Rule-based inference model for the Kansei Engineering system

Kansei Engineering has been applied to product development for customer sat isfaction based on ergonomic technology. The system is composed of three pa rts such as Kansei analysis, inference mechanism, and presentation technolo gies. The inference mechanism by which human Kansei is translated into desi gn elements plays an important role in Kansei Engineering. The reasoning lo gic in the system must satisfy several conditions. First, the whole aspects of design elements must be considered in the reasoning processes. Second, the reasoning logic can make the discrimination between Kansei words select ed by customers and other words. Third, the reasoned results must have reli ability high enough to reduce the difference between customer's image and r easoned design elements. In this paper, we propose a rule-based inference model which will cover the above-mentioned conditions. The rule-based inference model is composed of five rules and two inference approaches. Each of these rules reasons the de sign elements for selected Kansei words with the decision variables from re gression analysis in terms of forward inference. These results are evaluate d by means of backward inference. By comparing the evaluation results, the inference model decides on product design elements which are closer to the customer's feeling and emotion. Finally, simulation results are tested stat istically in order to ascertain the validity of the model.