Journal of Materials Processing Technology 139 (2003) 267–272

A holonic architecture of the concurrent integratedprocess planning system

Jie Zhanga, Liang Gaob, Felix T.S. Chanc,∗, Peigen Liba CIM Research Institute, Shanghai Jiao Tong University, China

b School of Mechanical Science and Engineering, Huazhong University of Science and Technology, Chinac Department of Industrial and Manufacturing Systems Engineering, The University of Hong Kong,

Pokfulam Road, Hong Kong, China

Abstract

This paper describes a conceptual architecture of applying holon to process planning (PP) for integrating design, PP, and shop floorscheduling so that a concurrent integrated process planning system (CIPPS) is constructed. The problem in PP has been mapped onto aholonic architecture based on multistage cooperation. A holonic architecture for CIPPS integrates all the activities of PP into a distributedintelligent open environment. Holons for CIPPS can be organized dynamically, and cooperate with each other to perform an appointedtask flexibly. In addition, using CORBA technology, a prototype system of the CIPPS is implemented.© 2003 Elsevier Science B.V. All rights reserved.

Keywords:Process planning; CAD; Production scheduling; Integration; Holon; CORBA

1. Introduction

Traditionally, the activities of product design, processplanning (PP) and shop floor scheduling are viewed asdiscrete stages in manufacturing enterprise and developed,respectively, into automation systems. However, decisionstaken in one stage affect the decision in downstream stages[1]. Computer aided PP has been recognized as playing akey role in computer integrated manufacturing (CIM) thatintegrates these islands of automation. It forms an impera-tive connection between design and manufacturing opera-tions in CIM. However, most of the PP is generated withoutconsideration of real-time dynamic status information in theshop floor, and assumes an infinite capacity of resources onthe shop floor, and usually a factory is empty without anywork-in-progress when assigning resources to jobs[2,3].These assumptions lead to the repeated commitment ofcertain popular resources to numerous process plans[4,5].Over the past two decades, considerable efforts have beenexpended in developing integrated computer aided design(CAD) and computer aided manufacturing functionalities.Based on the philosophy of concurrent engineering (CE), aconcurrent integrated process planning system (CIPPS) isproposed in this paper to integrate CAD, PP, and production

∗ Corresponding author.E-mail address:ftschan@hkucc.hku.hk (F.T.S. Chan).

scheduling system (PSS) so that it is expected to signifi-cantly enhance the ability of manufacturing companies toadapt efficiently to changing conditions, and yield signif-icant performance improvements (e.g., shorter lead times,increased resource utilization, enhanced due-date perfor-mance and coordination between customers and suppliers).In order to implement CIPPS, a holonic architecture ispresented, which has characteristics such as distribution, au-tonomy, interaction, and openness to meet the requirementof CIMS.

The rest of this paper is organized as follows.Section 2proposes a concurrent integrated PP model.Section 3de-scribes a conceptual architecture of applying holons to PPfor integrating design, PP, and production scheduling. Theneeded holons are defined.Section 4describes the cooper-ation between holons. Using CORBA technology, the CIPPprototype system is constructed inSection 5. Section 6isthe conclusions.

2. Concurrent integrated PP model

CE is an important method of the integration from infor-mation to function for CIMS, and an overall optional methodfor the complicated manufacturing systems. A problem inthe traditional product development cycle is that the commu-nication between different perspectives is linear in nature.

0924-0136/03/$ – see front matter © 2003 Elsevier Science B.V. All rights reserved.doi:10.1016/S0924-0136(03)00233-4

268 J. Zhang et al. / Journal of Materials Processing Technology 139 (2003) 267–272

Fig. 1. PP model.

