Proceedings of the 2006 IEEE Conference on Computer Aided Control Systems Design Munich, Germany, October 4-6, 2006
FrB02.3
Rapid Control Prototyping of Active Vibration Control Systems in Automotive Applications Konrad Kowalczyk, Hans-J¨urgen Karkosch, Peter M. Marienfeld, and Ferdinand Svaricek Abstract— This paper discusses the rapid controller prototyping approach used at Continental and the University of the German Armed Forces for the design and implementation of active vibration control systems. Continental has developed and implemented prototypes of active engine mounting systems on various test vehicles and demonstrated that significant reductions in noise and vibration levels are achievable.
180◦ out of phase from the original vibration signal, so that it cancels the effect of incoming excitations.
Engine Passive mounts
I. INTRODUCTION Chassis
Modern control applications are becoming increasingly important in the area of vehicle riding comfort. An attractive application in this area is the use of active vibration control in engine mounting concepts, particularly since conventional mounts are approaching their inherent limitations. The standard approach is to isolate the engine and the transmission vibrations from the chassis with rubber or hydro mounts. This mount design is always a compromise between the conflicting requirements of acceptable vibration isolation and engine movement. Engine mounts in automotive vehicles are designed according to the following requirements: • • •
holding the static engine load, limiting engine movement due to powertrain forces and road excitations, and isolating the engine and the transmission from the chassis.
In order to limit engine movement, it is desired to design a very stiff mount. However, to minimise transmission of engine vibrations into the passenger compartment, a very soft engine mount is required. These requirements contradict each other. A solution for this design conflict could be an active engine mount system. The active vibration control (AVC) system generates dynamic forces to cancel the effect of incoming excitations. Research and development activities have focused on the transmission of engine-induced vibrations through engine and powertrain mounts into the chassis. For references see [1] and [2]. A schematic representation of such a system is shown in Fig. 1. The basic principle is to synthesise a waveform that is identical in magnitude, but Konrad Kowalczyk and Ferdinand Svaricek are with Faculty of Aviation and Space Engineering, System Dynamics and Flight Mechanics, University of the German Armed Forces, Munich, 85577 Neubiberg, Germany, E-mail:
[email protected] Hans-J¨urgen Karkosch and Peter M. Marienfeld are with ContiTech Vibration Control GmbH, 30419 Hannover, Germany, E-mail: hans-
[email protected]
0-7803-9797-5/06/$20.00 ©2006 IEEE
Error sensor
Actuator
Amplifier Fig. 1.
ECU
Engine speed
Active vibration isolation in automotive vehicles.
The reminder of this paper will present an overview of model-based development of control algorithms, a short description of system components, as well as obtained experimental results and is organised as follows. In Section 2, the AVC system installed in the test vehicle is described. This is followed by an explicated presentation of tools for the development of control functions in automotive industry. Section 4 will give a presentation of experimental evaluations. The paper will end up with conclusions and overview of future works.
II. SYSTEM DESCRIPTION The disturbance force originating from the engine and transmitted into the chassis through the engine mount is actively cancelled by the force generated through an electromagnetic inertia-mass actuator of a kind reported in [3]. This actuator is driven by a power amplifier that converts the voltage signal from an electronic control unit into an actuator current. The control signal u is the input of the power amplifier (a voltage), and the output signal y is the acceleration sensor output (also a voltage). Since the controlled signal is the chassis acceleration, the output signal is scaled to m/s2 (which is more meaningful than the sensor voltage). Also, the current flowing through the actuator is more relevant than the amplifier input voltage, therefore, the input signal u is scaled to amperes. Fig. 2 shows the location
2677
of the acceleration sensor and actuator on the transmission crossmember in a test vehicle. The control algorithm is implemented on a rapid prototyping unit, the dSPACE MicroAutoBox. The electronic hardware consists of an amplifier and filter unit that contains the power amplifier and the anti-aliasing filter for the sensor signal, and the electronic control unit (see Fig. 3). A remote control on/off switch is used to turn the control algorithm on and off during vehicle tests.
to be developed according to the so-called ”V-Cycle”. The schematic diagram presented in Fig. 4 and the following subsections describe this development process based on the example of implementing algorithms for active vibration control in automotive applications [5], [6].
