Data-Mining Driven Reconfigurable Product Family Design Framework for Aerodynamic Particle Separators

Abstract

This paper presents a design framework that integrates data driven knowledge discovery with engineering design simulation to create an optimal product family of aerodynamic air particle separators. Two key data mining techniques are presented in this work to help guide the product design process. First is the RELIEFF attribute weighting criterion that identifies and ranks product attributes based on their importance within the raw data set. The second is the X-Means clustering approach that eliminates the need for the number of clusters to be stated a priori. The engineering product family optimization stage that follows will therefore be a true representation of the raw data set by attaining an optimal design based on the number of clusters generated by the X-Means clustering process while concurrently taking into account attribute weighting information. The resulting product portfolio will achieve cost savings and autonomous reconfiguration of product architecture under the notion of reconfigurable product family. A family of prototype aerodynamic air particle separators will be used to evaluate the final solution of the reconfigurable product family model generation.