Signal Processing, Automation and Control department – 2008, June 3rd©IF
P
Controlled CO2 | Diversified fuels | Fuel-efficient vehicles | Clean refining | Extended reserves
Prototyping and Deployment of Real-Time Signal Processing Algorithms for
Engine Control and Diagnosis
Fabrice Guillemin, Oliver Grondin, Emmanuel KuentzmannIFP
Signal Processing, Automation and Control department – 2008, June 3rd2
©IF
P
Outline
IntroductionRapid prototyping platform for real-time signal processing algorithmsAlgorithm implementation for combustion analysisDeployment on industrial DSP based targetConclusion
Signal Processing, Automation and Control department – 2008, June 3rd3
©IF
P
Outline
IntroductionRapid prototyping platform for real-time signal processing algorithmsAlgorithm implementation for combustion analysisDeployment on industrial DSP based targetConclusion
Signal Processing, Automation and Control department – 2008, June 3rd4
©IF
P
Introduction
Today issues :Increasing legal requirements (tailpipe emissions, fuel consumption) while keeping engine performancesDevelopment of innovative combustion concept: LTC, HCCI, CAIMore sensitive combustion processes according to
initial conditions (BGR, temperature, pressure) and fuel propertiesinjection system drift
No direct control of the start of ignition
A closed loop combustion control is needed to ensure fuel loop robustnessCombustion indicators (closed loop variables) can be computed from :
cylinder pressureinstantaneous engine speedion current
Need for higher sampling rate and suitable rapid prototyping platform
Signal Processing, Automation and Control department – 2008, June 3rd5
©IF
P
Real-time combustion analysis issue
Need of relevant combustion parameters to characterize
combustion phasing (efficiency, emissions)combustion noise
In-cylinder pressure analysis provides significant informationAcquisition and processing of in-cylinder pressure signals
cycle to cyclecylinder to cylinderengine synchronously
-100 -80 -60 -40 -20 0 20 40 60 80 1000
10
20
30
40
50
[°CA]
[Bar
]
Middle of Combustion
Start of Combustion
Energy of Combustion
in-cylinder pressure
Signal Processing, Automation and Control department – 2008, June 3rd6
©IF
P
Outline
IntroductionRapid prototyping platform for real-time signal processing algorithmsAlgorithm implementation for combustion analysisDeployment on industrial DSP based targetConclusion
Signal Processing, Automation and Control department – 2008, June 3rd7
©IF
P
Platform specification
IFP has developed a real-time platform for engine-synchronous algorithms prototyping. It fulfills the following specifications:
engine events synchronization (TDC or 6°CA)eight continuous or multiwindowed acquisition channelsacquisition at fixed frequency (400 kHz) or at fixed angular resolution (0.1°CA)cycle-to-cycle and cylinder-to-cylinder data availability for online signal processing and combustion analysis algorithmsdata recording for database acquisition and post-treatment purpose (over 1000 consecutive cycles)
Signal Processing, Automation and Control department – 2008, June 3rd8
©IF
P
Platform description
• Cylinder pressures• ...
ECU
Rapid prototyping platform :Industrial PC with real time kernel
• High frequency acquisitions• Engine TDC synchronous events• Algorithm rapid prototyping facility
Industrial PC based acquisition and rapid prototyping platform
Signal Processing, Automation and Control department – 2008, June 3rd9
©IF
P
Platform descriptionIndustrial PC based acquisition and rapid prototyping platform
• Cylinder pressures• ...
