Ion Sensing, Control, and Actuation in Multicylinder HCCI Engines

1
Ion Sensing, Control, and Actuation in
Multi-cylinder HCCI Engines
Hunter Mack Professors J-Y Chen, R.W. Dibble,
and J.K. Hedrick University of California -
Berkeley
HCCI University Consortium Meeting Sandia
National Laboratory Feb 8-9, 2006
2
Road Map for UC Berkeley
Build on Past Success of Ion Sensing (i.e.
Acetylene, EGR)
Fast Thermal Management (FTM) control of
multi-cylinder engine using Ion Sensors
Increase Intake Pressure !
First Supercharge, Then Turbocharge..
How does Ion Signal Change with Applied Boost?
Can we numerically model these results?
3
Outline

• HCCI and Ion-Sensor Technology
• Chemistry / Numerical Model
• Fuel Composition Effect
• EGR Effect
• 2. HCCI Multi-cylinder Control
• Actuation Approach
• Control Strategies
• Super / Turbo Charging HCCI
• Development of Isooctane Skeletal Mechanism
• 5. Future / Conclusions

4
Ion Sensors for HCCI
5
Ions in flames and Ion-sensor technology
Chemi-ionization reaction is the principal source
of ions in hydrocarbon flames
This reaction is rapidly followed by the charge
exchange reaction
H3O is the dominant ion in both fuel lean and
slightly rich hydrocarbon flames
6
The ionization mechanism focuses on the formation
and consumption of H3O .
The ion mechanism have 34 reactions involving 9
ionic species (CHO,O,N,OH,H3O, NO, O2,
N2, electrons)
7
We modified a 4 cylinder 1.9 Liter diesel engine
to use in HCCI mode.
VW TDI Engine specifications
Intake Manifold
VW TDI 1.9 Liter in HCCI mode
8
The ion signal accurately detects the combustion
event
Experimental Results Propane at varied
Stoichiometric Ratios
9
Using Acetylene increases the ion signal in an
HCCI engine
Remember the initiation reaction for ionization
Acetylene produces large number of CH radicals.
Electrical Aspects of Flames by Weinberg, 1969.
Ion signal should be stronger for acetylene
compared to propane.
10
The addition of acetylene to the fuel mixture
increases the ion signal

Constant Equivalence Ratio Acetylene in Propane
mixtures
11
Our Ionization Model Explains the Influence of
Fuel Composition and EGR on Ion-Signal
Fuels that produce higher CH radicals produce
stronger ion signal
Ion-Signal reaches a maximum close to
stoichiometric fuel-air ratio when EGR replaces
the intake air
12
Analogous to Fuel Composition Results, Increased
CH Radical Concentration is Responsible for
Stronger Ion-Signal with EGR
Numerical Model Results
13
Increased CH Radical Concentration is Responsible
for Stronger Ion-Signal
2-Orders of Magnitude increase
Note Logarithmic Scale
14
HCCI Control Strategies
15
Summary of Accomplishments in Control of HCCI
Combustion Timing

• Initially demonstrated two methods (electrical
  trim heaters and exhaust throttles) for
  controlling HCCI combustion timing
• Built real-time HCCI engine controller hardware
• Discovered low cost alternative
  start-of-combustion (SOC) sensors a) ion sensor
  and b) microphone sensor, essential for HCCI
  engine control.

16
Control requires an actuation method to change
combustion timing
Version 1 Electrical Trim Heaters

• Fast heat heaters coupled with mixing control
  of hot and cold air to control intake
  temperatures.
• Independent control of cylinder temperature by
  using separate intake runners.
• Effective method for achieving control of
  ignition time, engine balancing.

17
Control requires an actuation method to change
combustion timing
Version 2 Exhaust Flap Valves

• Exhaust throttles are used to adjust the
  combustion timing
• Closing throttle increases in -cylinder
• residual gases for next cycle and advances
  combustion timing
• Emulates variable valve timing (VVT)
• Sub-optimal, but rapidly implemented

18
Control requires an actuation method to change
combustion timing
Version 3 Fast Thermal Management (FTM) using
Intake Flap Valves
19
Intake System was Redesigned for FTM
20
Successful Test of FTM system for control of
combustion timing
21
Four Steps for Any Controller Design

• Develop a plant model (physical based or
  experimentally derived) Our plant is our
  engine. These essentially are the
  differential/difference equations describing
  plant dynamics.

