Abstract
Keywords
Sensory science, Design of experiment, Electric Power Assisted Steering, Drivers feel, Modelization
Abstract
To optimize future steering systems, we would like to sort out the parameters of Electric Power Assisted Steering (EPAS) according to their influence on the driver feel. We have a prototype vehicle whose EPAS is fully adjustable; 8 parameters can be tuned (2 parameters of travel, 2 parameters of force and 4 parameters of friction). This includes variable effort assist and wheel return characteristics. Besides sensory descriptive criteria are needed to characterize the driver feel when steering.
A preliminary study allowed us to build a customer-oriented sensory lexicon for steering composed of 9 terms. Among these 9 terms, we decided to adapt 8 attributes to the description of the EPAS laws. 2 terms relate to the force feedback intensity, 1 to the constancy of force, 2 to the steering wheel return characteristics and 3 to the precision.
In order to study the influence of the 8 parameters of EPAS system on these 8 terms, a Plackett-Burman design of experiments is used to generate 12 different steering configurations. The vehicle speed is not a parameter of our model. To take the vehicle speed into account, we build 3 Plackett-Burman designs. Each design is evaluated at a different speed: 10, 50 and 80 Km/h.
Using sensory profile methodology, these configurations are described by a sensory panel composed of 10 subjects. The task of the subjects is to evaluate the relative intensity for each attribute.
Individual data treatments are performed to assess quality of the data (repeatability and discriminant ability). Using multivariate analysis, the relationships between the products and their sensory description are illustrated on sensory maps. We find a consensus between the descriptions of the 10 subjects. The means of the scores are used for Partial Least Squares (PLS) regression. PLS allows to find the relation between 2 groups of variables, one group as predictors (matrix of the Plackett-Burman design), the other as dependent variables (matrix of attributes scores across steering configurations). PLS regression sorts out the coefficients of the model that explain the sensory scores according to the level of the EPAS parameters. This model shows the most influential parameters on the steering feel and how they influence the intensity of sensations. With such a model, we can predict steering feel according to EPAS parameters levels; we can easily tune EPAS according to customer steering feel preferences.