Product Variety Optimization Based On Clustering Analysis And Genetic Algorithm

Mass customization (MC) is a new production strategy which aims to provide customized products with near mass production efficiency and cost. The key is to solve the conflict between scale economies with batch production while meeting customers'individual requirements. Thus effective variety decision in the product design and production management process is critical. Reasonable variety can provide customers with specific products and service and at the same time increase the commonality among them. However, effective methods and tools are lack to support scientific product variety decision at present. The dissertation is to research and develop suitable methods and algorithms to solve key problems of product variety decision area, including optimization of both exterior product functions and interior design diversification and related common knowledge identification, to provide theory and method support for the implementation of MC strategy.Function variety decision of products is needed to improve customer satisfaction and customer requirements (CRs) analysis is imperative. In fact, this is also a chief step to realize MC. Research of this dissertation focuses on analyzing and mining historical transaction records to discover customers'eal preference and level. Customers and products are to be grouped from customer requirements and product features, respectively, and based on this to analyze and predict trends of dynamic CRs and functional requirements (FRs), aiming to facilitate product definition and function variety decision. Then FRs patterns are to be captured by analyzing existing product data, with the aim to help customization using information stored in existing products and consequently to deal with orders more effectively. A novel method integrating clustering analysis and information entropy is proposed to recognize FRs patterns with higher commonality by historical FRs data analysis. Steps of the method are demonstrated by a case study and its feasibility and validity are proved through the comparison of results derived by other published methods. Product design variety optimization, especially quantitative product platform planning and product family optimization, is another research focus of this dissertation. Many existing product family design methods assume a given platform configuration, i.e., the platform variables are specified a priori by designers. However, selecting the right combination of common and scaling variables is not trivial. Most approaches are single-platform methods, in which design variables are either shared across all product variants or not at all. While in multiple-platform design, platform variables can have special value with regard to a subset of product variants within the product family, offering opportunities for superior overall design. In this dissertation, a new quantitative platform planning method is presented combining clustering analysis, information theoretics, and validity analysis of fuzzy theory. The objective and obstacle of product family design lie in the optimal tradeoff of individual product performance and commonality among them. A successful product family design method should achieve an optimal tradeoff among a set of conflicting objectives, which involves maximizing commonality across the family of products without comporising the capability to satisfy customers'performance requirements. Multiple-objective optimization problems and suitable approaches for scaled-based multiple-platform product family design are to be studied in the dissertation, especially a single stage optimization approach. The single-stage approach can yield improvements in the overall performance of the product family compared with two-stage approaches, in which the first stage involves determining the best settings for the platform variables and values of unique variables are found for each product in the second stage. A case study of designing a set of universal electric motors is applied both to illustrate the methodology and verify its performance.To overcome disadvantages of traditional algorithms, computation algorithms are studied and developed to solve relevant topics of product variety decision under MC environment. An improved clustering algorithm integrating rough set model and entropy is proposed. It aims at avoiding the need to pre-specify number of clusters, and clustering in both numerical and nominal attribute space with the similarity introduced to replace the distance index. At the same time, the RS theory endows the algorithm with the function to deal with vagueness and uncertainty in data analysis. Shannon's entropy was used to refine the clustering results by assigning relative weights to the set of attributes according to the mutual entropy values. It's applied to discover commonality knowledge for product function variety optimization and platform planning. This dissertation also presents and develops a two-level chromosome structured genetic algorithm to simultaneously determine the optimal settings for the product platform and corresponding family of products, by automatically varying the amount of platform commonality within the product family during a single optimization process. The augmented scope of 2LCGA allows multiple platforms to be considered during product family optimization, offering opportunities for superior overall design by more efficacious tradeoffs between commonality and performance. Validity and effectiveness of both algorithms are verified through case studies and results comparisons against previous work.