Please use this identifier to cite or link to this item:
|Title:||An intelligent system for supporting affective product design|
|Authors:||Fung, Kai Yin|
Design -- Human factors.
Hong Kong Polytechnic University -- Dissertations
|Publisher:||The Hong Kong Polytechnic University|
|Abstract:||Affective design is currently an important aspect of new product development, especially for consumer products, to achieve a competitive edge in a marketplace. Affective design can help companies develop new products that can better satisfy the emotional needs of customers. In this research, a novel methodology for supporting the affective design of products is proposed. Through this methodology, an intelligent system is subsequently developed. The methodology primarily involves four processes, namely, survey conducting, rule mining, affective relationship modelling, and design optimisation. First, customer affections on product design have to be collected by conducting a marketing survey. The obtained survey data are used to model the affective relationships between customer affections and the design attributes of the products. An intelligent system for supporting affective design based on the proposed methodology can be developed to perform rule mining, affective relationship modelling, and design optimisation. The system involves three models, namely, a multi-objective genetic algorithm (MOGA) based rule-mining model, a dynamic neuro-fuzzy (NF) model, and a design optimisation model. A novel two-stage MOGA approach is proposed to perform rule mining for affective design. To consider the ambiguity of affective data, approximate rules are generated based on MOGA considering the three criteria of rule mining, including accuracy, comprehensibility, and extent of approximations. A two-stage rule-mining method is introduced to generate individual rules and subsequently refine the entire rule set considering rule interactions.|
A new dynamic NF approach is proposed for affective relationship modelling based on the survey data using a modified dynamic evolving neuro-fuzzy inference system (DENFIS). DENFIS is suitable for dealing with the ambiguity of affective relationships. Moreover, DENFIS can handle a large number of input attributes and data sets, whereas conventional adaptive neuro-fuzzy inference system (ANFIS) models cannot. In the dynamic NF approach, local models are established based on the evolving clustering method. Thus, the structures of DENFIS models are simple. A recursive least square is applied on DENFIS, enabling the dynamic NF models to be updated easily once new data sets are available. For the design optimisation model, a novel guided search genetic algorithm (GA) approach is introduced to determine optimal design attribute settings of affective design. With the use of the developed NF models and mined rules, a search strategy based on the guided search GA is formulated, resulting in the determination of better solutions of affective design and decrease in the searching time of design optimisation. A case study of the affective design of mobile phones was conducted to illustrate the methodology and the development of the intelligent system and to validate their effectiveness. A number of important results were obtained from the validation tests. First, the MOGA-based rule-mining approach performs better than the dominance-based rough set-based rule-mining approach in terms of accuracy and reliability in mining approximate rules. Second, the dynamic NF model outperforms conventional ANFIS models in modelling affective relationships. Third, the guided search optimisation model is capable of improving GA to generate better solutions for affective design.
|Description:||xviii, 142 leaves : ill. ; 30 cm.|
PolyU Library Call No.: [THS] LG51 .H577M ISE 2012 Fung
|Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|b26158383_link.htm||For PolyU Users||203 B||HTML||View/Open|
|b26158383_ir.pdf||For All Users (Non-printable)||3.45 MB||Adobe PDF||View/Open|
Checked on Mar 19, 2017
Checked on Mar 19, 2017
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.