Chapter 3 Customer-based Design for Mass Customization (CDFMC) Methodology 3.1 Introduction Product design has been considered as the deciding factor to achieve the goal of mass customization. The prerequisite of a successful product design is to have a right understanding of individual customer requirements. Traditionally, customer requirements are first collected and then analyzed by design engineers to develop products to meet these requirements. However, organizations have recognized that they cannot rely on designers, developers or specialists to know how to design products or services to meet the customer requirements since designers are seldom the primary users of the products and their own biases and rationalizations often interfere with assessing what customers truly need or want. Hippel (1998) pointed out that users, rather than suppliers, are the actual designers of the application-specific portion of a product. Thus, customer-centered design has become a business strategy used by many companies to gain competitive advantage and maintain economic viability (Smart and Whiting, 2001). Currently, an increasing amount of research and industrial applications discussed in the last chapter has appeared on customer-centered product development. The implementation of customer-based design strategy has involved all the factors of 35
product life-cycle development, including product supply chain (e.g., customers, sellers, manufacturers, distributors), design knowledge (e.g., functional specifications, product structure, product family design, design solution optimization), computeraided tools (e.g., knowledge-based systems, web-based systems), etc. However, there is a lack of a systematic theory that has comprehensively considered these issues for the implementation of the strategy. Therefore, the key factors and kernel technologies should be identified and a formal approach is needed to implement the strategy. 3.2 Customer-based Design Strategy To implement the customer-based design strategy, the key factors or roles have to be identified and analyzed to determine their requirements. In a product life cycle, customers and manufacturers play the two most important roles among all the factors. Thus, it is essential to identify their needs and actions in order to successfully implement the strategy. From the manufacturers viewpoint, they need to rationalize the product line, and design and develop product families for different market segments. Modularity in product family design has offered manufacturers greater flexibility for rapid and economical development of mass-customized products. On the other hand, manufacturers have to provide customers with a platform or environment to let them become actual designers of the application-specific portion of a product in order to implement the strategy. Using this platform, namely, a knowledge-based mass customization system, customers can easily retrieve valuable information, clearly express their needs and conveniently obtain their personalized products. However, these platforms involve complex knowledge and constraints of modular product 36
families in the application domain. Thus, an effective product family modeling approach is needed to support the design synthesis and product customization processes. In addition, advanced IT and AI technologies are also needed to develop these tools or systems. From the customers viewpoint, there are different market segments, such as low-end, middle-end or high-end markets, with different product requirements. Customers also need more information about the products, e.g., market-related information, besides the product architecture information and functional specifications. To easily and economically obtain the related information of products or services, customers should have effective tools to help them. Generally, these tools should be provided by the manufacturers or suppliers, and can easily be used by the common customers. When purchasing a product, namely implementing product customization, different customers may need different levels of support from the tools in terms of their familiarity with the product, and their buying habits. Accordingly, there are three methods to implement the customization process: 1. Customers provide all the requirements, e.g., functional specifications, and customized products can be obtained. This requires the biggest help from the tools or systems, such as a user-recommending system. 2. Customers provide some requirements, e.g., functional specifications, and simultaneously specify certain architecture features, e.g., components or assemblies, and the customized products can be obtained with partial support from the tools or systems, such as configuration systems. 3. Customers specify all the structural features, e.g., assemblies and components that meet all the constraints. This is a design-it-yourself (DIY) method. This 37
requires the least support from the tools or systems, which may only provide functions like catalog-based selection and order. According to the amount of user influence in the customization process, product customization can be divided into automated customization, semi-automated customization, and interaction-based customization. The automated customization corresponds to the first method, where computer-aided configuration tools or knowledge-based systems can create a solution automatically in terms of customer needs with minimum user influence. The interaction-based customization corresponds to the third method, where customers have the largest influence on the tools or systems, using a manual method to obtain their products. The semi-automated customization is between the two methods. In order to satisfy different customer requirements, a comprehensive product customization process is proposed and implemented as a computer-aided system or tool to realize the three methods. In addition, customers often have difficulties finding their desired products due to conflictive needs or incomplete requirement information in product customization. For the former case, these tools should give customers rational explanation and good advice on what to do next. For the latter case, the customers may obtain many feasible solutions because of the incomplete requirement information. In this situation, these tools should help customers select an optimal solution among these feasible solutions. Therefore, a successful computer-aided system or tool should solve these issues. According to the axiomatic design theory (Suh, 1990), product design is defined as a process mapping from a functional domain to a physical domain, in which a set of 38