Signal Processing, Automation and Control department – 2008, June 3rd10
©IF
P
• CPU (xPC real time kernel) • RAM (online data storage) • HDD (data recording)• IFP Timer board (PCI) : FPGA (ALTERA Stratix)
• Initially designed for generation of ignition and injection signals driven by ACEBox control system
• IFP Daughter acquisition board • maximum sampling frequency : 400 KHz • minimum sampling period : 0.1 °CA• 16 bit resolution (+/-10V with tuneable gain)
Platform description : Hardware architecture
Signal Processing, Automation and Control department – 2008, June 3rd11
©IF
P
Platform description : acquisition functionalitiesAcquisitions frame-based, continuous or windowed in engine cycleAcquisitions in time or angle Acquired data frames updated engine synchronously (cylinder-to-cylinder and cycle-to-cycle) Acquired data frames available for processing with fast recursive algorithms
TDC synchronous results sending to external devices (bench supervisors, ECU) Data can be saved on PC hard disk drive for a specified number of consecutive engine cycles
post-processing purpose
Signal Processing, Automation and Control department – 2008, June 3rd12
©IF
P
Platform description : software implementationTDC synchronous task implemented in Simulink
TDC event
Processing
Engine speed
cylinder number
Pcyl
Psural
Tadm
Engine_Control
angle LPP
Pcylmax
LMG
dPda max
CA10
CA50
CA90
ROHR max
ROHR max angle
IMEP
AVLNoise
Tmax
Tmean
CA01
TAVI2
Interrupt _Source
Board : IFP -TIMER V 2
Interrupt Source
IT TDC 1
IT TDC 2
IT TDC 3
IT TDC 4
CAN_TX
LPP
Pcyl max
LMG
dPda max
CA10
CA50
CA90
ROHRmax
ROHRmax angle
IMEP
AVL
Tmax
Tmean
CA01
TAVI2
CAN_RX
Psural
Tadm
Engine_Control_parameters
Acquisition
IT_TDC_cyl 1
IT_TDC_cyl 2
IT_TDC_cyl 3
IT_TDC_cyl 4
cylinder _number
Pcyls
Trigger
f()
Engine Speed
1
Signal Processing, Automation and Control department – 2008, June 3rd13
©IF
P
Outline
IntroductionRapid prototyping platform for real-time signal processing algorithmsAlgorithm implementation for combustion analysisDeployment on industrial DSP based targetConclusion
Signal Processing, Automation and Control department – 2008, June 3rd14
©IF
P
Combustion parameters computationOnline TDC synchronous computation tasks include combustion analysis :
indicated mean effective pressure, indicated Torquemaximum pressure gradient and its locationpeak pressure and its locationmaximum rate of heat release and its locationlocations of 10-50-90% of mass fuel burnt
Mass fuel burntRate of heat releaseCylinder pressure
Pmax
dP/dθmax
ROHR max
MFB90
MFB50
MFB10
Signal Processing, Automation and Control department – 2008, June 3rd15
©IF
P
AVL noise computation Noise computation algorithm is implemented using AVL Noisemeter
specifications for engine structure filter shape and human ear filter shapeRoot mean square value is computed on filtered frameModeled noise is obtained in dB
StructureFilter
EarFilter RMS
Acquired pressure frame
dB Computed noise
Signal Processing, Automation and Control department – 2008, June 3rd16
©IF
P
AVL noise computation: filters implementation
Denominator coefficients Numerator coefficients (LP) Numerator coefficients (HP)
Fixed angular resolution with varying engine speed give varying sampling frequencyFilters coefficients updated every TDC, in order to maintain absolute bandwidthApproximation : engine speed considered constant during a cycle
Signal Processing, Automation and Control department – 2008, June 3rd17
©IF
P
AVL noise computation: filters implementation Implementation in Simulink, using Signal Processing Blockset and
Embedded Matlab functions
AVLNoise1
dB conversion
dBbeta
b
alpha
a
Window selection
Windowing
RMS computation
RMS
Offset
Offset
Ne to Fs
-K-
Fp2
Fp2
Fp1
Fp1
Filt U analytic
Fs
Fp
Fc
a
b
HP_num4
HP_den4
LP_num4
LP_den4
Structure _Filter _Coeffs
Filt A analytic
Fs
Fp
Fc
HP_num4
HP_den4
LP_num4
LP_den4
Ear_Filter _Coeffs
Fc2
Fc2
Fc1
Fc1
Digital Filter 3
DigitalFilter
Num
Den
In
Out
Digital Filter 2
DigitalFilter
Num
Den
In
Out
Digital Filter 1
DigitalFilter
Num
Den
In
Out
Digital Filter
DigitalFilter
Num
Den
In
Out
Cylinder Number3
Engine speed2
Cylinder pressure1 data
data
Signal Processing, Automation and Control department – 2008, June 3rd18
©IF
P
AVL noise computation: filters implementation Implementation in Simulink, using Signal Processing Blockset and
Embedded Matlab functions
AVLNoise1
dB conversion
dBbeta
b
alpha
a
Window selection
Windowing
RMS computation
RMS
Offset
Offset
Ne to Fs
-K-
Fp2
Fp2
Fp1
Fp1
Filt U analytic
Fs
Fp
Fc
a
b
HP_num4
HP_den4
LP_num4
LP_den4
Structure _Filter _Coeffs
Filt A analytic
Fs
Fp
Fc
HP_num4
HP_den4
LP_num4
LP_den4
Ear_Filter _Coeffs
Fc2
Fc2
Fc1
Fc1
Digital Filter 3
DigitalFilter
Num
Den
In
Out
Digital Filter 2
DigitalFilter
Num
Den
In
Out
Digital Filter 1
DigitalFilter
Num
Den
In
Out
Digital Filter
DigitalFilter
Num
Den
In
Out
Cylinder Number3
Engine speed2
Cylinder pressure1 data
data
Signal Processing, Automation and Control department – 2008, June 3rd19
©IF
P
Model-based design, implementation and integrationUse of same Simulink model along all development process
Specification and SimulationSimulink is the receptacle of the functionalities to be developedSimulink allows to test algorithm reliability offline, using simulation
Development and test of the platform's driversIntegration with algorithms in a whole system model
system testing in real time, on xPC Target environmentValidation of the final executable application
