叶轮五轴加工中心结构设计(英文翻译)

CBIMS:Case-based impeller machining strategy support system

Min-Ho Cho a,Dong-Won Kim a,?,Chan-Gie Lee a,Eun-Young Heo a,Jae-Won Ha a,F.Frank Chen b

a Department of Industrial&Information Systems Engineering,Chonbuk National University,Jeonju,South Korea

b Department of Mechanical Engineering&Center for Advanced Manufacturing&Lean Systems,University of Texas at San Antonio,TX78249,USA

a r t i c l e i n f o

Article history:

Received16December2008

Received in revised form

9March2009

Accepted10April2009

Keywords:

Impeller

Machining strategy

Case-based impeller machining strategy

support system(CBIMS)

a b s t r a c t

Impeller machining strategies cannot be easily formalized due to the complex,overlapping and twisted

shapes that form impeller blades.Skilful machining process planners may generate appropriate

machining strategies based on their experiences and previous machining data.However,in practice,

most shop?oor data for impeller machining is not well-structured and it cannot be used effectively by

process planners to produce the required machining strategies and process plans.This paper presents

the development of a case-based impeller machining strategy support system(CBIMS)that employs

case-based reasoning(CBR)to obtain an ef?cient machining strategy for an impeller by using the

existing machining strategies from the shop?oor.The CBIMS generates impeller machining strategies

through a stepwise reasoning process considering the similarities of the blade shapes and machining

regions between existing impellers and a new one.The system can provide a process planner with

machining strategies such as tool speci?cations,machining area partitioning,and the machining

parameters including feed rate,depth of cut,RPM and machining tolerance.A case study is provided to

demonstrate that the CBIMS can generate useful machining strategies while ensuring that it can be

effectively used to support the process planner.

&2009Elsevier Ltd.All rights reserved.

1.Introduction

An impeller is a type of high-speed rotor that is used to

compress or transfer?uid under high speed,pressure,and

temperatures.The impeller is composed of an axial hub and

several blades attached along the hub.The weight and shape of an

impeller must be balanced;otherwise the imbalances can cause

noise and vibration,which can lead to the breakage of the

impeller blades during operation.Thus,the hub and blades of an

impeller are commonly machined in5-axis NC machines to obtain

the required surfaces.But the impeller machining strategy or

process plan cannot be easily formalized due to the complex,

overlapping and twisted shapes of the impeller blades.Thus,

according to machining strategies,there can be various types of

tool paths that can be taken in the5-axis numerically controlled

(NC)machining of an impeller.

Impeller machining strategies can include tool speci?cations,

machining area partitioning,and machining parameters.The

tool speci?cations indicate tool con?gurations and tool diameters.

The machining area partitioning is concerned with machining

volume partitioning methods for ef?cient rough-cut planning in

terms of reducing machining time.Finally,the machining

parameters are involved in NC machining conditions such as feed

rate,depth of cut,RPM,and machining tolerance.These machin-

ing strategies signi?cantly in?uence the quality of the machined

parts,as well as the machining time taken to produce the

impeller.The5-axis NC machining of an impeller,typically,is

performed by using ball and tapered end mills.Thus,it is dif?cult

to determine the tool paths to avoid collision between the tool

and the impeller blades,compared to3-or4-axis machining of

other mechanical parts.Among many factors that in?uence the

NC machining,the feed rate of the tool is the most important

factor that affects the machining productivity,as well as the

machinability of a part.

The impeller machining industry belongs to the so-called3D

(dirty,dif?cult,and dangerous)industrial sector.It is dif?cult to

?nd experts who can successfully build the required machining

process plans and most shop?oor data for the machining of

impellers is not well-structured.Hence,it is dif?cult to effectively

provide a process planner with good data so that effective

machining strategies and process plans can be produced.This

study presents a case-based impeller machining strategy support

system(CBIMS)that can provide unskilled process planners with

standardized,effective machining strategies for process planning

by employing a case-based reasoning(CBR)process.The method

uses a database and/or a knowledge base that systematically

stores and maintains up to date shop?oor data such as planning,

machining,and product data.

