Tools used for the development ofnew automatic climate control concepts
at DaimlerChrysler GermanyAndré Strobel, Research & Technology, REM/AC
dSPACE User Conference Japan 2004
André Manfred Strobel, REM/AC 2
Lecturer's background
André Manfred Strobel
29, married
1994 - 1999 Diploma Degree in Technical Cybernetics,Focus on Aerospace EngineeringUniversity of Stuttgart
1999 - 2002 Doctorand „Modelbased climate control“Research & Technology, DaimlerChrysler AG, Stuttgart
since 2000 Guest lecturer in Automotive EngineeringFocus on MechatronicsFHTE University of Applied Sciences, Esslingen
since 2002 Research EngineerClimate Comfort, Simulation and Concepts (SIL/HIL)Research Electronics & Mechatronics, DaimlerChrysler AG, Stuttgart
André Manfred Strobel, REM/AC 3
Contents
Tools to simplify the development of complexclimate control systems
� Rapid Control Prototyping (RCP)
� dSPACE AutoBox / MicroAutoBox
� dSPACE RapidPro System
� Hardware-in-the-loop (HIL)
� dSPACE HIL-Simulator
André Manfred Strobel, REM/AC 4
Abstract
In general the development of an electronic climate control unit (ECU) is not a straight forwardprocess. A lot of iterations are necessary between the OEM and the supplier of the ECU, becausethe specification of the desired system behavior is not only based on physical data, but also onsubjective perception. The well-known V-process model with the two branches control softwaredesign and ECU testing and calibration has to be executed several times.
In order to reduce the number of iterations that are necessary due to subjective, incomplete ormisinterpreted system specifications it is necessary to rapidly generate a variable prototype fromthe given system specification. Therewith the system specification and the subjective perceptioncan be tested even before the first prototype of the real ECU exists.
The remaining iterations should be executed very fast. This can be achieved by the intensive useof tools for all steps in the V-process model. In the ideal case a single, continuous tool chain is usedthroughout the whole development process. DaimlerChrysler Germany and many suppliers are usingdSPACE tools for climate control applications.
For the creation of a very good system specification dSPACE AutoBox and dSPACE RapidProsystems are used for Rapid Control Prototyping (RCP).
dSPACE Simulators are used for Hardware-in-the-loop (HIL) testing. This allows functional testingas well as virtual testing of customer relevant test cases. Test drives can be rerun. Integration testsare used to test the ECU prototype inside the overall ECU network.
André Manfred Strobel, REM/AC 6
ECU development processfor climate control applications
ECU development process
Implementation
Automatic code generationwith dSPACE TargetLink
Idea Product
Time
OEM
Supplier
Testing and calibration
Virtual test drives withdSPACE HIL-Simulatorand real test drives
Control software design
Control concept validationwith dSPACE Prototyper(AutoBox, RapidPro)
André Manfred Strobel, REM/AC 7
Real ECU development process
ECU development process
Idea
Product
Time
First loopSecond loop
Third loop
Final loop
Error loops
Subjective, incomplete or misinterpretedsystem specifications, general mistakes
Continuous tool chain
• Avoid error loops• Reduce time needed for one loop
André Manfred Strobel, REM/AC 9
Targets
Rapid Control Prototyping (RCP)
� High product quality in early development stages
Fully functional control system, preoptimization of parameters
� Testing the subjective perception ofa control system specification
Avoid high level error loops, e.g., misinterpreted system specifications
� Automatic generation of a prototypefrom the control system specification
Avoid low level error loops, e.g., programming mistakes
� Parallel development of the climate control systemand the according control unit
André Manfred Strobel, REM/AC 10
RCP system layout
Rapid Control Prototyping (RCP)
V220 test vehicle
• Standard 4 zone climate control system• Programmable user interface
dSPACE Tandem-AutoBox
• PowerPC 750 (480 MHz)• 32 A/D channels• 6 D/A channels• 40 Digital I/O lines• 4 RS232 channels• 4 CAN controllers• Automatic powerup/shutdown
