Electric power assist steering system parameterization and optimization employing CAE

Abstract: The vehicle brands of today need to develop high quality products in a short period of time to satisfy the consumer needs and to stand out in the competition. To be able to do this simulation tools have been used more and more in the development process. This would enable quick evaluation of different concepts and setups without the need of building physical prototypes (though physical prototypes are still common to validate development in different stages). The goal of this thesis is to contribute to this trend by developing and evaluating a process of optimizing the control parameters of an EPAS (electronic power assist steering) system by solely using computer aided engineering (CAE) tools. This tuning is done to improve the steering feel for the driver. The process could then be used to improve the initial tuning of such a system or in a later stage completely parameterize it. In this project, a complete process from setting up the simulation environment to developing the optimization process of the parameters was developed. A vehicle model was simulated using the real time simulation software IPG CarMaker, parameterized by multi-body simulation results and tyre measurements. The steering system itself was modelled by the supplier in Simulink so that the entire vehicle could be co-simulated using both CarMaker and Simulink. To evaluate the performance of the steering system objective metrics were used, each of which had a target range. The optimization process itself utilizes the optimization tool Tomlab, which is a powerful Matlab toolbox. Tomlab uses a complex strategy of evaluating the response of a system when the relationship between the inputs (in this case the EPAS control parameters) and the outputs (in this case the objective metrics) is unknown. A cost function using the difference between the target metric value and the actual metric value was set up and the weighing factors for each of the tested metrics were determined by their sensitivity to the optimized parameter. The developed process was tested on three specific metrics for two specific parameters of the controller. These metrics were the friction feel, the torque build-up and the torque deadband, all three of which relates to the torque feedback of the steering wheel. All three metric values had room for improvement but after going through the developed process they could be adapted to be closer to the optimal target. These results need to be validated, but as a proof of concept the work described in this thesis shows that it is possible to parameterize a vehicle subsystem by only using CAE tools.