Contents lists available at ScienceDirect

journal homepage:https://www.sodocs.net/doc/0b1537273.html,/locate/rcim

Robotics and Computer-Integrated Manufacturing

0736-5845/$-see front matter&2009Elsevier Ltd.All rights reserved.

doi:10.1016/j.rcim.2009.04.017

?Corresponding author.Tel.:+82632702327;fax:+82632702333.

E-mail address:dwkim@chonbuk.ac.kr(D.-W.Kim).

Robotics and Computer-Integrated Manufacturing25(2009)980–988

2.Related research

2.1.Machining of an impeller

Several research studies have been performed on impeller machining in recent years.They can be classi?ed into three main areas:the determination of the tool axis vector and tool path generation(TPG)without cutter collision,the integrated approach to obtain entire tool paths,and the ef?cient rough machining.To ?nd the tool axis vector that is free from cutter interference,Cho et al.[1]determined the tool center points by removing the interference between a tool and an impeller body using a modi?ed Z-map.They also proposed a tool axis vector determina-tion method through the linear blending of tool vectors on adjacent blade surfaces[2].

Young and Chuang[3]and Chuang and Young[4]suggested more integrated approaches to impeller surface machining,which are compared to other studies focusing on individual machining issues of an impeller,such as collision avoidance between a tool and blades,and the determination of cutter contact or CL data on a hub or blade surface.They considered the quality requirements of a machined part,tool collision or interference,and machining error comparison issues concurrently.Especially,they conducted graphic simulation for machining the blade,hub,and leading edge surface with a software package,Anvil Verify.

Morishige and Takeuchi[5]?rst presented the rough-cut issue associated with an impeller and generated the5-axis control rough-cutting CL data of an impeller-like shape represented by a free form surface model,resolving the cutter interference problem using2-dimensional C-spaces.Balasubramaniam et al.[6] suggested a general method of?nding5-axis rough-cut paths directly from tessellated geometric entities including impeller shaped parts,using visibility cones to avoid local and global interferences with embedded part shapes.

Young et al.[7]developed a5-axis rough machining module dedicated to impeller manufacturing,by focusing on the rough-cut path generation in the narrow and deep machining area based on a deep die cavity machining.They suggested an iso-parametric method to mill machining sections on the blades,considering the residual tool path to prevent the blade from over-cut.Further-more,they considered the minimization of the change in cutting forces in a rough machining process by using the concept of constant scallop height and uniform depth of cut.However,their study did not account for the ef?ciency of their rough-cut plan in terms of machining time.In this sense,Heo et al.[8]proposed a machining region partitioning method considering tool diameters and tool approach vectors concurrently so as to decrease the impeller machining time,producing partitioned unit machining regions with tool approach vectors that are free from tool collision with blade surfaces.

2.2.Case-based reasoning

CBR has many similar features to those of problem solving methods used by human beings.The general process of CBR is as follows:?rst,similar cases are extracted from a case database to solve a new given problem.Then,the extracted solutions are modi?ed to?t the new problem.Finally,the modi?ed,new solution is stored into the updated case database for future use[9].

The CBR method is different from other techniques in arti?cial intelligence or search heuristics and it is popularly used in the area of forecasting and for making recommendations within the management discipline,since it uses empirical knowledge associated within certain situations.Most expert systems using rule-based reasoning infer desired knowledge by employing the so-called IF-THEN rules to knowledge bases.On the contrary,the CBR method,as a data mining method,can be effectively used when the rule-based reasoning is not applicable,simply by storing new cases into the knowledge base.It has shown outstanding performances regardless of the number of cases that it deals with.

