Development of Statistical Models for Predicting Automobile Seat Fit of Drivers

Abstract

The present study is intended to develop statistical models for predicting automobile seat fit based on the relationships between seat dimensions and subjective seat fit. The evaluations of the subjective seat fit for 43 different driver seats and the seat dimensions at six cross-sectional planes (three for the seatback and the other three for the cushion) were measured and evaluated by eight seat-engineers. The best subset logistic regression analyses were conducted to quantify the relationships between the measured seat dimensions and evaluated subjective seat fit at each of the cross-sectional planes. As a result, significant seat dimensions, such as insert width or bolster height, on the subjective seat fit were identified. The developed logistic models show 90% overall classification accuracy at each section with 80% accuracy with five-fold cross-validation. The developed models would be particularly useful to support seat engineers by providing recommended seat dimensions, which could increase seat fit. In addition, the model is useful to reduce development costs for an automobile seat and increase work efficiency in the digital evaluation process of an automobile seat.