Design and Simulation
Parameter Calibration
Validation with Rapid Prototyping
Verification with Hardware-inthe-Loop
Target Code and Hardware
Fig. 4.
Implementation of control functions according to V-Cycle.
A. Design and Simulation Fig. 2.
Actuator and acceleration sensor on the car chassis.
Fig. 3.
Electronic hardware in the test vehicle.
III. V-MODEL CONCEPT FOR IMPLEMENTATION OF CONTROL FUNCTIONS The automotive industry is increasingly adopting modelbased design methods and using automatic code generators for software development. This is to cope with rising demands, such as the growing number of electronic systems in a vehicle, increasing complexity and shorter time-to-market [4]. A modern rapid prototyping system comes with tools for control design, simulation, automatic code generation and hardware calibration. These tools allow an application
The first step of the V-Cycle is the implementation of the control algorithms in a simulation environment. A widely used tool is the MATLAB/Simulink/Stateflow package. In such an environment, the developer can design and simulate his algorithms. A graphical block diagram description is used for the implementation of the algorithm. There is a number of possible control strategies for active vibration control, that have been already reported in the literature. A reference is given, for example, in [7] and [8]. Two approches for active control of engine-induced vibrations in automotive vehicles have been investigated and reported in [9]. These are an adaptive feedforward structure based on a digital filter with the least mean squares technique for adjusting the filter coefficients, and a disturbance observer combined with a state feedback. For the simulation of a control system, a plant model, typical controller inputs and disturbance signals are required. The model of the plant to be controlled in this application is the transfer function from the amplifier input to the (filtered) sensor output. The approach taken here is to excite the system with a test signal and record the response. Any of the discrete-time black-box system identification techniques (such as the least squares approach for equation-error models) can then be used to identify the model [10]. Such a model is then used for the controller design and for simulation studies. An example of the obtained transfer functions is presented in Fig. 5 and Fig. 6. As the further step, the developed control algorithm can be investigated in simulation studies. Signals recorded from the sensors installed in the car are used as controller input signals in the simulation. Through a simulation of the system with the designed controller, the designer can verify the stability and performance of the closed-loop system. Particularly, for
2678
5000 Engine speed [rpm]
4.5 4
3
−2
Magnitude [ms /A]
3.5
2.5
4000 3000 2000 1000
2
0
0
5
10
1.5
15 Time [s]
20
25
30
25
30
1 0.5 0
Fig. 7. 0
200
400 Frequency [Hz]
600
Engine speed plot.
800
Fig. 5. Magnitude plot of the plant transfer function (actuator current to filtered sensor output).
180
Acceleration [m/s2]
15 10 5 0 −5 −10 −15
0
5
10
Phase [°]
90
Fig. 8.
0
15 Time [s]
20
Measured acceleration signal on the chassis with control off.
Acceleration [m/s2]
15 −90
0
Fig. 6.
200
400 Frequency [Hz]
600
800
5 0 −5 −10 −15
Phase plot of the plant transfer function.
non-stationary conditions, such as ramping-up the engine, the advantage of the simulation environment is an easy comparison of the results for the control-off and control-on case. The Figures 7 to 10 present the simulation study for the engine-ramp-up case. The engine speed is driven from idle to the maximum value and back. The time history of the measured acceleration signal on the car chassis is presented in Fig. 8. The controller is activated after reaching the 1700 rpm (past 7 seconds). The obtained acceleration signal with control on and generated output of the controller are given in Fig. 9 and 10.
0
Fig. 9.
5
10
15 Time [s]
20
25
Simulated acceleration signal on the chassis with control on.
5 0 −5 −10
0
5
Fig. 10.
