Air Handling Considerations for Dynamic Skip Fire

Applied to a 1.8L Turbo 4-Cylinder Engine

North American GT Conference

November 6th, 2017

R. Wang, B. Wolk

Tula Technology

What is Dynamic Skip Fire?

• Firing decisions made on a cylinder-by-cylinder basis: “Dynamic Downsizing”

• Determination whether a cylinder’s torque is required, immediately prior to firing

• Firing Density (FD) chosen to minimize fuel consumption, subject to certain constraints

• Production level NVH, drivability, OBD, and emissions

Torque Demand

Firing Pulsetrain

Specifications

Engine 1.8L TGDI I4 (EA888)

Test Weight Class 3500 lb.

Transmission 6AT

Tailpipe Emissions SULEV

Evaporative Emissions Zero Evap

EMSDelphi MT92.1 +

Tula algorithms

Deactivation Hardware Delphi DRFF + OCV

Projected DSF CO2

Improvement 7 – 9 %

3

Tula-Delphi DSF Jetta Demonstrator

GOAL: Prove DSF gains on an I4 GTDI vehicle

DSF cylinder deactivation mechanization and trapping strategy

4

OCVDRFF

Oil Control Manifold

DRFF deactivation

Fire Fire

LPES

Skip

Int.

Exh.

Low pressure exhaust spring (LPES) strategy minimizes

thermodynamic losses when skipping

Enabling DSF control within GT-POWER engine model

5

Int. Man

Throttle

Exh. Man

Mufflers

Turbine

Catalysts

CAC

Airbox

Comp

DSF Block

Cylinders & Cranktrain

DSF Control Block

• Firing pattern algorithm

• Cylinder-specific valve lift and injection

deactivation

Miscellaneous

• Custom variable calculation window for

RLTs – variation from 720CAD window

• Custom fuel injection control – delay

with default injector inputs

• Cylinder blowby model – resolve

impact of LPES

• Transient thermal solver – steady

solver causes jumps between fired and

skipped cycles

Greater MAP variation observed with DSF than I4

6

FF = 3/4 Cyl 1 Cyl 3 Cyl 4 Cyl 2

Cycle N Fire1 Fire2 Fire3 Skip

Fire1

Fire2

Fire3

Skip

Image: eSelf-Study Program 920243, Audi of America, LLC (2013).

I-4

DS

F

Exp.

Sim.

Exp.

Sim.

• Reduction in MAP = reduction in trapped air mass

• Potential cylinder lambda imbalance if equally fueled

Cylinder refire incurs enhanced gas exchange

7

FF = 2/3 Cyl 1 Cyl 3 Cyl 4 Cyl 2

Cycle N Fire1 Fire2 Skip Fire1

Cycle N+1 Fire2 Skip Fire1 Fire2

Cycle N+2 Skip Fire1 Fire2 Skip

• Each Fire1 also a cylinder refire (previous cycle

cylinder skipped)

• Greater trapped mass for refire due to lack of valve

overlap

• During overlap, exhaust pressure drives backflow

into intake reducing gas exchange for Fire2 events

• Refire + MAP variation compounds challenge for

cylinder lambda control

2000 RPM, 3 bar BMEP, FF = 2/3

Cylinder pressures also show effect of air mass variation

8

• Sim able to dictate stoich fueling per

cylinder - more difficult on dyno

• Exp results show initial cal from Tula

fuel compensation algorithm

• Air mass variation leads to cylinder

pressure and IMEP variations

• Reasonable agreement between sim

and measured – observe

overprediction in refire air mass

Simulation provides insight into MAP variation impact

9

Engine operating points

Engine Speed 2000 rpm

Engine BMEP 0.5 – 4.5 bar

Firing Fraction 3/4

• Competition between cam advance and

MAP lead to peak RGF around 50kPa MAP

• RGF and MAC maintain a consistent

sequence between fires

• Normalized RGF and MAC spread from

mean relatively consistent

RGF = residual gas fraction

MAC = mass air charge

Simulation provides insight into refire + MAP fluctuation impact

10

Engine operating points

Engine Speed 2000 rpm

Engine BMEP 0.5 – 4.0 bar

Firing Fraction 2/3

• Again, peak RGF observed at 50kPa MAP

for Fire2

• Consistent decrease in RGF for Fire1 -

due to increasing int/cyl PR at IVO

• Normalized MAC spread from mean

consistently grows with increasing MAP –

due to Fire1

Incremental improvement when fueling each fire stoichiometric

11

Experiment

• Sim and measured depict similar benefit

compared to equal fueling

• Incremental fuel consumption

improvement up to 1% at 4bar

• Smaller improvement for FF=3/4

MAC spread increasingFF = 2/3

• Engine-out CO reduced to acceptable level

• NOx and THC emissions largely unchanged

FF = 2/3

Fu

el C

on

su

mp

tion

Imp

rovem

en

t over

I-4

[%

]

Conclusions

12

• GT-POWER engine model of 1.8L EA888 with DSF functionality developed

• Special considerations need to be taken with fueling and RLTs

• Custom cycle definition extending past 720deg would be helpful

• Simulation predicts MAP fluctuations well

• Refire breathing overpredicted – requires further investigation

• Model provides insight into RGF and MAC variations in DSF

• Appropriately fueling each fire leads to meaningful improvements in BSFC and

engine-out CO

• Simulation supported algorithm development and calibration on a production-type

ECU – more details to be presented at SAE WCX18

DSF Technology Roadmap

MGU torque mitigation/assist

Multi-level (i.e. Miller)

Lean combustionAutonomous