Much work has been done in the area of CBR for engineering and business applications.In the engineering?eld,Virkki-Hatakka [10]proposed a well-known engineering-based research to apply CBR for the effective selection of engineering equipments. Schmidt[11]employed CBR to create systematic production plans.Haque[12]developed a type of decision support system through CBR,namely Project Planning Support System(PPSS)by using concurrent engineering for product development processes. Kwon[13]introduced an XML-based CBR system that provides preliminary plans for Enterprise Resource Planning(ERP).

On the other hand,in the?eld of commercial business,Jung

[14]adopted CBR to analyze the risk of electronic commerce.Lee

[15]developed a monitoring model of Electronic Data Interchange (EDI)to improve the performance of EDI for EDI user companies, showing that the CBR-based model outperforms a Model Driven Architecture(MDA)based EDI monitoring model.Chiu[16]used CBR to classify the groups of customers for the direct marketing. Hsu[17]and Changchien[18]developed CBR-based systems for the success of forecasting outsourcing and for marketing plans, respectively,in telecommunication industries.Thus,CBR has been applied to many applications as described before and today it is popularly used for case-based search.This paper presents a study applying the CBR in the development of a decision support system for impeller process planning.

3.Machining process planning for an impeller

3.1.Characteristics of an impeller

Turbo machinery parts,which improve the energy consump-tion ef?ciency in power engines,are widely used in the automobile,aircraft,and shipbuilding industries and they are even used in jet engines and launch rockets.Thus,many research studies on high performance and high ef?ciency turbo machinery parts have been carried out.An impeller,composed of a hub body and several blades,is a high-speed rotor used to compress or transfer?uid in high speed,high pressure,and high temperature environments.Fig.1shows the two types of typical impellers:a splitter and a non-splitter type.The splitter type impeller has shorter blades between major impeller blades.

As seen in Fig.2,an impeller can be largely divided into a hub and blades.Blades consist of pressure surface,suction surface, leading edge,and trailing edge,in which the pressure and suction surfaces are a type of ruled surface that use hub.A leading edge is produced by lofting a curve that is interpolated by the shroud and hub.Thus,the characteristic curve of a leading edge can be found in the shroud and hub.The characteristic curve of a trailing edge

Fig.1.Types of an impeller.(a)Splitter type.(b)Non-Splitter type.

M.-H.Cho et al./Robotics and Computer-Integrated Manufacturing25(2009)980–988981

can be formed by the intersection of an exit surface of a blade with the bottom cylinder of a hub.

The hub and blade surfaces of an impeller are conventionally machined by 5-axis NC machines,as the weight and shape imbalance of the impeller often causes noises and vibrations which can lead to blade breakage [19].However,it is very dif?cult to generate 5-axis machining data because of the geometric and mechanical characteristics of the impeller and for the effective teaching of 5-axis NC machines.Furthermore,it takes more than a week to set up a 5-axis NC machine for actual impeller machining,as well as to prepare the machining process plans.Therefore,it is essential to develop a good decision support system that can be used to reduce the set up and planning timess,based on the existing machining experiences and available experimental data.The system has to support a network-based web environment allowing various personnel such as designers,process planners,production and quality engineers,and even sales staff to have access at their convenience,at any time and place,in support of modern enterprises competing effectively in the global market-place.

3.2.Machining features of an impeller

Machining features of an impeller can be de?ned by consider-ing impeller shape characteristics in view of machining informa-tion (see Table 1).These machining features can support ef?cient machining case retrieval as well as machining process planning,namely:(1)they can support more accurate,similar case retrieval,compared to Boolean-based search or conventional shape feature-based search;(2)they can facilitate standardized,feature-based machining so that machining work load can be properly distributed in advance;and (3)this standardized,feature-based machining process planning can decrease unnecessary delay in the shop ?oor.