• Design a controller using the plant model

• Implement the controller on the actual plant
  after successful performance on the plant model

• Test the controller on actual plant

22
Comprehensive Model Was Created in Simulink
intake manifold 9 differential equations
4 cylinders 20 differential equations
rotational inertia we 1 differential equation
4 exhaust manifolds 12 differential equations
23
The Model is Useful for Analysis butToo Large
for Control Synthesis

• Modeling for control design
• Want the simplest model that adequately captures
  the input-output behavior
• 42 states is much too comprehensive (and not
  necessary) for control
• Solution
• Generate a simpler model using experimental data

24
A Simple Linear HCCI Model is Needed

• Experimentally-derived model based on
  input-output data
• Model was developed using subspace system
  identification
• Canonical variate analysis weights (CVA) are used
• Multi-variable output-error state-space (MOESP)
  produces similar results

25
The Controller Performs Poorly on the Engine

• Experimental results
• Controller is unable to track 8 CAD combustion
  timing

• Modified Simulation Results
• A plant disturbance is added accounts for
  unknown cylinder wall and engine temperatures

controller on
26
Integral Action is Added to the LQG Controller
27
The controller significantly improves balancing
among the cylinders
Simulation
Experiment
28
Super/Turbocharged HCCI
29
The upper limit of the HCCI operating Regime can
be extended through Turbocharging
30
Ignition Temperature decreases at all RPM for
increased normalized P intake
Simulation CA50 1 ATDC
Source Killingsworth et al, A Simple HCCI
Engine Model for Control
31
An increase in Boost Pressure causes a decrease
in required Intake T
Simulation 1800 RPM CA50 1 ATDC
Source Killingsworth et al, A Simple HCCI
Engine Model for Control
32
First step is to supercharge the VW TDi(donated
by Eaton Automotive)
Source www.automotive.eaton.com/product/engine_co
ntrols/superchargers/M45.asp
33
Key Questions we want to answer.
How does an increase in Intake Pressure affect
Combustion Timing
Ion Signal
Cross-talk between cylinders
Power Output
34
Development of Isooctane Skeletal Mechanism
35
Skeletal Mechanisms can cut down on computational
costs

• Remove unimportant species and steps (targeted
  regimes)
• Collect data by using detailed mechanism (LLNL)
• ? Rate analysis
• ? Sensitivity analysis
• ? Computer Singular Perturbation (CSP)
• Sensitivity analysis CSP
• Need Jacobian matrix (?T/?ki)
  Eigen-values/vectors
• ? Expensive calculation (large of
  species and steps)
• e.g., 10hrs for single-zone
  Well-Mixed Reactor
• (one HCCI cycle)

Source Y.-H. Chen et al, Development of
Isooctane Skeletal Mechanisms for Fast and
Accurate Predictions of SOC and Emissions of HCCI
Engines based on LLNL Detailed Mechanism
36
Two skeletal mechanisms (258 and 291) are
developed

• Skeletal258 258 species 621 steps
• Aim for SOC (?0.6, P2atm ? get best result)
• Not accurate for emission predictions (HC CO)
• Skeletal291 291 species 875 steps
• Aim for both SOC and HC, CO emissions
• ?0.2 and 0.6 ? extra species and steps due to
  emissions
• Validation between detailed skeletal mechanisms
• Ignition-delay time (wild-range conditions are
  tested 36,750 runs)
• ?LLNL detailed mechanism 10
  mins/run
• ?Skeletal mechanisms 0.5-0.7
  mins/run
• Transient multi-cycle engine simulations with
  WMR
• CO, HC, HCOx emissions

