Document Sample
45 Powered By Docstoc

                                                           Suthep BUTDEE

    Leather fashion design is the one of the main project for                                         Introduction
Thailand. The objectives are to improve skills of people,
improve the Bangkok city to be fashion city and improve                    CaseXPert system is computer online software that
industry to be competitiveness. Fashion is a concept of a fast             assists designing of leather products in order to win the
change and short life cycle of product as well as various styles.          time for preparing of works, planning to production,
Fashion is combined between lifestyle and engineering in                   making a pattern, selecting good materials, calculate cost
order to meet human requirements and market constraints.                   and time. The main principle of the CaseXPert is a
Three criterions are commonly concerned. They are low cost                 knowledge expert system using case based reasoning.
but high price, fast production and high quality. Therefore,               This system
good knowledge and experiences are needed. However, human                  The goal of CaseXpert is to make a new design based on
experts in this field are rare and very expensive. The new                 the customized products for individual requirements and
effective concept that is used to replace the traditional method           meet products at low cost, best quality and delivery short
is intelligent system in order to capture industrial knowledge,            time. In this case, the configuration design and variant
formulate, manage and use the knowledge. CaseXPert system                  design are used as a primary method. A abundant design
is presented. Database of products and processes of leather                knowledge is needed together with huge cases as well as
good fashion design are created and stored with existing data              design experiences. Therefore, it is necessary to collect
in the library. The new product design is asked to get the                 and manage the design knowledge generated from the
requirements from the system. It is then matched and                       product customization design process and thus provide
retrieved the most similar previous products. The similarity               good mechanism of knowledge accumulation and reuse
percentages are shown. One of the shown products is selected               so as to support designers decision making. Chunli and
to modify according to the customer’s requirements which is                Yu [1] study on product knowledge management method
followed up with the verification. The satisfaction product is             supporting product agile customized design. The
indexed and stored in the library for the future used. The                 comparison between traditional design and product
system is successfully tested.                                             agility customized design is using more knowledge-
                                                                           intensive. Designers are not only exchanging geometric
Indexed Term: CaseXPert System, Case-Based Reasoning,                      data, but more knowledge about design and design
Industrial Knowledge Management                                            process, including specification, design rules, constraints,
                                                                           rational and so on which is needed for an integrated
                                                                           knowledge resource management environment to
                                                                           support the representation, capture, sharing and reuse of
                                                                           design knowledge among designers becomes more

 Suthep Butdee, Integrated Manufacturing System Research Center (IMSRC) Department of Production engineering, Faculty of Engineering King Mongkut’s
 Institute of Technology North Bangkok (KMITNB) 1518 Piboonsonkram Rd. Bangsue Bangkok, Thailand 10800 Tel: +662-5858541 ext.2520 Fax: +662-
 5870029 Email: stb@kmitnb.ac.th