10
15 Time [s]
20
25
Generated controller output.
B. Controller Validation with Rapid Prototyping Once the control algorithm has been designed and simulated, it can be easily implemented on the prototyping hardware. The designer needs only to specify which I/O channel will be used (i.e., the interface to the real world) and all other implementation aspects are left to the rapid prototyping software, thus freeing them from hardware restrictions and tedious low-level software details. The designer can configure the I/O interfaces straight in the block diagram. Since the code is directly generated from the Simulink model, there is no risk of misinterpreting the block functions. A simple example of an implemented controller is presented in Fig. 11. 2679
30
10 Control signal [A]
−180
10
Fig. 11.
Controller implementation with Matlab/Simulink.
30
The prototyping hardware includes plenty of analog and digital I/O channels as well as CAN, LIN and FlexRay interfaces and the capability of measuring and generating PWM and FM signals. Supplementary hardware can be added for more demanding applications. It uses automotive compatible signals and can be easily installed in a vehicle. At this point the controller can be tested together with other components of the system: amplifiers, actuators, sensors, analog filters and the car itself. Usually the controller has to be tuned, but in some cases it might be also necessary to change the control strategy due to the physical and electrical characteristics of the system. Finally, a suit of tests will be carried out to validate the controller. At this step, a very useful tool is an experiment software. Such a tool provides a possibility for the designer to manage the real-time experiment on-line, change parameters in real-time, and track all the data generated during the experiment. Some results of the experimental evaluation are presented in the Section 4. The performance achieved by the rapid prototyping system will be used as a reference for the final system. C. Production Code Generation In all previous steps, the control algorithm has been considered independent of the hardware platform and has been designed without taking limitations into account, since the prototyping hardware is several times more powerful than the target system. The controller function has been validated and reviewed in the real-time, but it is time to take a down-toearth approach and analyse other aspects of the target system: limited memory size, limited speed and limited precision. Traditionally, the function developer handed over the specifications to the software specialist who wrote the code and optimised it to meet the hardware requirements, but this process was time consuming and introduced new bugs. New automatic production code generation tools produce target-specific or ANSI C code from Simulink models, thus the consistency with the original design is kept. The target system is usually based on a fixed-point processor and numerical stability is guaranteed by performing Softwarein-the-Loop and Processor-in-the-Loop simulations. Flow, interblock and target-specific optimisations are performed during the code-generation process so that the size and speed of the generated code are comparable to the manually written one [11]. It is also well structured and readable. The latest versions of the tools also assist the developer in integrating the code with the operating system and configuring the task manager of the microprocessor system. D. Hardware-in-the-Loop Simulation When an operative target hardware is available, it is necessary to test this target system intensively. In vehicle tests are both time consuming and expensive. An alternative way to verify the ECU is hardware-in-the-loop (HIL) simulation [12]. In HIL simulation, the outside environment (i.e., the vehicle) is replaced by a simulation model,
but the real target hardware is used. In this simulation, real driving scenarios are tested. The most important advantage of hardware-in-the-loop simulation is that the simulation can run automatically. Therefore, the function verification and diagnosis can be carried out very efficiently. E. Calibration The calibration itself begins early in the development process and parameter description files can be seamlessly exchanged between different steps of the V-Cycle. The tool chain supports test automation, calibration according to ASAM standards [13] and interaction with other widely used calibration tools. Data management, evaluation and test report capabilities are invaluable when coping with a large amount of information. F. Benefits for Developers and Customers A development environment for ECUs that supports the V-Cycle strategy makes the development process easier and more transparent. One single development environment that offers a common interface for all steps including automatic code generation eliminates the need for the developer to familiarise themself with different tools. Furtheremore, the necessity for rewriting code at different stages (which would be a source for further errors) is made obsolete. Therefore, the time-to-market for a product decreases and special customers requirements can be realised faster. There is no need to change the low-level C or assembler code manually if the control algorithm is modified.