Source Y.-H. Chen et al, Development of
Isooctane Skeletal Mechanisms for Fast and
Accurate Predictions of SOC and Emissions of HCCI
Engines based on LLNL Detailed Mechanism
37
Accurate Prediction of Ignition Delay for both
Skeletal Mechanisms
291
258
Source Y.-H. Chen et al, Development of
Isooctane Skeletal Mechanisms for Fast and
Accurate Predictions of SOC and Emissions of HCCI
Engines based on LLNL Detailed Mechanism
38
CO emissions validation done for both mechanisms
Skeletal291 predicts CO better than Skeletal258
Source Y.-H. Chen et al, Development of
Isooctane Skeletal Mechanisms for Fast and
Accurate Predictions of SOC and Emissions of HCCI
Engines based on LLNL Detailed Mechanism
39
Validation of HC and HCOx emissions
Skeletal291 predicts HC HCOx much better than
Skeletal258
Source Y.-H. Chen et al, Development of
Isooctane Skeletal Mechanisms for Fast and
Accurate Predictions of SOC and Emissions of HCCI
Engines based on LLNL Detailed Mechanism
40
Success in development of the skeletal isooctane
mechanisms
1. Both Skeletal mechanisms agree well for SOC
with LLNL detailed mechanism 2. Skeletal291 is
necessary for accurate predictions of HCCI
emissions 3. Speed-up factor of 15-20 by using
Skeletal mechanisms ? much less
computational cost than detailed mechanism
Source Y.-H. Chen et al, Development of
Isooctane Skeletal Mechanisms for Fast and
Accurate Predictions of SOC and Emissions of HCCI
Engines based on LLNL Detailed Mechanism
41
Conclusions

• Inexpensive ion sensors can be used to sense SOC
  in HCCI engines.
• Our Ion Mechanism numerical model results are in
  good agreement with the experimental findings.
• Engine control can be achieved through Fast
  Thermal Management and Ion Sensor technology
• A skeletal mechanism for isooctane provides
  accurate (and more computationally efficient)
  predictions for HCCI

Work-In-Progress All of these developments and
tools will be focused on greater understanding of
HCCI at elevated intake pressures
42
END END END
43
END END END
44
(No Transcript)
45
Extending Single Cylinder Results to
Multi-Cylinder is Surprising

• Closing an exhaust throttle on cylinder 1 causes
  the combustion timing to advance on 1, but also
  on 2

46
Cross-Talk is Related to the Firing Order

• 4 cylinder firing order 1 ? 3 ? 4 ? 2
• Closing 1 ex. throttle ? advance CA50 on 1 and 2
• Closing 3 ex. throttle ? advance CA50 on 3 and 1
• Closing 4 ex. throttle ? advance CA50 on 4 and 3
• Closing 2 ex. throttle ? advance CA50 on 2 and 4
• Conclusion
• Closing an exhaust throttle on a cylinder
  advances the combustion timing on that cylinder
  and (to a lesser extent) the preceding cylinder
  in the firing order

47
Ion sensors are Installed on the Engine and
Machining of Other Parts is in Progress
Flap Valve
48
Ion Sensors Provided by Collaboration with
Woodward
Modified spark plugs, 10 mm diameter threads
tapped into glow plug hole
49
Valve is sucked in, Piston Jams on upstroke,
rod emerges from side of Engine block. (main
damage to cylinder 1 on RHS)
50
Valve is sucked in, Piston Jams on upstroke,
rod emerges from side of Engine block. view
of valve ports on damaged cylinder Note hole into
water jacket
51
We are Not Only Ones with HCCI Damage Damage to
Diesel Piston in HCCI operation Possibly caused
by stoichiometric operation?
52
We Successfully Demonstrated Two Methods of
Individual Cylinder Combustion Timing Control
Timing Control via Exhaust Throttle EGR
Timing Control via Intake Air Electrical Heater
53
We Balanced the 4 Cylinders via Intake Air
Heating
54
Important NOX generation mechanisms.
Zeldovich Mechanism Reactions
Fenimore/Prompt Mechanism Reactions
N2O Mechanism Reactions
55
Although extremely low NOX 2 ppm, butN2O
mechanism is as important as thermal mechanism
for HCCI !
Fenimore Mechanism contribution is
insignificant. N2O mechanism, in the case of HCCI
type combustion, is as important as the Zeldovich
or thermal mechanism. N2O mechanism contributes
more than thermal NOX in some cases.
56
We have operated three different engines in HCCI
mode

• First experimental prototype
• CFR engine
• SAE 2000-01-0328

• Volkswagen TDI engine
• High-speed, 4-cylinder small displacement engine
• SAE 2001-01-1894 and 1895

• Caterpillar 3401 engine
• Representative of heavy-duty diesel engines
• SAE 2002-01-1758