Bypass KLA
• ECU with bypass on CAN• Selectable bypass functions
– 14 hardware outputs– 37 KLA CAN messages– 7 sensor inputs– 16 user interface keys
• KLA calibration data(314 values)
André Manfred Strobel, REM/AC 11
Rapid Control Prototyping (RCP)
dSPACE AutoBox / MicroAutoBox
� What we like
� Designed and feasible for automotive applications
� Choose between flexibility and small size
� Same software on all platforms (dSPACE RTI blocksets)
� Powerfull front-end software (dSPACE ControlDesk)
� What we miss
� Direct parameter manipulation in the Simulink modelin connection with the real-time hardware
� Power electronics not included => New: dSPACE RapidPro System
André Manfred Strobel, REM/AC 12
Rapid Control Prototyping (RCP)
Control software
� Separation of hardware access fromsoftware routines (algorithms, logic)
Transparent structure, exchangeable hardware
� 100% multitasking approach
Flat turnaround time profile, high flexibility
� Global variable structure
Management of more than 1300 variables
� Special features
FlighRecorder, nonvolatile data, menu structurefor display, powerup and shutdown routines
V220_KLA_RPT.mdl
Current number of blocks: 65026Future number of blocks: 150000
Model sample time: 10 msTurnaround time: 5 ms
André Manfred Strobel, REM/AC 13
New: dSPACE RapidPro System
� Targets
� Programmable, variable and reusable power electronics
� Integration into dSPACE tool chain
� Usable without special E/E knowledge
� No soldering (except connectors)
� Application
� Standardized, portable RCP-System
� Test of new sensor and actuator concepts
Rapid Control Prototyping (RCP)
André Manfred Strobel, REM/AC 14
Rapid Control Prototyping (RCP)
dSPACE TargetLink
� Implementation is done by a supplier
� Suppliers use dSPACE TargetLinkfor automatic code generation
� First products are on the road(fan control, intelligent solar sensor)
� Next generation of cars will containmuch more automatically generated code
� Future? 32-Bit processors with floating point arithmetics!
Idea Product
TimeImplementation
André Manfred Strobel, REM/AC 16
Targets
� Real-time simulation of the climate control system
HVAC system (heating, ventilation, air conditioning), cabin, environment
� Functional test of a prototypic ECU
Switchpoint tests, customer relevant test cases
� Integration tests for the complete ECU network
Communication tests
� Virtual rerun of test drives
� Reduce the number of test drivesfor finding basic mistakes
Hardware-in-the-loop (HIL)
André Manfred Strobel, REM/AC 17
Basic principle
Hardware-in-the-loop (HIL)
+
HIL-SimulationVirtual test drives
MeasurementReal test drives
+
André Manfred Strobel, REM/AC 18
dSPACE HIL-Simulator
� What we like
� Support including complete E/E setup of HIL-Simulator
� Flexibility of hardware layout
� Integration into dSPACE tool chain
� Powerfull front-end software (dSPACE ControlDesk)
� Automated test runs with dSPACE AutomationDesk
� What we miss
� Standard macro language(Python is powerful and for free, but everybody has to learn it first)
� Direct parameter manipulation in the Simulink modelin connection with the real-time hardware
Hardware-in-the-loop (HIL)
André Manfred Strobel, REM/AC 19
vm_hil_BR211.mdl
Model structure
� Separation of hardware access andsignal conditioning from plant models
Transparent structure, exchangeable hardware,error simulation
� 100% multitasking approach
Flat turnaround time profile, high flexibility
� Intensive use of Simulink bus signals
Busses are more than just a structured vector
� Model exchange with supplier
HIL-Simulators are built-up in cooperation with the supplier
Hardware-in-the-loop (HIL)
Plant model blocks: 33621Total model blocks: < 100000
Sample time: 1-100 msTurnaround time: < 1 ms
André Manfred Strobel, REM/AC 21
Conclusions
� We will continue to use dSPACEsystems for RCP and HIL
� The interface between the OEM and the suppliermust be managed carefully
� RCP and HIL will become standard proceduresin the development process
� Multitasking increased the performanceof our systems dramatically
� The availability of accurate real-time plant modelswill be a key success factor in the future
Summary