This may lead to long and costly development cycles asdesign mistakes are often not discovered until downstreamstages of product development. In contrast to the traditionalsequential approach to design and manufacturing, CE advo-cates a rapid, simultaneous approach, where concept devel-opment, design, manufacturing, and support are carried outin parallel[6]. This approach prompts developers, from theoutset, to consider all the relevant element of product life cy-cle from conception to disposal, including quality, cost, sche-dule, and user requirements[7]. From the implementationpoint of view, the CE paradigm can be viewed as an integra-tion of functional software tools. Based on the philosophyof CE, in this paper an integrated PP system is proposed tointegrate CAD, PP, and PSS. It can provide not only a toolfor the early stage of design but can also reach nearly an op-tional process plan by working with PP and PSS in parallel.

PP involves multistage decision making so that its tasksmay be divided into three levels, namely initial planninglevel, decision making level, and detailed planning level tocomplete the task of PP. The activities with each level takeplace in different time periods, as shown inFig. 1. Initialplanning should be executed in a very early stage, e.g., assoon as the product design is finished. Initial planning levelgenerates a serial of alternative processing routes of partbased on processing potential in the shop floor. Decisionmaking is executed in a later stage, i.e., when the orders havebeen released to the shop. In the decision making level, onesuitable process plan can be selected for a part from alter-native process plans generated in the initial planning levelbased on the current status of the shop floor. The detailedplanning is executed just before manufacturing begins. Thedetailed planning level generates a detail PP for the processroute by the decision making level. The interaction betweenPP and CAD takes place in the first level, but the interactionbetween PP and production scheduling takes place in all thethree levels (Fig. 1).

3. Holonic architecture for CIPPS

3.1. Applying holon to PP

In order to implement CIPPS, a holonic architecture ispresented, which has characteristics such as distribution,

autonomy, interaction, and openness to meet the require-ment of CIMS. The holonic manufacturing system (HMS)consortium introduced “holonic” as a hierarchical archi-tecture of self-consistent, cooperating modules. A holon isautonomous and co-operative and sometimes intelligent. Itcan be made up of other holons. Thus, holon means simulta-neously a whole and a part of the whole. This feature distin-guishes holons from agents. Sousa and Ramos[8] introducea dynamic scheduling holon for manufacturing orders.

From software engineering point of view, holons can bethought of as a natural extension to object-oriented pro-gramming. Each holon has its own thread of control, andruns independently of the other holons. Holons interactamong themselves by sending and receiving messages, butthe messages are not in the form of function calls. It is pos-sible to multicast a message to a group of holons, and eachrecipient holon is free to ignore a message or deal with itat a later stage. Using the holon conception, a complicatedsystem can be divided into different and efficient holonsthat are more realizable. By this way, the flexibility, expand-ability and stability of the CIPPS are improved. In order toimplement a holonic architecture for CIPPS, the first task isto partition the holons in a system. The principles of divisionmainly lie in function grouping and sizing. Different func-tional modules will be grouped into different holons. How-ever, the size of a holon cannot be too large or too small. Ifit is too large, it will be insufficiently configurable and diffi-cult to change. If it is too small, it will need to define moreinterface messages and will be more difficult to integrate.According to the above principles and a new PP model(as shown inFig. 1), the CIPPS is partitioned into severalholons, such as task holon, initial planning, decision holon,detailed holon, etc. These holons can be mapped to differentfunctions and objects of the PP. They can be organized dy-namically, and perform flexibly and cooperatively the taskof PP.

3.2. Holonic architecture

Holonic architectures are robust and dynamic; they canquickly react to unexpected events, and adapt to changingconditions. They are inherently distributed and scaleable:more holons and more computers can be added as necessaryto increase the performance or the capacity of a system.

J. Zhang et al. / Journal of Materials Processing Technology 139 (2003) 267–272 269

Fig. 2. Holonic architecture of CIPPS.

Compared to centralized systems, holonic architectures areeasy to maintain, modify, and extend as the requirementsfrom the system change and grow with time. Therefore,holonic architecture can be used to construct a CIPPS.