IV. EXPERIMENTAL EVALUATION In the last years, several vehicles - with different vibration problems - have been equipped with active absorber systems to attenuate the transmission of the engine vibrations into the vehicle cabin. The developed control prototyping systems were tested in miscellaneous real-time experiments with test vehicles. The following subsections present some recent results. A. Comparison between Experimental Results and Simulations As mentioned above, the performance obtained through the simulation studies can be used as a reference for the real-time implementation. In this subsection, a comparison between acceleration signals achieved by a drive test and simulations is presented. Fig. 12 shows an acceleration signal on the engine mount taken for a constant engine speed. The control-off part of the signal is later used in the simulation. The attenuation performance of simulation, Fig. 13, and drive test, Fig. 12, is comparable.
2680
10
0.6
0 −10
0.5 2
RMS amplitude [m/s ]
Acceleration[m/s2]
20
−20 0
2
4
6
8
10
Time[s]
Fig. 12. Time plot of the acceleration signal taken during a drive test. Control on after aproximately 5 sec.
Acceleration[m/s2]
20
0.4
0.3
0.2
0.1
10
0 0
0
50
100 150 200 250 300 350 400 450 500 Frequency [Hz]
−10 −20
0
1
2
3
4
Fig. 15. Amplitude spectrum of the acceleration signal. Thin line: control off; bold line: control on.
5
Time[s]
B. Drive Test at 3000 rpm, in 3
rd
2
Gear
A number of real-time experiments for a constant engine speed have been maintained. Fig. 14 presents the time history of the vibration signal amplitude taken while driving on the motorway at the approximate vehicle speed of 80 km/h and the engine speed of 3000 rpm. The control is applied after about 15 seconds from the beginning of the experiment. 76 % of the power in the acceleration signal on the car chassis was eliminated by applying the control.
10 0 −10 −20
0
5
10
15 Time [s]
20
25
30
Fig. 16. Time plot of the acceleration signal. Control on after aproximately 15 sec.
The corresponding amplitude spectrum plot is presented in Fig. 17. The particular engine orders, respective to the 4., 7., and 8. of the engine firing frequency, are clearly damped.
5 0
3 −5 5
10
15 Time [s]
20
25
30
RMS amplitude [m/s ]
0
2
Acceleration [m/s2]
Acceleration [m/s ]
Fig. 13. Time plot of the acceleration signal for control on obtained through simulation.
Fig. 14. Time plot of the acceleration signal. Control on after aproximately 15 sec.
The effect of the active control system on the vibration signal on the car chassis can be also seen in the amplitude spectrum plot for the case with control-off and control-on (Fig. 15). The controller has been tuned to cancel integer multiples of the firing frequencies (i.e., half the engine orders), and this control objective is clearly achieved. The dominant frequencies in this case, corresponding to the 4., 7., 9., 10., 11., and 12. of the engine firing frequency, are significantly attenuated.
2.5 2 1.5 1 0.5 0
0
50
100 150 200 250 300 350 400 450 500 Frequency [Hz]
Fig. 17. Amplitude spectrum of the acceleration signal. Thin line: control off; bold line: control on.
D. Sound Pressure Level Measurement
C. Drive Test at 4000 rpm, in 4 th Gear This experiment is maintained while driving a test vehicle on the motorway at the approximate speed of 160 km/h. 91 % of the power in the acceleration signal on the car chassis was eliminated by applying the control. Fig. 16 presents the time history of the amplitude measured by the sensor.
It is well-known that part of the transmitted vibration energy through the mounts passes through the chassis and emanates in the vehicle passenger compartment in the form of structure-borne noise. For the car passengers is not only the attenuation of the acceleration signal on the engine mount of interest, but also the reduction in the sound pressure level
2681
inside the car compartment. The active vibration control system presented above achieves a significant attenuation in the sound pressure level. The improvement can be seen in the confrontation of sound pressure measurement taken while speeding-up the test vehicle at full throttle for the case with control-off and control-on. The obtained result, given in Fig. 18, presents an attenuation up to 5 dB(A) in sound pressure level. During the drive tests, one could clearly notice the sound reduction. This fact confirms that active vibration control applied to engine mounting system provides an attenuation in sound pressure level inside the car compartment.