design specifications, referred to as product structure features, is found to meet the functional requirements and related constraints. In the first production customization strategy, the customer requirements involve only the functional domain, while the customer needs are related to both the functional and the physical domains in the second product customization strategy. In the third product customization strategy, customer requirements are only associated with the physical domain. According to the classification of product variety design proposed by Fujita and Ishii (1997), namely, the system-level, configuration-level and parameter-level varieties, the first strategy can be mapped onto the system-level variety design problem, while the second and third strategies may be mapped onto the configuration-level and parameter-level variety design problems respectively. A successful computer-aided system or tool needs to support these underlying design principles. The issues and challenges mentioned above need to be solved through a systematic approach to successfully implement the customer-based design strategy. In the next section, the framework for customer-based design for mass customization, namely CDFMC, will be proposed. 3.3 CDFMC Framework The CDFMC framework was constructed based on the approaches for solving the issues of the representation schemes of the domain knowledge of a modular PFA, problem-solving technique for design synthesis, and the optimization technique for design solution optimization. In the following section, these issues will be discussed. 39
3.3.1 Knowledge Acquisition and Representation Knowledge acquisition and representation are among the most critical factors for the successful development of knowledge-based systems or tools. This is especially true for systems that can help customers implement product customization since product customization involves the knowledge in product life-cycle development, namely, the domain knowledge of a modular PFA. Generally, the sources of knowledge are distributed among different people within an organization. Thus, knowledge acquisition has long been viewed as a bottleneck in the design and implementation process (Walczak, 1998). In this research, the configuration stage of product family design is the focus, where a modular PFA is built based on other researchers work (e.g., Jiao and Tseng, 1999a; Gu and Sosale, 1999; Stone et al., 2000; Dahmus et al., 2001). With a modular PFA, knowledge acquisition in the application domains has been completed at this stage. Thus, the main task is to effectively organize and represent the configuration knowledge of the product families. As described in Section 2.1.1, a modular PFA has been widely adopted in product family design. There are two characteristics of a modular PFA: hierarchy and classification. A product family is a modular system that is composed of different kinds of modules. These modules can be further decomposed and organized into a hierarchical architecture. This decomposition can be based on the different viewpoints such as customer functional needs and product physical architecture. To effectively support design synthesis and product customization process, there are many ways to organize the configuration knowledge. Soininen et al. (1998) proposed a general ontology of configuration knowledge that is independent of the problem 40
solving methods, while Jiao and Tseng (1999b) presented another classification of configuration knowledge in which a product family model consists of three classes: Entity, View and Constraint, and these classes are further divided into sub-classes. Different classification methods of the configuration knowledge form different relationships between the classes. Sabin and Weigel (1998) pointed out that a minimal list of relations among the classes of an effective representation scheme in computeraided configuration tools have to include: classification relationships (e.g., is-a), aggregation relationships (e.g., part-of), local constraints (e.g., structural, arithmetic, cardinality), global constraints (e.g., resource constraints, optimization criteria), etc. Based on the review of product family representation in Section 2.1.3, the objectoriented approach includes several distinct characteristics that enable a natural decomposition and hierarchical structuring of complex configuration knowledge, represent complex constraints between classes or objects, and support the dynamically evolving nature of domain knowledge (Gorti et al., 1998). Furthermore, objectoriented approaches are widely combined with other representation schemes (e.g., CSP-based framework, case-based framework, other knowledge-based scheme) to increase the flexibility of knowledge representation and modeling and improve the capability of design synthesis. Hence, the object-oriented approach can be used to model the configuration knowledge of a modular PFA. The critical issue is the use of object-oriented techniques to define a meta model to effectively represent a modular PFA and support design synthesis in product customization. In this research, a meta model has been defined for a modular PFA and a frame-based representation scheme has been used to describe the meta model. The detailed description of the representation scheme is given in Chapter 4. 41
Although the object-oriented approach can be used to represent the hierarchical configuration knowledge of product families from multiple viewpoints such as the functional and structural viewpoints, a method to structure the configuration knowledge is needed. As a structuring method that has been proven in practice, quality function deployment (QFD) can translate the voice of customer (VoC) into product design or engineering characteristics, and subsequently into parts characteristics, process plans and production requirements (Hauser and Clausings, 1988). The basic QFD tool is a product-planning matrix, which is frequently referred to as the house of quality (HoQ). Generally, there are four sets of matrices, in which the first two matrices focus on the determination of design characteristics for the entire product and its features. Rao et al. (1999) stated that QFD is well compatible with knowledgebased systems since this matrix fits nicely with the equivalent premise/conclusion matrix (rule sets) frequently used in knowledge-based systems. The QFD techniques are used in the knowledge base development to convert customer requirements into design and manufacturing specifications (Ngai and Chow, 1999; Rezayat, 2000a). Thus, combining with the object-oriented model approach, QFD has been used as a structured method to map the customer needs onto the product technical specifications in this research. Schmidt (1997) identified the shortcomings of traditional QFD applications, one of which is that the traditional QFD neglects the existence of individual part-related customer needs. However, these customer needs are important to realize the customerbased design strategy described in Section 3.2. Thus, these customer needs must be considered in modeling the configuration knowledge of a modular PFA. 42
3.3.2 Problem-solving Technique Problem-solving techniques are important for effectively addressing the design tasks. As reviewed in Section 2.1.4, many reasoning or searching techniques have been proposed to generate or select design solutions from the knowledge bases that contain the domain knowledge described in the corresponding representation schemes. However, these techniques are not able to effectively support the product customization process due to lack of complete and uniform representation schemes for the configuration knowledge of product families. In this research, an object-oriented product family modeling approach has been proposed as the knowledge representation scheme. The corresponding problem-solving technique, namely, a constraint-based framework, has also been formulated to handle the product customization process, since the CSP-based techniques are effective problem-solving techniques and include many domain-independent approaches to reduce the search complexity. The combination of the object-oriented representation and the CSP-based techniques can fully exploit the advantages of approaches, namely, the knowledge expressiveness and the search efficiency. As discussed in Section 3.2, the issues that need to be addressed in product customization include design solution generation, design solution optimization, and conflictive needs handling. The corresponding algorithms have been formulated to handle these issues respectively. The algorithms constitute the constraint-based framework. A detailed description of the constraint-based framework is given in Chapter 5. 43
