Process Process Model input - output Residual generation Residual evaluation + Knowledge of faults Aug. 2017 - Feb. 2021 March 2021- Model-based development with “eFMI” From Physical Models to ECU Software Recorded July 17, 2021 Oliver Lenord (Bosch Research) voice
Transcript
Tips & Tricks Project Presentation <Delete this slide from
your final slide set>Model-based development with “eFMI” From
Physical Models to ECU Software
Recorded July 17, 2021 Oliver Lenord (Bosch Research) voice
Presenter
Presentation Notes
Welcome to “Model-based development with eFMI”. In the next 8
Minutes you’ll learn about the new eFMI standard its motivation,
application and benefits. eFMI is one of the key achievements of
the publicly funded European ITEA project EMPHYSIS over the past 3½
years. The eFMI specification has been handed over to the Modelica
Association for further development in preparation of the first
official release expected within 2021.
2
Developer
Su pp
lie rseFMI
Publicly Funded Project EMPHYSIS Partners by Country and Position
in the Value Chain
OEM
Germany Bosch DLR ETAS ESI ITI AbsInt PikeTec dSPACE EFS
Sweden Dassault Systèmes AB Volvo Cars Modelon Linköping University
SICS East
France Siemens SAS Dassault Systèmes SE Renault CEA University of
Grenoble FH Electronics OSE Soben
Belgium Siemens NV Dana University of Antwerp
Canada Maplesoft
Presenter
Presentation Notes
The EMPHYSIS project with 25 partners from 5 countries was led by
Prof. Martin Otter from DLR and Oliver Lenord from Bosch. The
partners of the consortium and the members of the associated OEM
Advisory Board cover the entire value chain from physical modeling
to ECU software including tool vendors as well as automotive
suppliers and OEMs. Leading companies ensured a good market access.
The OEM Advisory Board provided the market pull and direction. All
partners, even competing companies, joint forces to work out a
conceptually compelling and technically outstanding open standard
to enable new ways of model-based development of ECU
software.
3
5
Benefits of Lower SW Complexity
Less SW maintenance effort
Less SW calibration effort
Less ECU resources demand
Area LoC (Lines of Code) Color BMI (Bosch Maintainability Index
)
Engine Control SW Complexity Measurement
Presenter
Presentation Notes
The motivation for Bosch Corporate Research to initiate this
project back in 2015 was driven by the business needs expressed by
our industrial partners in the Bosch business units. A growing
software stack whose complexity has become a threat in terms of
maintenance costs and a limiting factor for new innovative
functions for systems of increasing complexity. New ways of
developing embedded software in a faster and more sustainable way
was the ambitious project goal.
4
Presentation Notes
Starting point of the eFMI approach is the observation that
understanding how things work is fundamental for developing
successful products. <click> Modeling & simulation has
been used over decades to build-up and utilize this knowledge.
Thousands of engineers at Bosch leverage simulation techniques to
<click> build the high quality products we’re selling world
wide. With “code” being the fuel of the 21st century, <click>
software has become a driver for new innovations for smarter
products. Finding ways to dissolve simulation models into software
to inject them directly into the products is the key to derive
advanced functions faster. <click> Bridging the gap between
the modeling and simulation domain and the embedded software world
is the core idea behind the EMPHYSIS project leading to eFMI as new
exchange format between these domains.
5
Presentation Notes
This idea has not only been specified on paper in form of the eFMI
specification, but also realized by a large number of tools from
modeling tools over ECU code generating tools to validation and
verification tools. With 13 tool prototypes at a high level of
maturity one could say that eFMI is established before its official
release. The developed tool chain has been thoroughly tested and
cross checked. An eFMU Compliance Checker is available open source
for everybody to verify their implementation against the
standard.
6
ECU Runtime Performance: In all cases the eFMI generated code
is
below the +25% KPI margin. In 5 of 6 examples an eFMI exists
that
outperforms the hand code. In average the best performing eFMUs
are
26% faster than the hand code.
eFMI Readiness D7.2 eFMI Performance Assessment (Bosch)
# Name Difficulty* Average Min. Max. M03 PID low -7% -27% +29% M04
Drivetrain medium +9% -21% +44% M15 Air System medium +38% -7%
+132% M10 Inverse Slider Crank high -65% -66% -64% M16 ROM high +4%
+1% +6% M14 Rectifier high +3% -33% +44%
Average -3% -26% 32%
ECU Runtime Performance
eFMI
KPI
*Difficulty for an automated procedure to achieve same quality as
manual implementation.
Relative ECU Runtime
Textual modeling compact formulation
Presentation Notes
Key for Bosch was to assess the quality of the generated code. It
was clear to us that compromising on runtime performance and
resource demand would undermine the user acceptance. A better
productivity through an automated process was considered acceptable
at the cost of 25% loss. Six test cases have been defined by our
expert embedded software developers and manually implemented to be
the baseline for the benchmark of the eFMU code evaluated on a
Bosch MDG1 control unit. As a result of this assessment did the
auto-generated code in all case reach the KPI of less than 25% and
to our surprise even outperformed the hand coded solution in 5 out
of the 6 examples. This is a technically outstanding result.
