Realistic Dynamical Testing of EPS Motors and ECUs by Means of a Hardware-in-the-Loop Test Bench

chassis.tech 2015, Munich, 16.-17.06.2015

H.Briese, E.Farshizadeh, S.Oedekoven,
DMecS - Development of Mechatronic Systems GmbH & Co. KG, Köln;

T. Schubert, Hermann Henrichfreise
CLM - Cologne Laboratory of Mechatronics, FH Köln

Abstract

Conventional test benches with real steering mechanisms often have non-negligible parasitic properties and multiple control systems to apply stimuli. As a consequence the EPS motor and ECU interact with a system that has a different dynamic behavior than the steering mechanism in the vehicle. Depending on the actual test scenario, this might reduce quality of testing results significantly.


That was the motivation for the development of a test bench for realistic testing of EPS motors and ECUs as a complement for conventional test benches. In the test bench described here, the real steering mechanism is replaced by a test bench actuator that is mechanically connected to the EPS motor as device under test (DUT) by means of a measuring unit with a specifically tailored sensor concept. The cutting torque, which would act on the EPS motor in a real steering system, is computed by real-time simulation of a detailed model of the steering system that includes all relevant properties. Depending on the development stage, the EPS motor can be driven either by the full EPS ECU or a power stage only with e.g. a steering control implemented on a rapid control prototyping system. For HiL tests, the test bench might be used in combination with any suitable commercial vehicle model.


The realistic testing environment for the EPS motor and ECU is achieved by a highly dynamically controlled test bench actuator. The actuator control consists of an observer based state space control with extensions for compensation of nonlinear characteristics of the test bench. It is designed accounting for all dynamic components of the test bench setup: test bench actuator, measuring unit, and DUT. The actuator control adjusts the cutting torque in the test bench which would act on the EPS motor in the real steering system. This results in the EPS motor being operated in a realistic environment. The simple mechanical setup in combination with a sophisticated control leads to a high bandwidth and a testing quality that can hardly be achieved with conventional test benches. Additionally, it enables testing in early phases of development, even when the real steering mechanism might not be available yet. The model based approach allows for detailed investigation of the impact of special properties of the mechanism like backlash or altered friction on the EPS behavior. Furthermore, the influence of torque ripple of the EPS motor on the steering feel can be evaluated.

Download Publication:

DMecS_HiLTestbench4EpsMotorEcu_chassis.tech2015.pdf (420 KB)

1 Introduction

For more than a decade, DMecS is working in the field of control development for EPS systems. During this period DMecS gained considerable experience with many types and performances of test benches for steering systems. The rather high effort for development and operation of test benches for the full steering system, as well as their sometimes limited performance plus the increased requirements for HiL tests induced the motivation to develop this new test bench approach. It has been developed in cooperation with the Cologne Laboratory of Mechatronics (CLM).

The core development target was to provide a test bench for realistic testing of EPS motors and their ECUs. The test bench was designed to apply that load torque (cutting torque) to the EPS motor that would act on it when operated in the real steering system. This cutting torque is highly dynamically controlled so that the dynamics of the test bench, in the relevant range of frequencies, is not visible in the results of the HiL tests. The cutting torque acting on the EPS motor is computed by real time simulation of the steering system. According to that, the quality of tests primarily depends on the used simulation model.

To achieve a sufficiently high performance it is necessary to account for the dynamics of the entire system (test bench and DUT) for controller design. This again means that all relevant properties and parameters of the DUT must be known. For the identification of the sometimes unknown parameters of the DUT, automated parameter identification is introduced [2]. For that, additional types of controllers for the test bench have been developed. This offers additional options for parameter identification and model validation like the determination of the torque-speed characteristics without disturbing vibrations and the investigation of the torque ripple of the DUT under different load conditions.

2 HiL Test Bench

The test bench shown in figure 1 consists of the test bench actuator, a proprietary measuring unit, and the EPS motor with ECU (DUT). Furthermore the figure shows a dSPACE system used for test bench control and real-time simulation of the steering system mechanics.
Instead of a real EPS motor and ECU, a high quality permanent-magnet synchronous machine is used to emulate different types of EPS motors during test bench development. A dSPACE MicroAutoBox is used to emulate the ECU.

Figure 1: Test bench with DUT (EPS motor and ECU replacement) and dSPACE system

To achieve the high control dynamics required for realistic HiL testing the individual components were selected to satisfy special requirements. This includes, among other properties, low moments of inertia, low friction and low torque ripple at the test bench actuator. The latter is further reduced by specific control measures. In addition the sensors provide high accuracy and the actuator and sensor interfaces introduce only negligible signal delays. The actually used test bench actuator allows a maximum torque of 30 Nm. Thereby it is applicable for a high range of tests and various types of DUTs, even motor racing components.

Figure 2 shows the structure of the HiL test bench with the components described above and their interaction.

 

Figure 2: Structure of the HiL test bench

Besides the simulated test environment, in this case a simulation model of steering mechanics and driver, the test bench actuator control is implemented on the dSPACE system. It generates the control signal for the test bench actuator and provides additional signals for the simulation of the steering mechanism. The latter generates the reference signals for the test bench actuator control and the input signals for the EPS ECU.

3 HiL Test Bench Control

The realism of the testing environment for the EPS motor and ECU is achieved by a highly dynamically controlled test bench actuator which adjusts the cutting torque at the EPS motor. In addition to a high controller bandwidth, high control accuracy is requested. This must be realized despite disturbances like friction and torque ripple at the test bench actuator and measurement noise. Natural oscillations of the system should be damped actively by the control.
To achieve these goals, the cutting torque control consists of an observer based state space control. It is extended by a nonlinear model of the test bench actuator torque ripple. Figure 3 shows the corresponding control structure.

Figure 3: Structure of the test bench actuator control

The actual controller is built by a static feedback of the system state variables in the vector xp as well as feedforward of the reference variables uref and of disturbance variables in the vector

xd. Since not all system state and disturbance variables can be measured, an observer is required to estimate unknown variables by using the measurement variables in the vector ypm and the control variable yctrl. Furthermore, the estimator (Kalman filter) calculates additional variables in the vector umech that can be used for the simulation of the DUT test environment, e.g. a steering mechanism.


The controller and estimator gains are determined by the LQG/LTR design approach of an optimal robust control [3], [4].

To compensate the torque ripple of the Test bench actuator, the output signal of the torque ripple model ycog is added to the control output yctrl.