3.3.3 Optimization Techniques In the configuration design stage, optimization techniques have been used to derive an optimal configuration of the modules for a product variant from all the feasible module configuration spaces according to conflictive goals and criteria. Optimization techniques are used to formulate the optimization models that are abstracted from engineering design problems. As reviewed in Section 2.1.5, many optimization models have been built to solve the product configuration problem from different viewpoints. In product customization, customers often encounter situations where many products of a product family can satisfy their requirements. Thus, an optimal product should be selected among these feasible products for the customers. This issue should be tackled to realize the customer-based design strategy described in Section 3.2. In this research, optimization models have been developed based on a modular PFA to quantitatively handle customers preferences on product features such as product cost, quality and performance. According to the different preferences, there are two optimization models, namely, the single-objective and multi-objective models. In the constraint-based framework, the preferences are treated as software constraints and the optimization models are transformed into optimization evaluation functions to select an optimal solution from the feasible candidates. These will be described in Chapter 5. 3.4 CDFMC Implementation 3.4.1 System Implementation Technology 44
To implement the customer-based design strategy, the development of knowledgebased mass customization systems or tools has to utilize the latest IT and AI technologies. The advancement of IT, especially in the web technology, makes it possible for knowledge-based systems to realize this strategy. In fact, the system should be an enterprise-web portal for life-cycle support, and acts as a distributed design and manufacturing environment that enables integrated product, processes, and protocols development (Rezayat, 2000b). The system should be able to handle the issues such as enterprise-wide access or sharing of information, updating of information, and management of information and resources. Rezayat (2000b) has identified the main components of these systems, which include a client/server system architecture, the web as its backbone and technologies such as browsers, HyperText Markup Language (HTML), Transmission Control Protocol/Internet Protocol (TCP/IP) and so on. These technologies have been used in this research to develop the system prototype. On the other hand, developments in the AI technology allow knowledgebased systems to effectively handle complex knowledge and help customers obtain a customized product timely and easily. The knowledge representation and reasoning algorithms mentioned in Section 3.3 have been applied in the prototype system. 3.4.2 System Development Method The system prototype development method is important since it influences the system migration and extensibility, reusability of system components, testing, etc. Recently, the component-based development (CBD) approach has become an important and widely adopted approach in the software industry, as it allows large-scale codes to be used or reused across multiple platforms, in various languages and at different locations across networks (Rosenman and Wang, 2001). The reuse of existing software 45
components allows system development to achieve the efficiency of time, quality and cost. The CBD approach also makes the system open and flexible. With these characteristics, the CBD approach can meet the needs of complex large-scale software systems, and has been adopted in the various business domains such as manufacturing, telecommunication, banking, etc. In the design field, research and implementation are currently being conducted. For example, Rosenman and Wang (2002) developed a component agent-based open CAD system for construction design. Pahng et al. (1998) created a distributed modeling and evaluation framework based on Common Object Request Broker Architecture (CORBA). This research uses the CBD approach to develop the knowledge-based mass customization system for product customization. In the CBD approach, a software component is defined as a part of an application that has been developed and tested independently, and later integrated into the application through simple communication. A component can perform a particular function and interact with other components in the system through a standard interface. The standard interface enables encapsulation so that the component may hide its inner details and provide public functions or services for the external world. Through standard interfaces, a component allows any application in any language on any platform to access its features and use its services. A consistent and standard interface enables a component to work well with others, and eases the integration of the components into a system. Based on the type of reusability, a component can be classified into two types. One type is reusable components that are used as servers. The other is application components that will use the services that are offered by the servers. Correspondingly, the CBD approach has two aspects: developing the servers, 46
i.e., the server side development, and developing applications, i.e., the client side development. To support component development, a number of standards and development tools are available today. Among such tools, the Microsoft s (distributed) Component Object Model (COM or DCOM), the Object Management Group s CORBA, and the Sunsoft s Enterprise Java Beans (EJB) are standards for the building of software components that can inter-operate in a reliable way. These standard interfaces allow CBD to be a more reasonable strategy in tackling the increasingly complex task of software development. Based on the system development platform (e.g., Win32 systems) and the development tool (e.g., Visual C++), the Microsoft s component object model has been used in this research. 3.5 Summary In this chapter, the key factors in the customer-based design strategy have been identified and their requirements analyzed. Based on the analysis, a framework for CDFMC methodology has been proposed for addressing issues such as knowledge acquisition and representation, the problem-solving technique and the optimization technique. Finally, the implementation technologies and the development method of the system prototype were also discussed. 47