presented at the: 8th International CTI Symposium Innovative Automotive Transmissions, Berlin, 1 - 2.12.2009.
Test-driven Development of DCT Control Software
Dr. Nikolaos Papakonstantinou, Dr. Sigrid Klinger
GIF - Gesellschaft für Industrieforschung mbH
{nikolaos.pap, sigrid.klinger}@gif.net
Dr. Mugur Tatar
QTronic GmbH
Abstract
Transmission systems, such as DCTs, require systematic test and validation methods in order to
guarantee correctness and quality despite shorter development times and increasing complexity.
Therefore, during the development of a new dual clutch transmission at GIF, a process for extens-
ive test and optimization of the control software has been adopted. For this, system correctness
and quality criteria are evaluated for thousands of simulated test scenarios (driving maneuvers).
Problems and system weak points are identified and corrected as early as possible and the pro-
cess is repeated after each correction or system change. The tool TestWeaver [2] from QTronic
was used for the systematic generation and evaluation of the test scenarios. This allows achieving
high test coverage with minimal test specification effort. This extensive test process drives and ac-
celerates the optimisation of the control software. Beside MIL tests also intensive SIL, HIL and
tests with the real hardware components in the loop must be performed in order to guarantee the
functionality of the automatically generated code, the proper functionality of the TCU including
electronics and CAN communication, and the robustness of the software functions under real oper-
ating conditions [3].
1 Introduction
Transmission systems are under continuous improvement with respect to efficiency, robustness,
costs and comfort. Many of these requirements have to be addressed also by the transmission
control software, which is becoming more and more intelligent. Many driving conditions have to be
detected fast and reliably. Specific, optimized actions and control strategies have to be performed
in order to achieve the optimum balance between often conflicting goals: efficiency, agility, comfort.
Many modules are interacting and cooperating in order to achieve the system requirements. The
test and the validation of the control strategies become, however, also more difficult. Module tests
performed in isolation, no matter how extensive, cannot assess the emerging system behaviour,
while the test of the system in only a limited set of predefined test scenarios does not scale with
the huge number of differing situations that need to be tested and validated.
The software of the new GIF DCT is developed using a model-based development process with
Simulink and TargetLink. The complete control software model can be simulated in closed-loop
with a realistic plant model developed also in Simulink - including transmission hydraulics, mechan-
ics and vehicle model. For faster simulation the complete model is compiled with Real-Time Work-
shop from Mathworks. This way, the simulation runs about 20 times faster than real-time on a com-
mercial PC. The simulation includes, beside the control model and the plant model, also several
correctness and quality observers implemented in Simulink. TestWeaver then controls and evalu-
ates the simulation runs. Thousands of scenarios are automatically generated and assessed in a
reactive loop in which TestWeaver actively attempts to reach all possible system states with at
least one scenario and to find scenarios with correctness or quality problems. For instance, it will
try to reach all transmission shifts, simple (e.g. 4-5) and multiple (e.g. 6-4), with many differing
speed and torque loads, with differing street slopes, with and without braking, with or without chan-
剩余12页未读，继续阅读
资源评论
qinjianqiu
- 粉丝: 0
最新资源
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
