Modeling and Optimization of Trajectory-based HCCI Combustion

018 CCEFP IEC Summit at the University of Minnesota Modeling and Optimization of Trajectory-based HCCI Combustion 018 CSSCI Spring Technical Meeting Chen Zhang Abhinav Tripathi Professor Zongxuan Sun Department of Mechanical Engineering University of Minnesota

Presentation Outline Motivation, Background and Objective Trajectory-based HCCI Combustion Control Chemical Kinetics Driven Model Control-oriented Model Optimization of Piston Trajectory Controlled Trajectory Rapid Compression and Expansion Machine (CT-RCEM) Conclusions Chen Zhang 018 CSSCI Spring Technical Meeting

Motivation Large amount of energy is consumed by transportation sector annually: In the US, transportation sector consumes ~30% of total energy There are 800 million passenger vehicles in the world now The above number is projected to reach 8 billion by 050 Environment and human health impact: CO effect on climate (greenhouse gases) NOx emission forms smog and acid rain PM causes serious respiratory disease Chen Zhang 018 CCEFP IEC Summit Meeting 3

Emission performance Background HCCI Combustion Spark Ignition Engine HCCI Engine Diesel Engine Fuel efficiency HCCI Combustion Thermal heat release ü Enable higher compression ratio than SI engines Process Reaction production ü Extremely Reaction fuel-lean rate condition In-cylinder Gas Chemical ü Low temperature Kinetics of fuel Pressurecombustion Dynamics Temperature ü Shorter combustion duration Species concentration Chen Zhang 018 CCEFP IEC Summit Meeting v Still short of controllability in ICE to adjust the HCCI combustion v Existing control methods, eg regulating exhaust gas recirculation (EGR), varying valve timing and stratifying charge, can only execute combustion control at specific time instants in each cycle 4

Background FPE in UMN Opposed Piston Opposed Cylinder (OPOC) Design Direct Fuel Injection Uniflow Scavenging q Variable compression ratio Inner Piston Pair On-off Valve Check Valves Advanced combustion strategy Multi-fuel operation LP Servo Valve Hydraulic Chambers Intake Ports Exhaust Ports Exhaust Ports On-off Valve Outer Piston Pair Chen Zhang HP Intake Ports q Reduced frictional losses q Higher power density q Fast response time Virtual crankshaftbalanced mechanism q Internally Enables the FPE pistons to track prescribed q Modularity trajectories reference precisely 018 CCEFP IEC Summit Meeting 5

Trajectory-based Combustion Control Chemical Kinetics of Fuel Check Valves Inner Piston Pair On-off Valve Hydraulic Chambers LP Servo Valve Thermal heat release Reaction product Reaction rate Pressure Temperature Species concentration In-cylinder Gas Dynamics Volume Variable Piston Trajectories Intake Ports Exhaust Ports Outer Piston Pair Exhaust Ports On-off Valve HP Intake Ports Virtual Crankshaft Mechanism Chen Zhang 018 CCEFP IEC Summit Meeting 6

Trajectory-based Combustion Control Model Based Trajectory Optimization Optimal Trajectories Active Motion Control (Virtual Crankshaft) Control Signals Task 1: Model the in-cylinder processes with chemical kinetics and thermodynamic states along variable piston trajectories Piston displacement Task : Develop a control-oriented In-cylinder gas volume, model pressure, to temperature realize the and trajectorybased combustion control in major chemical spices concentration practice Overall system configuration of trajectory based combustion control Task 3: Based on the above models, design optimal piston q Inner Loop: piston motion control - virtual crankshaft trajectories by leveraging the dynamic between chemical kinetics and qgas Outer dynamics Loop: Trajectory-based combustion control Chen Zhang 018 CCEFP IEC Summit Meeting 7

Presentation Outline Motivation, Background and Objective Trajectory-based Combustion Control Chemical Kinetics Driven Model Control-Oriented Model Optimization of Piston Trajectory Controlled Trajectory Rapid Compression and Expansion Machine (CT-RCEM) Conclusions Chen Zhang 018 CSSCI Spring Technical Meeting 8

Chemical Kinetics Driven Model The entire model is separated into 3 parts Fuel Number of species Number of reactions Resource A W cos(pf t) X = + B W cos(pf t) + sin(pf t) Energy conservation: First law of thermal dynamics Convective heat loss Methane 53 35 GRI-30 Propane 50 44 UC Santiago Ethanol 57 383 LLNL a Mass conservation DME 79 683 LLNL Ammonia 3 98 Cal-tech Ideal gas law n-heptane 160 1540 LLNL NASA polynomial parametrization: C v i ( T), 3 4 = a + a T + a T + a T + a T 0 1 3 4 R h ( T), a a 1 a3 3 a i m = 4 4 a T T T T + 0 + + + + RT 3 4 5-1 a 5 Chen Zhang 018 CCEFP IEC Summit Meeting 9