7
93 %
52 %
18 %
-20%
0%
20%
40%
60%
80%
100%
0
20
40
60
80
100
120
140
Manual eFMI Manual eFMI Manual eFMI Manual eFMI Manual eFMI Manual
eFMI
PID Drivetrain Inverse Slider Crank Rectifier Air System ROM
M03 M04 M10 M14 M15 M16
De ve
lo pm
Better code, less effort!
Presentation Notes
Furthermore the working hours have been counted for both the eFMI
workflow and the state-of-the-art manual implementation. As a
result it could be shown that especially for the examples taking
advantage of a high level of reuse based on Modelica libraries a
tremendous gain in productivity can be achieved. Bottom-line:
<click> eFMI enables better code with less effort.
8
Binary Code
Advanced Emergency Braking System controller
Hybrid engine torque prediction using scale model Neural
Network
Kalman Filter air filling estimation using scale model Neural
Network predictor
D7.06 Renault Demonstrator
Vehicle dynamics control by Parameterized Nonlinear Model
Predictive Control for semi-active control with Neural Network
prediction model
Transmission model as virtual sensor
EMPHYSIS Demonstrators
D7.4 Model-based Diagnosis of Thermo System
eFMI DAE Eq. Code
D7.14 Dassault Systèmes Demonstrator
Presenter
Presentation Notes
Over the course of the project nine demonstrators have been
developed to show case and evaluate the industrial grade readiness
of the eFMI tool chain. The impressive results include a wide
variety of physics-based state estimators that convinced not only
our industrial partners but also the ITEA review board to consider
the EMPHYSIS project with excellent ratings. Finally we're looking
forward to the ITEA Award Ceremony in September 2021, when the
EMPHYSIS project will be honored with the Vice-chairman Award of
Excellence for outstanding results in all three categories:
innovation, business impact and standardization.
9
Automation Model Transformations Code Generation
Seamless Tool Chain
Separation of Concerns Physical Behavior Data Flow Embedded
Code
Software Innovations Tool Vendors Added Value Expand Market in MBD
Domain
Supplier/OEM New Advanced Functions Replace HW with SW New Modes of
Collaboration
eFMI Outlook Business Impact
Component Libraries
Data Flow
Services Functions RTPC, e.g. ETAS RTPC
.bin
Presentation Notes
This brings us back to our initial goals. As briefly outlined in
this talk: <click> By tearing down the walls between physical
modeling and embedded software, the overall productivity can be
largely improved. The eFMI workflow helps to manage complexity
better by a separation of concerns and allowing to manage complex
system on a higher level of abstraction. Finally all this is an
enabler of new software innovations for suppliers and OEMs to sell
more and more advanced functions and tool vendors to be among the
first to provide tool support for this breaking technology.
10
Voice of the Customer Statements of Members of the OEM Advisory
Board
“Reusing the same, universal and physical plant model as eFMU in
MiL, SiL, HiL and on the ECU is a technological breakthrough with
considerable potential to reduce the development time.” (Zdenk
Husár, Daimler)
“What we demonstrated using eFMI for the model-based development of
a virtual sensor is the way to do it.” (Per Jacobsson, Volvo
Cars)
“eFMI will revolutionize the translation of models to embedded SW.”
(Yutaka Hirano, JSAE)
JSAE: Japanese Society of Automotive Engineering
Presenter
Presentation Notes
As a closing note I want to refer to what the members of our OEM
Advisory Board said. These positive statements encourage us to
expect a strong market pull for eFMI so that in a couple years from
know we can truly say: Yes, eFMI has revolutionized the model-based
development of embedded software. Thanks for watching.
Process
Model-based development with “eFMI” From Physical Models to ECU
Software
Recorded July 14, 2021 Oliver Lenord (Bosch Research)
Look-up EMPHYSIS results: https://emphysis.github.io/ Visit us on
https://efmi-standard.org/ Join the Modelica Association Project:
MAP-efmi
https://modelica.org/
Presenter
Presentation Notes
Does this sound exciting to you also? Don’t hesitate to contact us.
We're looking forward to welcome you in the eFMI community. Wanna
get involved? Become member of the new Modelica Association project
MAP-efmi.
Model-based development with “eFMI”From Physical Models to ECU
Software
Publicly Funded Project EMPHYSISPartners by Country and Position in
the Value Chain
Why?Challenges in the Field of Automotive Embedded Systems
Why?Bridge the gap
ReadinesseFMI Tool Chain
Slide Number 8
eFMI OutlookBusiness Impact
Voice of the CustomerStatements of Members of the OEM Advisory
Board