Research On Control Strategy For Auto Electric Power Steering System

Auto Electric Power Steering (EPS) System is a new technique and construction in Power Steering , and the analyze method of traditional Hydraulic Power Steering System can not be completely applicable to Electric Power Steering System. So we need to research the model and control strategies of the EPS drastically. The three-freedom nonlinear strong coupling dynamic math model of the system is established in this paper. And it researches on influence to system characteristic produced by the predigest model and its parameter. The assist character, turning following character and turning road feeling character are obtained by simulation on the predigest model. The assist character is a key technique in EPS system. Based on analysis on discrete nonlinear assist character curve, a kind of new assist character analysis method ground on Multi-layer Foreword Neural is put forward. It makes use of the generalization function of Neural to fitting the assist characteristic curve at any speed, overcoming effectively blind zone generated by discrete assist character. Validated by simulation and experiment, this method proved to be workable, valid and in general use to different model.The EPS system must be equipped with adequate frequency band to fulfill follow performance of assist motor when steering wheel turned fleetly .At the same time, it should well materialize steering road feeling. For this purpose, according to the system's construction, illustrating by dynamic compensator analysis of torque sensor, we apply root locus method to design dynamic compensator of the torque sensor, validated by experiment and simulation. Researching on the system's robustness is the important work of the paper. While the EPS system is a multi-input and multi-output nonlinear system, considering the error during modeling, methods of designing the system controller are put forward with the H∞ robust control theory. In designing H∞ controller, we adopt two main methods. First we introduce mixed –sensitivity method based on frequency-domain. The Second is linear matrix inequalities method. The validation of the designed controller is verified by the comparison between simulating result of system characteristic and ideal characteristic. It is also confirm by the ability of restrain disturb. which controller is fit to the system is validated by the simulation.