  © 2006 GCMM                                                                   November 19 - 22, 2006, Santos, BRAZIL
                                          Global Congress on Manufacturing and Management
                                                            management with several tables and many fields. A table
Presently, one of the active analytical learning in cases   can be linked to another table. However, the frame
system is known as Case-Based Reasoning – CBR.              representation for case-based reasoning might be
Schank’s [2] worked in dynamic memory which used            different from the traditional database. The frame for
the philosophy based on an arrangement that human           case must show the event like story from the beginning
beings reason with the assistance in the form of cases      until the end. It is commonly called script with scene.
rather than rules.                                          Types of memory can be classified into two types; Static
CBR is essentially analytic reasoning from the earlier      and dynamic memory. The fixed knowledge is stored the
cases of experience which are used to solve current         in the static memory. It is consistent but it has not ability
problems, evaluate solutions and explain erroneous          to change the knowledge. It causes some problems and
solutions. There are two ways of implementing the           inefficiency. They can not modify and adjust the
case-based reasoning approach [3] One is the problem        previous knowledge. The knowledge does not link each
solving approach in which new situations are derived        others. For example, one type of knowledge might be
from the old cases. This type was applied to a lot of       adapted from the other knowledge. By contrast, the
applications in design, planning, diagnosis, etc.           adapted knowledge is dynamic knowledge. The basic
another is the interpretation approach in which the         knowledge is stored in library. Then, such knowledge is
new situation is evaluated rather than derived in           used, modified, adapted, and stored in the library. This
context of the old available cases. This type is            knowledge contains a specific condition, situation and
especially used in the cases of if any computational        event. MOP and TOP are the two types of memory
tools are available to evaluate the solution.               organization that have been applied in the expert system
CBR concept has been widely applied to design               applications.
domains such as mechanical design, manufacturing            Inference engine can be classified in several types. They
design, architectural design and aerospace design. It       are rule-based reasoning, case-based reasoning, model-
enables users to retrieve a previously known design         based reasoning and so on. The rule-based reasoning is
from the memory and adapt it to the current problem.        normally used logic decision making in the format of “if
Maher [4] developed case structural design which            – then – elf”. The rules are generally linked like chains.
contains four modules. CaseCAD is object-oriented           Therefore, they can be conflicted to each others. Another
cae representation. Each object includes CAD                problem is time consuming in order to get a result from
drawings and images, groups of attribute value pairs        the system due to checking the huge rules and conditions.
and text-based case description. CADsyn focuses on          On the other hand, the case-based reasoning is performed
case adaptation. Win concerns with a case memory            by the concept of parallel instead of serial chain like rule-
context to assist a structural engineering perspective.     base. The knowledge is stored in case like slots. It can be
Demex is used for flexible retrieval and memory             completed in one slot or linked to the other slot. Each
exploration. Perce [5] developed ARCHIE for                 slot is indexed by its feature. The slot can be derived
architectural design system.                                from the other slot by modification and adaptation. This
CBR was shown the applications in several domains of        is an advantage because knowledge can be checked by
design [6- 10]. The critical summaries are that the CBR     the relation to each other.
can use past experiences to solve current problems. It is
known as a combination method of reasoning and                         CaseXPert System Structure
learning. The CBR remembers previous problems and
either adapts their solutions or uses their outcome to      The system structure contains with three major parts.
evaluate new cases. Basically, the CBR procedure            The first one is input data for problem description. This
contains six steps; assigning indexes, matching and         is the part of user interface. The users can give their
retrieving old similar cases, modifying the retrieved       requirements. The second part is knowledge library
case to fit with the current case requirements, testing     which is linked to CAD product model library. The
and validating the modified cases, storing with             product models are classified by feature-based concept
assigning indexes to the successful cases, repairing and    and indexed them in groups. Each model is linked to
testing again if the cases are not satisfactory.            production process design. Third part is inference engine
                                                            which is used case-based reasoning. In this system it is
     Case-Based Reasoning and Memory                        called case-based reasoner.
               Organization                                 Products descriptions concerned with products, processes
Expert system is successfully applied to assist many        and people. Products contain functions, part in details,
industrial fields. They are several types of the expert     quality, materials, forms and styles, values and costs, life
systems depending on the types of knowledge                 cycle. Process consists of step of production, machines,
representation, types of memory and types of inference      tooling, pattern and prototypes. People involves with
engine.                                                     requirements, prices, sex, ages, professional.
Knowledge representation can be a production rule           Knowledge library consists of three main parts. They are
type, semantic net, frame-based type, model-based and       product descriptions, prototypes, processes and market.
etc. Case-based reasoning is mostly used with frame         As it has explained above, the product descriptions
based knowledge representation. Frame can be a type         contain not only product itself but also prototypes,
of database relations using the concept of database         processes and market. It is extremely significant that

© 2006 GCMM                                                            November 19 - 22, 2006, Santos, BRAZIL
                                 Global Congress on Manufacturing and Management
inside the knowledge library must have the stored cases    trend with the identity of their own products. Finally, the
which are used in the previous time. Such cases can be     new design concept can be created.
good cases or bad cases. For Example, the bad cases
consist of bad processes, bad prototypes and can not
sale in the market. These cases can be a warning for a
new designer.