3.3.Impeller machining process planning

The 5-axis machining of an impeller starts from the user input associated with raw materials and the ?nal product shape through a CAD database (see Fig.2).Based on this input,detailed machining plans are normally performed by using hub,shroud,and boundary curves of an impeller body and blades.Three common types of input methods can be used,namely (1)a direct CAD drawing,(2)a neutral data format such as Initial Graphics Exchange Standard (IGES),and (3)a point data format considering characteristic curves of an impeller.When the input data are given by the ?rst and the second methods,curves required for NC machining are produced by extracting point data from the characteristic curves of the drawing data,or by generating curves

directly through CAD software.A curve ?tting process is performed to derive the required curves from the point data in the third method.

Based on these curves and point data,key-hole machining is ?rst tried to grip an impeller through the turning operation of the cylindrical blank,as well as 3-axes milling.Then,5-axes NC machining is performed to obtain a practical impeller shape as shown in Fig.1.Finally,machined products are measured in the Coordinate Measuring Machine (CMM)to inspect the important,characteristic points of the impeller blades (see Fig.3).4.Impeller machining process planning using case-based reasoning

4.1.Case-based reasoning for machining strategies

A CBR method is utilized in this study to support the building of the machining strategies.The characteristic curves of an impeller are provided in advance as described in the previous section.Then,the characteristic points and features of the impeller are automatically determined,and the CBR module

Fig.2.Features of an impeller.(a)Shape feature.(b)Characteristic curve.

Table 1

Machining features for rough machining of an impeller.Rough machining module

Major features

Remarks

Machining volume

calculation

Check surface determination in hub and blade

Machining volume:?Raw

stock àHub check surfaces–Blade check surfaces

Depends on

impeller size and type

Machining area

partitioning and tool selection

Machining area partitioning

J Depth direction:0–1,Height direction:0–1

J

Calculation of max/min width

Available tool diameters

Machining area

partitioning [8]

Machining planning

Cutting conditions

J Feed rate (approach,

machining,retract),RPM J Depth of cut,path interval,machining allowance J

Hub,blade,leading edge machining allowance

Tool path pattern

Impeller

machining

process planning [19]

M.-H.Cho et al./Robotics and Computer-Integrated Manufacturing 25(2009)980–988

982

functions,as shown in Fig.4,to search referable,past cases by imposing weight values on the computation of similarity between the existing cases and the current case being considered.

A process planning engineer should enter blade,boundary,and leading edge point data of an impeller into the CBIMS.When a splitter type of impeller exits,the blade and leading edge point data of the splitter have to be keyed in.These input data are used for the automatic computation of the Euclidian distance of the similarity function that was established in advance in the CBR.After the CBR process has been completed,appropriate machining strategies are generated via the decreasing order of the similarity values.From this,the engineer can then create a new solution,based on these alternative machining strategies.

4.2.Representation of cases

Each machining strategy or case should be properly repre-sented for the CBR process in advance.The machining process plans are to be derived based on this case representation.Thus,if a set of impeller shape features is given in the CBR process,several machining strategies are usually generated as alternative solutions.

For the representation of the current case being considered,the characteristic shape features of an impeller shape are de?ned as shown in Fig.5.The big diameter of the impeller body (hub)and the small diameter of the hub are used for these characteristic features,including the height of the impeller blank and the length

Material & Geometry Information

User Input Information

Final Geometry Analysis Hub, Blade, Leading Edge Feature Curve Generation

CAPP

Lathe, 3 axis, 5 axis machining 3 Dimension CMM Calibration

Machining Information Storing

Geometry Feature Curve

(Machining Strategy)

Final Fig.3.Flowchart of the impeller machining process.

Fig.4.Generation of the machining strategies in CBIMS.

M.-H.Cho et al./Robotics and Computer-Integrated Manufacturing 25(2009)980–988

983

of a leading edge.For a blade,both the upper and the bottom distances of two facing,neighboring blades are used for these features,and the blade length and the angle between the blades.A blade volume is also used for the comparison of similarity function values.