Combustion Phasing Control CR = 03 CR = 46 CR = 346 CR = 48 Chen Zhang 018 CCEFP IEC Summit Meeting 10

Combustion Phasing Control Ω = 05 Ω = 10 Ω = 15 Increasing both the CR and the Ω can realize multiple fuels combustion in FPE Increasing CR and Ω can both advance the SOC timing A new control means is achieved for combustion phasing control piston trajectory Chen Zhang 018 CCEFP IEC Summit Meeting 11

Indicative Output Works Methane Propane Ethanol DME n-heptane Ammonia Both CR and Ω have direct effects on the amount of indicative work output The peak of output work surface is located at zone with larger CR and smaller Ω Chen Zhang 018 CCEFP IEC Summit Meeting 1

Heat loss amount Methane Propane Ethanol DME n-heptane Ammonia The peak of heat loss amount is located at zone with larger CR and larger Ω Chen Zhang 018 CCEFP IEC Summit Meeting 13

NOx Emission Asymmetric trajectories Temperature profiles NOx emission profiles NOx emission Ø The majority of NOx emission is produced after the major heat release Ø Quick expansion reduce the temperature and thus reduce the NOx emission Ø We can reduce NOx emissions and increase efficiency simultaneously Chen Zhang 018 CCEFP IEC Summit Meeting 14

Control-Oriented Model Comprehensive model Detailed reaction mechanism Higher computational burden Control application Existing controloriented model Over-simplifying the chemical kinetics HCCI combustion simulation New controloriented model Best balance between computational burden and prediction accuracy Chen Zhang 018 CCEFP IEC Summit Meeting 15

Modeling Approach Variable piston trajectories Geometric part Physicsbased part The first law of thermodynamics and heat loss process Chemical kinetics part (Phase Separation) Combustion process Chen Zhang 018 CCEFP IEC Summit Meeting 16

Phase Separation Method In order to reduce computational time and keep sufficient chemical kinetics information, the engine operation cycle is separated by several phases Phase separation within an engine cycle Phase 1: Pure compression Phase : Ignition phase R : + H 1 CH4 + 05O CO 9 05 15 15095 RR1 = 44 10 [ CH 4 ] [ O ] exp( - ) T Phase 3: Heat release phase R R 3 : CO + : H H + 05O O CO H O 16 60 10 0 5 w x = [ N ][ O ] NO 05 T Phase 4: Pure expansion 7 10065 RR = 75 10 [ CO] [ H O]exp( - ) T 9 05 17609 RR3 = 15 10 [ H ][ O ] exp( - ) T Sub-phase: NOx production + H - 69090 exp( ) T Chen Zhang 018 CCEFP IEC Summit Meeting 17

Computational Cost The proposed model is compared with the other two models: Simplified model Entire chemical kinetics is represented by a global reaction Assuming the heat release is instantaneous after the combustion Detailed model The chemical kinetics is represented by a detailed reaction mechanism includes 53 species and 35 reactions Utilized model Computational time [ms] Detailed model 070 Proposed model 98 Simplified model 6 q The detailed model needs 070 ms to simulate an engine cycle q The proposed model reduce the computational turnaround time by 95% q The simplified model only takes 6 ms Chen Zhang 018 CCEFP IEC Summit Meeting 18

Prediction Accuracy v Good agreement between the proposed model and the detailed one v The simplified model fails to represent the combustion precisely due to its oversimplifying of the chemical kinetics o later start of combustion o higher peak temperature Comparison of the accuracy of the prediction Temperature and NOx production (inserted) v Simplified model cannot provide any information on NOx emission while the others predicts similar results of NOx production Chen Zhang 018 CCEFP IEC Summit Meeting 19

Piston Optimization one Ω Identical BDC Identical BDC CR Different TDCs Ω Identical TDC No combustion Larger heat loss Ø At given CR, various piston trajectories as well as their output work can be presented as a function of Ω Ø Smaller Maximize Ω will : result J 1 = in Wan incomplete ouput = f ( W) combustion and larger Ω will increase the heat loss Ø higher CR, lower optimal Ω Chen Zhang 018 CCEFP IEC Summit Meeting 0

Piston Optimization two Ω q Asymmetric trajectories o Increase thermal efficiency o Reduce NOx emissions q Two Ωs represent compression and expansion trajectories respectively q Forming a two dimensional optimization problem at a fixed CR W C( NOx) Maxmize : J = w1 - w = f ( W1, W ) W C ( NOx) max v Cost function J takes both work output and NOx emissions into account Piston Trajectory Ω 1 Ω Work [J] NOx Emission [mol/m 3 ] Asymmetric 35 040 40191 3e-6 v W max and C max (NOx) are the simulated maximum values respectively, working as the normalized coefficients Symmetric 111 111 3988 407e-6 max v w 1 and w are weight coefficients Chen Zhang 018 CCEFP IEC Summit Meeting 1

