A value driven decision support framework for the front-end product design and development

• Main Author: Withanage, Chathura
• Other Authors: Park Taezoon
• Format: Theses and Dissertations
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Systems engineering; DRNTU::Engineering::Manufacturing::Product design; DRNTU::Business::Marketing::Consumer behavior; DRNTU::Engineering::Industrial engineering::Engineering management
• Online Access: https://hdl.handle.net/10356/54748
• Institution: Nanyang Technological University

Value-attribute models are used at the front-end phase of the customer driven product design to support the design decision making by eliciting customer preferences. Currently used models are not capable of handling the high dimensional technological product attribute data, plagued by the multicollinearity, with limited observations. Furthermore, they are not equipped to deal with the heteroscedasticity and the dynamic nature of the technological product market systems. As a result, especially, the technological products with longer design cycle times are heavily affected by model uncertainties coming in various faces. Robust, dynamic, value-attribute models are needed to accurately replicate the market systems, to overcome the deficiencies of the data, and ultimately, to predict the future product values for the front-end concept screening. Therefore, a strategic decision support framework is proposed in this thesis to integrate the hitherto overlooked time variant properties of preferences, into the front-end design decision making process. In the proposed framework, the Partial Least Squares Regression and Path Modelling techniques, robust soft modeling methods, are used as the main decision support tools. And, Customer Revealed Value, a perceived value estimation obtained from a demand-price analysis, is used as the design objective. There are four main contributions in this thesis. Firstly, a theoretical basis is provided for the multivariate modeling of the value-attribute relationship. Secondly, a robust Partial Least Squares algorithm is introduced to handle the heteroscedasticity presence in the market systems. Thirdly, a dynamic value-attribute model is formulated by combining Partial Least Squares and Time Series Forecasting techniques. Finally, a dynamic value-characteristic model is formulated by extending the earlier model by including higher or system level product characteristics, using Partial Least Squares Path Modeling. All the contributions are validated using the US automobile market data. And, the results of the case studies depict the potential of the framework as a design decision support method.