Furthermore,to determine the solutions of CBIMS,the machining strategies for the rough machining of impeller blades and a hub,as well as the ?nish machining of blades,leading edges,and hub surfaces,need to be provided together with machining history data,with information of any product defect that occurred in the machining process.Herewith,tool information such as tool diameters,and tapered tools,need to be produced.The machining parameters such as feed rate,RPM,depths of cut,and cut clearances are also provided (see Fig.6).Note that rough machining strategies may have one or several alternatives

according to the partitioning methods of a rough machining area between the blades.

The CBR enables a process planner to share the machining strategies of past,similar parts,by parsing characteristic data,namely machining strategies,associated with the impeller’s shape features,as shown in Fig.6.The CBIMS in this study can ?nally produce NC codes that are suitable for a speci?c NC machine,considering the machining strategies informed by the selected cases through the CBR.

4.3.Case-based search

A case-based search is a process to identify similar cases from a case database or knowledge base,selecting the best matching

Fig.5.Shape features of an

impeller.

Fig.6.CBR of the machining strategies by using shape features.

M.-H.Cho et al./Robotics and Computer-Integrated Manufacturing 25(2009)980–988

984

cases in terms of the similarity value between the existing cases and the new given case that needs to be solved.

The similarity value can be computed through a similarity estimation function and a nearest neighbor(NN)method conventionally used to de?ne the similarity function[20].The NN method produces quantitative measures that represent the extent of similarity between entities or cases,which can be determined by a speci?c standard such as a Euclidian distance.

A similarity(estimation)function and a similarity index for the case-based search in this study are de?ned as follows: Similarity estimation

Max f Ob i g?SI11W1tSI12W2...SI ij W i;0Ob i1;1i n;1j m

(1) Similarity index

SI ij?aàE ij

;0SI ij1;1i n;1j m(2)

Weight values

W?

X n

i?1

W i?1;1i n(3) Notations

V NC

j

:j th shape feature value of a new case

V C

ij

:j th shape feature value of the i th case

E ij?j V NC

j àV C

ij

j

Max f E ij g?a;Min f E ij g?b

An NN method selects the best similar entity from the stored entities in the database by comparing the attributes of the entities.The NN method has been popularly used in various applications such as text mining,cluster analysis,search engines, even more recently in machine learning[12].It is often known as an NN-identi?cation rule where,W i is the i th weight value and the sum of these weight values become1,a is the similarity index, and SI i denotes the biggest shape feature difference between the i th shape feature values of the existing cases and that of the current one.On the contrary,b denotes the smallest shape feature difference in this sense.SI i has a value of1if an existing case exactly matches the one being considered.After the CBR process, the system lists the cases in turn by the order of the best case?rst in terms of the similarity estimation value.

The weight values are imposed based on the user’s intent and these vary according to the relative importance of the shape features of an impeller.The value of each shape feature can be changed by re?ecting the user’s intent on the shop?oor.The default weight values are basically pre-determined by the CBIMS system but,in general,a Boolean operation based search does not consider the value of the weight.

5.Implementation of CBIMS

The proposed CBIMS system runs on a Windows XP platform using an IIS web server supporting Windows XP.Active X of Visual Basic is used for the graphical encoding of the system on the web, while C++is used for the generation of the tool trajectories.For the web page coding and CGI functions,PHP4.XX compatible with IIS has been employed.Finally,a graphic library OpenGL1.1is used for the simulation of the tool paths and tool con?guration,while a server version of MySQL 5.0is used for the entire system implementation.

The preparation of the impeller machining strategies and NC code generation by the proposed CBIMS is as follows:step1:input the design information into the CBIMS system such as the basic shape features of an impeller;step2:search similar cases by using CBR with weight values re?ecting user’s search intents,as shown

Fig.7.CBIMS interface screen.

M.-H.Cho et al./Robotics and Computer-Integrated Manufacturing25(2009)980–988985

Fig.8.Creation of a new rough machining

strategy.

Fig.9.Tool path veri?cation on the web browser.