system validation in HIL conditions, with an engine signals simulatorfunctional validation on a test bench or in a vehicle calibration online, using Simulink external mode or GUIs
Signal Processing, Automation and Control department – 2008, June 3rd20
©IF
P
Model-based design, implementation and integration
Acquired data exploitation in model-based designThe platform's data storage ability permits to feed a full database with a complete engine mappingWith an offline analysis of this data, computation algorithms can be tested and pre-tuned from the early development and simulation phases
Signal Processing, Automation and Control department – 2008, June 3rd21
©IF
P
Outline
IntroductionRapid prototyping platform for real-time signal processing algorithmsAlgorithm implementation for combustion analysisDeployment on industrial DSP based targetConclusion
Signal Processing, Automation and Control department – 2008, June 3rd22
©IF
P
Issue and functionalities
Need of an embedded industrial and energy-cost effective solution to deploy developed algorithms rapidly on vehicles and test benches to provide efficient standalone tools for combustion closed-loop control achievement
Cost and time effective achievement Minimize modifications while moving validated application from prototyping platform to deployment hardwareExploit existing targets and IDE
Signal Processing, Automation and Control department – 2008, June 3rd23
©IF
P
Issue and functionalities
IFP has developed an industrial solution based on FPGA-DSP (TI C6727) hardwareThe functionalities to be embedded are:
in-cylinder pressure acquisition and samplingcombustion analysis and noise computation cylinder-to-cylinder and cycle-to-cycle updating of combustion parametersresult availability to external devicesalgorithm parameterization from external devices
Signal Processing, Automation and Control department – 2008, June 3rd24
©IF
P
Hardware description
This device is composed of:A customized timer board
includes an FPGA, which manages:engine and angular coder signals, such as cycle trigger and 0.1°CA trigger acquisitionscommunication protocols with external devices (mainly CAN)memory mapping
An 8-channel acquisition board 8 dedicated 16-bit ADCs high-frequency acquisition (800 kHz or 0.1°CA)
A TI C6727 DSP moduletargeted by Real-Time Workshop to execute a Simulink modeled application
Signal Processing, Automation and Control department – 2008, June 3rd25
©IF
P
Hardware description
User interfaceLCD display
Digital conditioning
CAN
USB 2.0 interface
Analog conditioning
Synchronization signals :- angular coder- camshaft- reference cylinder TDC
Sensors :- cylinder pressure- AFR- Fuel pressure
External devices :- ECU- bench supervisor- prototyping platform- Monitoring facility
Targeted DSP
TI C 6727
FPGA
LAN 10/100M
Deployment platform scheme
Signal Processing, Automation and Control department – 2008, June 3rd26
©IF
P
Software implementation
Aim: bring the application from a PC-based rapid prototyping environment to an industrial DSP-based targetDirect use of the Simulink code implemented on prototyping platformAdditional tools exploited in the code generation process:
Real-Time Workshop® Embedded Coder™Target Support Package™ TC6 (for TI’s C6000™ DSP) and its corresponding target function library (TFL)Embedded IDE Link™ CC (for TI’s Code Composer Studio™)
Signal Processing, Automation and Control department – 2008, June 3rd27
©IF
P
Software implementationReal-Time Workshop Embedded Coder configuration using TI C67x library
Using TFL instead of complete ANSI C code generation has improved code execution performance by a factor of 5, for frame-based AVL computation algorithm
Signal Processing, Automation and Control department – 2008, June 3rd28
©IF
P
Software implementationThe application embedded in the DSP gets acquired data and givescomputation results
Data accessibility management by mapping DSP memoryMemory addresses specified by code variablesAdd of specific headers and initialization code in the model and in the custom target configuration.
Signal Processing, Automation and Control department – 2008, June 3rd29
©IF
P
Outline
IntroductionRapid prototyping platform for real-time signal processing algorithmsAlgorithm implementation for combustion analysisDeployment on industrial DSP based targetConclusion
Signal Processing, Automation and Control department – 2008, June 3rd30
©IF
P
Conclusions & PerspectivesFuture engine technologies applying new combustion systems require a closed loop combustion controlIFP has developed a rapid prototyping platform for high frequency data acquisition and signal processing algorithm developmentBy acquiring in-cylinder pressures, online analysis is possible in real-time TDC synchronously, allowing for the estimation of combustion phasing and noiseValidated algorithms are being deployed on a standalone industrial DSP-based solution developed at IFPOther issues addressed with signal processing platform :
Fuel pressure measurement for diagnostic and control purposeAFR rapid measurements for control purposeInstantaneous engine speed measurement for torque estimationCAI applications
Implementation, integration, and calibration phases of processing algorithms have been simplified and accelerated thanks to the use of The MathWorkstoolchain and model based design approach
Signal Processing, Automation and Control department – 2008, June 3rd31
©IF
P
Thank you!
2008-01-0790
Signal Processing, Automation and Control department – 2008, June 3rd32
©IF
P
Annex : ACEbox Control System