Holonic architecture for CIPPS, as shown inFig. 2,consists of an initial planning holon, a decision makingholon, a detailed planning holon, a task holon, a resourceholon, a C/P coordination holon, and a P/P coordinationholon. It integrates all the activities of PP into a distributedintelligent open environment. Holons cooperate with eachother to implement system functionality. All the holonswork together to solve the whole PP problem cooperatively.Koestler [9] used the term “holarchy” to describe holonichierarchies. Based on the PP model (Fig. 1), the holonicarchitecture of CIPPS is, respectively, divided into threeholarchies, namely, initial planning, decision making, anddetailed planning. They are employed to perform the spec-ified PP task in the three different levels. Each holarchy ismade up of many holons.

Some of the basic holons are defined as follows.The task holon. To automatically generate process plans

from a CAD, a task holon is needed to animate humangeometric reasoning capabilities in order to transfer designdata into manufacturing specific semantic information. Thetask holon accepts product models, and translates them intothe CIPPS internal shared common model. The task holonknows about its work material, due date, and penalty forlateness. When the design information has been properlyinterpreted, the task holon will trigger the initial planningholon to create a process plan for manufacturing the de-signed part. Because PP is divided into three different lev-els, the decision sequencing and flow of control in CIPPSis controlled by the task holon.

The C/P coordination holon. It provides an open and uni-form coordination between CAD and PP systems. It sup-ports customizable functionalities that enable the user at theinitial planning level within the PP model to interactivelymanipulate and evaluate product design while selectivelycoordinating the requirement of the initial planning holon.On the other hand, when designing a new product, the CADdesigner can inquire of the related manufacturing process

information by C/P coordination, to ensure product manu-facturability and profitability.

The P/P coordination holon. It provides an open anduniform coordination between PP and PSS in all the threePP levels. It supports customizable functionalities that en-able the user at all the three planning levels to interactivelymanipulate and evaluate alternative PP and final PP whileselectively coordinating the requirement of the productionscheduling holon. The holonic architecture of CIPPS aimsat providing a framework for coordinated development andmanipulation of PP.

The resource holon. It is shared in the PP system andthe PSS. It provides the resource information in the currentmanufacturing environment.

The initial planning holon. It is the core at the initial plan-ning holarchy. It cooperates with other holons to finish thetask at the initial planning holarchy. The data of the input tothe initial planning holon consists of two types. One of themis the product model from a design system that represents thegeometrical shape of the product, design tolerances, surfacerequirement, and the properties of the material. The otherone is the resource information from the PSS. The task ofthe initial planning holon is to seek all feasible operations bypart information and resource information. This means thatthe initial planning holon does not carry out the completePP, instead it generates many feasible blocks that consist offeasible processes based on individual operations. A block isgenerally manufactured on an individual machine unless itis mass type production. Therefore, the output of the initialplanning holon is a series of the blocks. These blocks are as-signed to each machine based on feedback from productionscheduling. In fact, all these alternative blocks form the al-ternative process plans. Another task of the initial planningholon is manufacturing evaluation as a tool for the designer.The designer can use the tool to check the manufacturing fea-sibility and cost of alternative designs. Thus, designers canmodify their designs so that they are manufacturable and costeffective in terms of the current shop floor environment. Themanufacturing evaluations produced by CIPPS can be con-trasted to other approaches in that its evaluations are basednot only on the design (namely process capabilities that can

270 J. Zhang et al. / Journal of Materials Processing Technology 139 (2003) 267–272

be described by the part shape, dimensions, tolerances, sur-face finish, geometric and technological constraints, and eco-nomics of a process), but also on information on the shopfloor resources. Shop floor resources greatly affect the man-ufacturability analysis. A design may be manufacturable un-der one set of shop floor resources, but not index another.