M.-H.Cho et al./Robotics and Computer-Integrated Manufacturing 25(2009)980–988

986

in Fig.7;step3:select a speci?c case from the listed search results,considering the characteristic features of an impeller;step 4:create a stepwise,new machining strategy suitable for the current case,based on the selected machining strategy,as shown in Fig.8;and step5:once the machining strategy is completed,rough-cut machining paths are produced by CBIMS and tools paths are veri?ed through graphic simulation on the web(see Fig.9).

The data?ow of CBIMS is as follows:as an initial data input,a blade geometry?le,a boundary contour geometry?le,an edge geometry?le,and a splitter geometry?le that marks splitter types,are provided as shown in Fig.10.Then,impeller and blade shape information are extracted based on the input data. Case-based reasoning is performed based on the extracted characteristic features of an impeller.Finally,NC machining codes for an impeller are generated by using the retrieved and modi?ed,rough and?nish machining strategies.

6.Conclusion

The paper has presented a case-based impeller machining strategy support system(CBIMS)for effective NC machining process planning by using the machining features of an impeller, which can be applied in the network-based,geographically distributed e-manufacturing environment.Existing machining strategies can be shared,regardless of time and place,through the CBIMS,and a new machining strategy can be generated quickly by using this cooperation-based,e-manufacturing frame-work of CBIMS.Consequently,expert knowledge can be reused to reduce the time of carrying out process planning,which,in turn, increases the competitiveness of a company,and saves the cost associated with the planning phase.

The characteristics of CBR,when compared to Boolean-based search methods,or case-based search methods by geometric shapes,are as follows:?rst,the machining features in each machining stage are de?ned by a template format,in advance, from the aspects of impeller machining and its characteristic shape features.This can give a more exact,similar case search among existing machining strategies,which facilitates fast process planning such as rough machining planning,as well as consistent machine assignment planning in the downstream processes.Furthermore,the formal shape features of an impeller can assist engineers to build standardized machining process planning and to perform pre-designated,more reliable machining operations.

Second,a prototype decision support system,CBIMS,has been implemented,which provides the machining strategies for the 5-axis NC machining process planning for an impeller by using the so-called CBR,a type of variant-based alternative solution generation method.Production skill information,widely spread in a shop?oor,can be fed into a data-based system to save and maintain the machining strategies for the CBR,as well as to establish the shape features of an impeller.Further,new machining strategies are suggested through the CBR,which are used in the process planning for rough machining,leading edge, blade and hub?nish machining of an impeller.Furthermore,this cooperation-based e-manufacturing environment enables the cooperation of various departments through e-communication in the process planning stages.

However,some limitations still exist in the CBIMS;?rst,it is assumed that considerable?oor data and machining information, such as machining strategies for rough and?nish machining,from many cases,have been stored in the data and/or knowledge bases. However,an inexperienced planner may not be able to?nd or generate an appropriate machining strategy due to failure to search similar previous cases.In this case,the inexperienced planner cannot but ask the support of an experienced,impeller process planner.Second,this study has not addressed the information retrieval speed or accuracy of the CBIMS,even the validity of the solutions of the system.Note that the CBR method normally does not guarantee the optimality of the output solutions.However,the quality of solutions can be improved earlier if data and/or knowledge bases are well managed by engineers,and if useful knowledge and information is continually accumulated.Furthermore,the CBR engine of CBIMS could be enhanced by using data mining technologies such as Bayesian inference and neural networks,which can be good topics of further study.

Acknowledgements

This research was?nancially supported by the Ministry of Education,Science and Technology(MEST)and Korea Institute for Advancement of Technology(KIAT)through the Human Resource Training Project for Regional Innovation.

References

[1]Cho H,Park J,Yoon M,Choi D,Shin B,Lee S.A study on the5-axis CNC

machining of impeller.Journal of the Korean Society of Machine Tool Engineers1997;6(4):19–26.