Fig. 18. Sound pressure level at the rider’s left ear position. Measurements taken at full throttle, from 1800 rpm to 4500 rpm, in 3rd gear
R EFERENCES [1] H.-J. Karkosch, F. Svaricek, R.A. Shoureshi and J.L. Vance, ”Automotive applications of active vibration control”, Proceedings of the European Control Conference, Karlsruhe, Germany, 1999. [2] C. Bohn, A. Cortabarria, V. H¨artel and K. Kowalczyk, Active control of engine-induced vibrations in automotive vehicles using disturbance observer gain scheduling, Control Engineering Practice, vol. 12, 2004, pp 1029-1039. [3] C. Hartwig, M. Haase, M. Hofmann and H.-J. Karkosch, ”Electromagnetic actuators for active engine vibration cancellation”, Proceedings of the 7th International Conference on New Actuators ACTUATOR 2000, Bremen, Germany, 2000. [4] H. Hanselmann, Beschleunigte Mechatronik-Entwicklung durch Rapid Control Prototyping und Hardware-in-the-Loop-Simulation, Automatisierungstechnik 46, 1998, S. 113-119. [5] C. Bohn, H.-J. Karkosch, P.M. Marienfeld, F. Svaricek, ”Automotive applications of rapid prototyping for active vibration control”, Proceedings of the 3rd IFAC Workshop on Advances in Automotive Control, Karlsruhe, Germany, 2000. [6] A. Cortabarria, C. Bohn, J. Hanna, H.-J. Karkosch, P.M. Marienfeld and F. Svaricek, ”Design, Simulation and Implementation of Active Vibration Control Systems in Automotive Vehicles”, Proceedings of the 3rd VDI/VDE-GMA ”Rechnergest¨utzter Entwurf von Regelungssystemen”, Dresden, Germany, 2001. [7] C.H. Hansen and S.D. Snyder, Active Control of Noise and Vibration, E & FN Spon, London; 1997. [8] S.M. Kuo and D.R. Morgan, Active Noise Control Systems, John Wiley and Sons, New York; 1996. [9] K. Kowalczyk and F. Svaricek, ”Experimental Robustness of FxLMS and Disturbance-Observer Algorithms for Active Vibration Control in Automotive Applications”, Proceedings of the 16th IFAC World Congress, Prag, 2005. [10] L. Ljung, System Identification. Theory for the User, Prentice Hall, New Jersey; 1999. [11] H. Hanselmann, U. Kiffmeier, L. K¨oster, M Meyer and A. R¨ukgauer, ”Production Quality Code Generation from Simulink Block Diagrams”, Proceedings of the International Symposium on Computer Aided Control System Design, Hawaii, USA, 1999. [12] F. Svaricek, Automatic Valuation and Verification of ABS Controllers by Using a Hardware-in-the-Loop Simulation, SAE Paper No.980241 1998. [13] ASAM Homepage: http://www.asam.net
V. CONCLUSIONS AND FUTURE WORKS A. Conclusions The control design and implementation step for such systems at Continental is largely supported by computeraided control system design tools (CACSD) and rapid prototyping systems. The use of high-level simulation and control design tools and the implementation on rapid prototyping system constitute the first two steps of the VCycle approach to automotive ECU development. At the present, these steps are very well supported by tools such as MATLAB/Simulink/Stateflow with the appropriate toolboxes and the dSPACE MicroAutoBox rapid prototyping systems. Tools for remaining steps have been also become available (such as autocode generators for automotive microcontrollers, hardware-in-the-loop simulators and calibration tools). B. Future Works Future work at Continental will focus on the suitability of the components and the algorithms for the implementation of active vibration control systems in series production vehicles. 2682