Chen Zhang 018 CCEFP IEC Summit Meeting Piston Dynamic Optimization û ù ë é = ø ö ç ç ç ç ç ç ç ç ç ç è æ û ù ë é F - - - - - n N n CH n n n N n CH n n N CH N CH n n X X T P X X T P X X T P X X T P u u u u ] [ ] [ ] [ ] [ ] [ ] [ ] [ ] [ 4 1 1 4 1 1 4 1 1 4 1 1 1 1 Due to the repetitive nature of the FPE operation: Ø the time differential equations in the model can be numerically solved in cycle base Ø It forms a mapping converting the piston trajectory (u) in each cycle to other states of the model The original dynamic optimization problem is transformed into a static nonlinear programming problem and solved by SQP algorithm: 6) 3 ] max(0,[ ) ( ) ), ( ( 101 100 1 1 - - + å - = - F = + e X r dt u u P u u f x NO i i i i 101 u ÎR 0 ) ( = u h 0 ) ( u g U L u u u Minimize: over Subject to Work output NOx emission r works as a penalty if the final NOx emission > 3e-6 mol/m 3 h(u) limits the start, middle and end points of the trajectory g(u) ensures the piston velocity < 8m/s u L and u U are the TDC point and the BDC point respectively

Piston Dynamic Optimization Ø New optimal trajectory has even shorter residential time around the Trajectory TDC and more aggressive Work expansion NOx Emission afterwards output [J] [mol/m 3 ] Asymmetric 40191 45e-6 Ø With the assistance of virtual crankshaft, Direct Optimization the piston can stay 41670 at the TDC 31e-6 location for a desired duration to realize the ideal constant volume combustion Chen Zhang 018 CCEFP IEC Summit Meeting 3

Presentation Outline Motivation, Background and Objective Trajectory-based Combustion Control Chemical Kinetics Driven Model Control-oriented Model Optimization of Piston Trajectory Controlled Trajectory Rapid Compression and Expansion Machine (CT-RCEM) Conclusions Chen Zhang 018 CSSCI Spring Technical Meeting 4

Controlled Trajectory RCEM System Architecture and Specifications Maximum Combustion Pressure 50 bar Minimum Compression Time 0 ms Maximum Compression Ratio 5 Combustion Chamber Bore 508 mm Maximum piston travel 19 mm TDC clearance 8 mm Hydraulic Working Pressure 350 bar Hydraulic Piston Bore 40 mm Mass of Piston Assembly 17 kg Hydraulic actuator unit: high pressure accumulator, servo-valve Combustion chamber unit: combustion cylinder with a creviced piston Fueling and exhaust purging system: different set of check valves Control module: centralized data logging and motion control unit Diagnostics system: GCMS and PLIF system Chen Zhang 018 CCEFP IEC Summit Meeting 5

Controlled Trajectory RCEM Characterization of CT-RCEM Tracking accuracy is the key to repeatability Four repetitions for CR: 167, compression time 0ms Repeatability Analysis for CR167 Stroke: 131 mm Compression time: 0ms Peak velocity: 15 m/s Average velocity: 7 m/s Peak tracking error: 06 mm Chen Zhang 018 CCEFP IEC Summit Meeting 6

Controlled Trajectory RCEM Preliminary Case Study: trajectory effect on the auto-ignition of DME Identical air fuel mixtures were compressed (DME:O:N = 1:4:40) The trajectories are with the same CR = 167, but different compression time (0ms and 30ms) The first-stage ignition delay is 09ms in 0ms case, while it is 16ms in 30ms case The CT-RCEM is a perfect facility to validate the trajectory-based combustion control: Ø Precise and fast piston motion control Ø Comprehensive information on fuel properties and the related emissions Ø Accurate measurement on pressure and species concentration via the optical diagnostics system Chen Zhang 018 CCEFP IEC Summit Meeting 7

Presentation Outline Motivation, Background and Objective Trajectory-based Combustion Control Chemical Kinetics Driven Model Control-oriented Model Optimization of Piston Trajectory Controlled Trajectory Rapid Compression and Expansion Machine (CT-RCEM) Conclusions Chen Zhang 018 CSSCI Spring Technical Meeting 8

Conclusion Ø Proposed the trajectory-based combustion control to achieve real-time control on the combustion in the FPE Ø Developed a dynamic model to systematically investigate the effectiveness of the proposed combustion control Ø Realized the enhancement of the thermal efficiency and the reduction of NOx emission simultaneously as well as the combustion phasing control for multiple fuels Ø Designed, manufactured and tested a unique CT-RCEM to enable the experimental validation of the proposed combustion control method Ø Provided a new platform to realize co-optimization on both fuel production and engine performance Chen Zhang 018 CCEFP IEC Summit Meeting 9