[2]Cho H,Jung D,Yoon M,Choi D,Shin B,Lee E,et al.The development of

exclusive CAD/CAM system for impeller blades formed by ruled surface II

(a study on the5-axis machining).Journal of Korean Society of Machine Tool

Engineers2002;11(3):1–7.

[3]Young HT,Chuang LC.An integrated machining approach for a centrifugal

impeller.International Journal of Advanced Manufacturing Technology 2003;21:556–63.

[4]Chuang LC,Young HT.Integrated rough machining methodology for

centrifugal impeller manufacturing.International Journal of Advanced Manufacturing Technology2007;34(11–12).

[5]Morishige K,Takeuchi Y.5-axis control rough cutting of an impeller with

ef?ciency and accuracy.In:Proceedings of the1997IEEE international conference on robotics and automation.1997,pp.1241–6.

[6]Balasubramaniam M,Laxmiprasad P,Sarma S,Shaikh Z.Generating5-axis NC

roughing paths directly from a tessellated https://www.sodocs.net/doc/0b1537273.html,puter Aided Design2000;32:261–77.

[7]Young HT,Chuang LC,Gershwiler K,Kamps S.A?ve-axis rough machining

approach for a centrifugal impeller.International Journal of Advanced Manufacturing Technology2004;23:233–9.

[8]Heo EY,Kim DW,Kim BH,Jang DK,Chen FF.Ef?cient rough-cut plan for

machining an impeller with a5-axis NC machine.International Journal of Computer Integrated Manufacturing2008;21(8):971–83.

Fig.10.Data?ow of CBIMS.

M.-H.Cho et al./Robotics and Computer-Integrated Manufacturing25(2009)980–988987

[9]Aamodt A,Plaza E.Case-based reasoning:foundational issues,methodologi-

cal variations,and system approaches.In:Arti?cial intelligence communica-tions,Vol.7(1).IOS Press;1994.p.39–59.

[10]Virkki-Hatakka T,Kraslawski A,Koiranen T,Nystrom L.Adaptation phase in

case-based reasoning system for process equipment https://www.sodocs.net/doc/0b1537273.html,puters and Chemical Engineering1997;21(6):643–8.

[11]Schmidt G.Case-based reasoning for production scheduling.International

Journal of Production Economics1998;56(57):537–46.

[12]Haque BU,Belecheanu RA,Barson RJ,Pawar KS.Toward the application

of case-based reasoning to decision making in concurrent product develop-ment(concurrent engineering).Knowledge Based System2000;12(2–3): 101–12.

[13]Kwon SB,Shin KS.Case-based reasoning support for ERP pre-planning.Korean

Intelligent Information Systems Society2003;9(2):171–84.

[14]Jung C,Han I,Suh B.Risk analysis for electronic commerce using case-based

reasoning.International Journal of Intelligent Systems in Accounting,Finance, Management1999;8(1):61–73.[15]Lee SJ,Han IG.The design of EDI controls using case-based reasoning.

International Journal of Intelligent Systems in Accounting,Finance and Management1998;7(3):135–52.

[16]Chiu C.A case-based customer classi?cation approach for direct marketing.

Expert Systems with Applications2002;22(2):163–8.

[17]Hsu CI,Chiu C,Hsu PL.Predicting information systems outsourcing success

using a hierarchical design of case-base reasoning.Expert Systems with Applications2004;26(3):369–441.

[18]Chanchien SW,Ming-Chin L.Design and implementation of a case-based

reasoning system for marketing plans.Expert Systems with Applications 2005;28(1):43–53.

[19]Kim D,Gil G,Heo E,Kim B,Chen FF.Development of A CAPP system for5-axis

machining of impellers.In:The16th International Conference on Flexible Automation and Intelligent Manufacturing,University of Limerick.Ireland, 2006.

[20]Zezula P,Amato G,Dohnal V,Batko M.Similarity search-the metric space

approach.New York:Springer;2006.

M.-H.Cho et al./Robotics and Computer-Integrated Manufacturing25(2009)980–988 988