The decision making holon in the CIPPS. It is the core ofthe entire system. One of the important activities between PPand production scheduling is how to select machines basedon the real-time shop floor states. The decision making holonmatches the required job operations with the operation capa-bilities of the available production resource. It is dynamicallyconstrained and requires a close link to the PSS. Once thecurrent shop floor information is obtained, the process plansare constructed at the last minute when the job is about tobe loaded to the shop floor. The task of the decision makingholon is to decide an optimal PP from alternative process-ing routes that are generated by the initial planning holon.The decision making holon first lists all the possible processplans; then based on the real-time machine status which isprovided by the resource holon and certain criteria, such asminimize manufacturing cost and minimize manufacturinglead time, calculates their manufacturing cost or manufac-turing lead time; lastly, it selects the process plan with leastmanufacturing cost or least manufacturing lead time. An op-timal process plan will eventually be transmitted to the PSS.

The detailed planning holon. It is executed just beforethe beginning of manufacturing. It will generate the detailPP, which includes selecting the cutting speed, feed rate,etc., and calculating machining time, programming the NCprogram for each operation of processing route generatedby the decision making holon. The output of the detailedplanning holon is an entire document of PP to be sent to theshop floor to guide the production.

In addition to the holonic system having some commoncharacteristics such as distribution, autonomy, interaction,

Fig. 3. Cooperation between CAD, PP, and PSS.

and openness, the architecture possesses some other charac-teristics as follows:

1. Reconfiguration. It supports easy reconfiguration toaccommodate the introduction of new manufacturingenvironment.

2. Customization. It supports customizable functionali-ties that enable user at all the three planning levels tointeractively manipulate PP.

3. Hybridization. It is hybrid architecture that is composedof the peer-to-peer relationships (task holon with ini-tial planning holon) and the hierarchical relationshipsbetween holons (such as between planning holons).

4. Cooperation between holons

A holonic system provides a cooperation mechanism be-tween the different autonomous intelligent entities. Differentholons are designed according to the intelligent model of dif-ferent functions. When the prerequisites are satisfied, theseholons execute tasks defined by designer in a dynamic envi-ronment so as to response continuously the change of man-ufacturing conditions and production tasks. These holons inCIPPS can be organized dynamically, and cooperate witheach other to perform an appointed task flexibly. Negotiationis the key activity when holons cooperate one another. Co-operation between holons mainly concentrates on the threeaspects: PP with CAD and PSS, the initial planning holarchy,and decision making holarchy. Cooperation between CAD,PP, and the PSS is illustrated inFig. 3. The initial planningholon cooperates with the task holon, the C/P coordinationholon, the resource holon, and the P/P coordination holon tofinish together the task at the initial planning holarchy andimplement the integration with CAD and PSS. The cooper-ation model between holons in the initial planning holarchyis shown inFig. 4. The decision making holon cooperates

J. Zhang et al. / Journal of Materials Processing Technology 139 (2003) 267–272 271

Fig. 4. Cooperation in initial planning holarchy.

Fig. 5. Cooperation in the decision making holarchy.

with task holon, resource holon, and P/P coordination holonto finish together the task at the decision making holarchyand implement the integration of PP and scheduling systemin the decision making holarchy. The cooperation modelbetween holons in the decision making holarchy is shownin Fig. 5.

5. Implementing CIPPS

One of the major goals of establishing the concurrent in-tegrated PP prototype system is the enabling of applicationintegration by providing services of sharing data and syn-

chronizing the operation of applications. CORBA based ar-chitecture of the CIPP prototype system is shown inFig. 6,which provides an organized environment for running a col-lection of holons in CIPPS. The architecture helps the userto integrate the software holon and other applications. Eachholon can be located in a different machine or the same ma-chine, or runs on a different operating system, depending onthe requirements of CIPPS application in the real environ-ment, is shown inFig. 6. The decision making holon canbe installed with the PSS in the PIII Two computer. Otherholons can be located in the PIII One computer. All theholons work together to solve the entire PP problem coop-eratively. An initial prototype of CIPPS has been developed,

272 J. Zhang et al. / