418 / Bunting et al., SAE 2006-01-0872
Bifurcation diagram further illustrates unstable transition from conventional SI to HCCI combustion Combustion evolves between two distinct states representing SI and HCCI combustion. Ignition and propagation processes for both SI and HCCI are highly nonlinear. Transition region requires spark ignition. Is there any reason we would want to operate within this region? Better understanding of dynamics necessary to navigate transition region and speed/load transients. 24 Managed by UT-Battelle
Heat release rates reveal that desirable and undesirable sequences involve different mixtures of SI-like and HCCI-like combustion Spark timing: 25 BTDC SI ref HCCI ref 2 1 4 3 Different forms of combustion visited during unstable excursions. Possibilities include flame propagation, HCCI, and hybrid modes. 1 346 kj/m3 2 448 kj/m3 3 896 kj/m3 4 317 kj/m3
Operation at an intermediate dilution level could allow for HCCI-like benefits with reasonable pressure rise rates Spark timing: 25 BTDC SI ref HCCI ref 2 4 5 3 1 6 Multiple reaction modes appear to be occurring during a single combustion event. Tension between competing modes leads to characteristic patterns of instability. 1 321 kj/m3 2 390 kj/m3 3 887 kj/m3 4 786 kj/m3 5 774 kj/m3 6 752 kj/m3
Heat release profiles associated with fixed points are much different than observed for pure SI or HCCI Spark timing: 25 BTDC SI ref HCCI ref 3 1 7 2 5 6 4 Unstable period-1 fixed point is possible control zone. Trajectory in/out of stable zone appears predictable. May be possible to rate shape heat release through trajectory control. 1 772 kj/m3 2 771 kj/m3 3 745 kj/m3 4 754 kj/m3 5 702 kj/m3 6 708 kj/m3
Opportunity for control is to exploit naturally occurring hybrid states which exhibit an optimal balance of SI and HCCI 2 2 4 5 3 1 3 2 5 7 6 4 1 4 3 1 6 Unstable 3-state pattern due to excursions in SI HCCI balance Optimal balance of SI & HCCI occurs naturally but is not stable Frequent recurrence of optimal state suggests potential control target Patterns are complex but not random and are short-term predictable. 28 Managed by UT-Battelle
Combustion variations under intermediate dilution conditions exhibit recurring multi-cycle sequences Some burn variations are almost periodic and large in amplitude (e.g., a 3-cycle sequence) Combustion often stabilizes near one state for several cycles but eventually diverges When combustion finally diverges from this 'nominal' state, it does so in specific ways 29 Managed by UT-Battelle
Simple physical model yields integrated HR patterns similar to those observed in experiments Tracking cycle-by-cycle events reveals switching between two distinct HCCI-like combustion modes. Hypothesize switch mechanism depends on shifts in residual composition. Random noise added to EGR fraction to simulate stochastic in-cylinder processes. Experiment Simple Physical Model Such models are expected to be useful for short-term prediction and correlation of instability patterns with fuel properties and chemistry. 30 Managed by UT-Battelle Daw et al., ICEF2007-1685, 2007.
Outline Stability challenges Control opportunities Operation near the edge of stability Transitioning combustion modes Avoiding abnormal combustion events Forward directions 31 Managed by UT-Battelle
Abnormal combustion events observed for low-speed high BMEP operation Often referred to in the literature as Low Speed Pre-Ignition (LSPI) or Superknock SwRI Example LSPI* Point 1250 rpm @ 12.5 bar Stoichiometric operation with no dilution. 30,000 engines cycles necessary to observe multiple occurrences. BMEP [bar] 20 18 16 14 12 10 BSFC [g/kw-hr] 230 300 270 240 8 Data courtesy of Manfred Amannand Terry Alger from Southwest Research Institute. 6 4 250 270 300 350 2 1000 1500 2000 2500 3000 3500 4000 4500 5000 Engine Speed [rpm] 32 Managed by UT-Battelle
Experimental data illustrating the onset of abnormal combustion Each pre-ignition event consists of 1-20 consecutive engine cycles before returning to normal operation. Engine damage may be caused within few engine cycles depending on the severity of the events. Cycle: 53033 53034 53035 53036 53037 53038 53039 53040 53041 53042 53043 53044 Pre 53032 LSPI normal normal cycle cycles cycles 140 150 120 125 ) Cylinder Pressure (bar) 100 80 60 40 Peak Pressure (bar) 100 75 50 20 25 0 270 300 330 360 390 420 450 CAD 0 53030 53040 53050 Cycle Number 33 Managed by UT-Battelle
Advanced time-series analysis techniques used to find patterns with purpose to develop methods for LSPI avoidance Example methods include data symbolization (coarse-grained inter-cyclic pattern identification) and phase-space reconstruction techniques (intracyclic dynamic evolution). Phase-space reconstruction shows pronounced prior-cycle deviations, and the evolution of the LSPI event is highlighted early. LSPI event Phase-space reconstruction Pre-LSPI 34 Managed by UT-Battelle Analysis by Charles Finney, ORNL.
Outline Stability challenges Control opportunities Operation near the edge of stability Transitioning combustion modes Avoiding abnormal combustion events Forward directions 35 Managed by UT-Battelle
On-going and Future Research Toolkit development for assessing, predicting, and controlling unstable behavior Algorithms based on well-understood principles from nonlinear dynamics and information theory. Includes addressing amplification of cylinder-to-cylinder dynamics due to high cyclic dispersion. Abnormal combustion avoidance Informal collaboration with SwRI for prediction and avoidance of undesirable excursions. Reactivity Controlled Compression Ignition (RCCI) combustion Operational space bounded by regions of instability. Instabilities strongly affected by dilution level, intake temperature, mixing, etc. Cylinder-to-cylinder imbalances are exacerbated by cyclic dispersion. Predictive control under development and based on better understanding of instability mechanisms and cycle-to-cycle interactions. 36 Managed by UT-Battelle
Final Thoughts A dynamic perspective provides new insight into combustion instabilities and control opportunities. Impact of stochastic processes on combustion stability a high priority topic in upcoming DOE predictive simulation initiative. Deterministic processesalso of extreme importance for operation near edge of stability. Potential benefits: Operation closer to the edge of stability. More efficient mode transitions. Operation within inherently unstable regions for efficiency/emissions benefits. Prediction and avoidance of abnormal destructive combustion events (e.g., Superknock). Control comments: Reduce stochastic variations as much as possible. Many useful tools from nonlinear dynamics and information theory. Low-order qualitative models able to predict the complex dynamics observed experimentally. Appropriate control target is stable dynamic manifold region as opposed to a point. 37 Managed by UT-Battelle
Robert Wagner Acting Director Fuels Engines and Emissions Research Center (865) 946-1239 wagnerrm@ornl.gov Acknowledgements Special thank you to the DOE for funding portions of this research, numerous ORNL staff for valuable contributions to this topic, and SwRI for sharing LSPI data. Questions? 38 Managed by UT-Battelle