Products -   prototypes --   processes -   market

Case-based reasoner contains four modules; the             Figure 4 The examples of the Fashion Trends Design for
matching and retrieving module, modification module,                         the New Seasons
testing module, indexing and storing module.
                                                           The conceptual design is created by the designer in order
             Knowledge Management                          to tell some story and value for the customers. It is
                                                           followed up with the prototypes both in digital and in the
                                                           physical forms. The figure 5 shows the prototypes of a
Knowledge is critical for design. Leather good design
                                                           new product in correction.
knowledge can be classified into several parts. They are
users or customers, product functionality, product
components both outside and inside, quality, machines
and tools, raw materials, pattern and prototypes, forms
and styles, production processes, packaging, image and
Raw material knowledge is one of the most significant            Figure 5 The example of standard bag types
knowledge. Types and quality of materials must be
clearly understood for the designer. It means that the     Figure 5 shows the standard types of bags which are
designer must know material properties, textures. The      stores in the knowledge library for comparing and
figure 1 and 2 show the example of material texture        indexing. In addition, the accessories which are used
types and styles.                                          together with bag are stored in the accessory library.

  Figure 1 Raw Material Colors with same Textures

 Figure 2 Raw Material different Colors and Textures

                                                                Figure 6 The CAD product data management
   Figure 3 Raw Material Textures of Animal Skills
                                                           Figure 6 shows the product data management. The
Products and fashion trend knowledge are also              standard bag types are arranged by attributes of
influenced and very impact to new products design.         dimensions such as height, width and thick. The CAD
Therefore, in the knowledge base library must be           program is selected one of the simple types of CAD,
dynamic in order to update the new fashion trend.          named SolidWorks. CaseXPert system is developed by
Normally, the trends are set by the fashion leader.        VS Basic. The system linked to the CAD system.
They are European, Asian, and American. The figure 4       Feature-based method is employed to assist the CAD
shows the new fashion trend of the leather good for the    product modeling in order to modify the similar existing
new seasons of 2007. From the trend in the knowledge       standard products inside the knowledge base library to fit
library, the designer can look and study the world         with the new customer requirements.
fashion trend. It is then trying to compare the fashion

 © 2006 GCMM                                                            November 19 - 22, 2006, Santos, BRAZIL
                                  Global Congress on Manufacturing and Management
               Case Based Reasoner
Case based reasoner is performed as the human expert
in order to make decision to solve problem from                         Product
customer requirements. The system decision making
consists of several steps. It starts with inputting details
of customer requirements. The CBR then compares the
inputted information with the existing knowledge in
the parts of product descriptions. The product                   Figure 7 The Flow Chart of the CBR – Case-based
attributes are matched. The nearest neighbor concept is                             Reasoner
employed to match between the present and previous
data. The similar existing bags are retrieved. The                                  Case Study
similarity percentages of matching are shown. The user        The CaseXPert system development is tested by the
is then selected the most preferable. The picture images      customer requirements. The first case is to find the lady
of all retrieved bags are also shown. The user can then       bag which is made of genuine leather for classic
modify followed the design concept. New parts of              collection. This bag is proposed to use for every day of
some accessories are selected from the library. The           work. The colors are in the group of black or brown or
processes of modification are performed under the             gray or white. The size is 7 x 16 x 28 cms of the width,
CAD system. The modified bag is checked by the user.          height and length respectively. The information is
It is then confirmed or modified again. The approved          matched. The first stage of result is shown as the figure 8.
bag is then matched with existing pattern in library as       All of the retrieved bags are checked with the
well as product processes. It can now be made for a           requirements. They are then selected by the most similar
physical Mok-Up following the product processes and           one. Therefore, the bag type Hobo, the id. HO-004, the
pattern. The minor change from technical production           brand named is Matostto with the size of 14 x 14 x 28
can be modified the design in order to meet low cost,         cms is selected as shown in Figure 9. The selected bag is
good quality and short processing time. The successful        then modified to meet the customer concept. Another
prototype bag is stored in the library which is classified    case study is shown in the Figure 10. It is the flap style
by groups and clusters. However, the system can not           with the size of 8 x 14 x 28 cms. The system is matched
match every thing for the customer requirements. In           and retrieved the most similar bag from the library. The
this case, the expert must give information to the            type id is FP003.
system. Figure 7 shows the flow chart of the case-based
reasoner making decision.            The new product
description requirements are given by users. The data
are matched with existing bag styles in the knowledge
library. The most similar cases are retrieved and shown.
The users now are able to select and modify according
to conceptual product design. The modification is then
test more in product styles and forms. If the design is
satisfactory, it will be indexed and stored in the
knowledge library.

