Data Processing

Application of Data Driven Techniques to the Prediction of Seat Comfort Performance

Previous studies report and analyse the relationship between the subjective ratings of the perceived seat comfort and the Body – Seat Interface Pressure Distribution (IPD) or other objective measurements (postural). In this work we explored the application of advanced and powerful predictive modeling techniques to the problem cited above.

Two of different data driven inference techniques were considered and enhanced for this specific analysis and modeling task: Artificial Neural Network (ANN) and the Parzen Method. Based on the potential of these predictive models, the concept of their role in the product development process is described.

Compared to the traditional multivariate regression analysis, the ANN and, especially, the Parzen model have the following important inherent advantages:

  • Modeling capability for nonlinear relationships
  • Estimation of the reliability of the prediction
  • Detection of multi-modal distributions

A preliminary pilot experiment was designed and carried out to provide the data required to establish and test a simplified version of the model. We point out that the study here presented is dedicated to analyse biomechanical comfort.

Although is the most important, the biomechanical comfort is only one of the components of the overall comfort performance of the seat. A method for the specification of the model on an available data set has been developed. Also a method to test the predictivity effectiveness and accuracy of the model has been defined.

The preliminary results are encouraging and allowed us to collect important indications for further developments. The difficulties in operating with subjective quantitative data have been highlighted.