           Cases Query

                                                                Figure 8 The Product Description Case Query
            CBR                      Knowledge
           Reasoner                   Library

             Existing                  Indexing
           Similar Bag                 & Storing

 © 2006 GCMM                                                            November 19 - 22, 2006, Santos, BRAZIL
                                  Global Congress on Manufacturing and Management
                                                             [1] Chunil Yang and Ming Yu (2006) A Study on
                                                                  Production Knowledge management Method
        Figure 9 The Similarity Matching                          Supporting Product Agile Customized Design,
                                                                  IDMME 2006, Grenoble, France, May 17-19.
                                                             [2] Schank, R.C. (1982) Dynamic Memory: A Theory of
                                                                  Learning in Computers and People, Cambridge
                                                                  University Press, Cambridge.
                                                             [3] Kolodner (1992) An Introduction to Case-Based
                                                                  Reasoning , Artificial Intelligent Review, Vol.6,
                                                             [4] Maher, M.L. and Garza, A.G.S. (1996) Developing
                                                                  Case-Based Reasoning for structural Design, IEEE
                                                                  expert, June, pp.42-52.
                                                             [5] Pearce, M., Goel, A.K., Kolodner, J.L., Zimring, C.,
                                                                  Sentosa, L. and Billington, R. (1992) Case-Based
                                                                  Design Support: A Case Study in Architectural
                                                                  Design, IEEE Expert, October, pp.14-19.
   Figure 10 The Result of Retrieval Bags From the           [6] Gowri, K., Illiescu, S., and Fazio, P. (1992) Case-
                 CaseXPert Library                                based reasoning in Building Envelop Design, in
               Concluding Remarks                                 Grierson, D.E., Rzevski, G. and Adey, R.A. (eds),
CaseXPert system for leather good fashion design has              Applications of Artificial Intelligence in Engineering
been presented. Case-Based Reasoning concept is                   VII, Computational Mechanics Publications, Boston,
reviewed and applied to this system. In CBR, the case             pp.1221-1238.
memory organization is important. MOPs and TOPs              [7] Faltings, B. (1991) Case-Based Representation of
have been explained. The system structure and                     Architectural    Design      Knowledge,      Artificial
knowledge management are also discussed. Finally,                 Intelligence in Design, NL.
the case-based reasoner is illustrated as well as the case   [8] Domeshek, E.A., Hherndon, M.F., Bennett, A.W.
study. Two cases are shown together with the                      and Kolodner, J.L. (1994) A Case-Based Design
similarity matching cases. The most similar is simply             Aided for Conceptual Design of Aircraft
selected following with modification in order to meet             Subsystem’s, IEEE, pp. 63-69.
the customer satisfactory. This system can assist young      [9] Rong, R.W.and Lowther, D.A. (1994) Storage and
designer very much. It can save a lot of time. In                 Retrieval of Solutions in the Design Aid for
addition, the system can linked to CAD for changing               Conceptual Design of Aircraft Subsystem, IEEE, pp.
the product model.                                                63-69.
                                                             [10] Navinchandra, D., Sycara, K.P. and Narasimhan, S.
                                                                  (1991) Behavioral Synthesis in CADET, a Case-
                                                                  Based Design Tool, IEEE, pp 2063-2068.

  © 2006 GCMM                                                            November 19 - 22, 2006, Santos, BRAZIL
                                   Global Congress on Manufacturing and Management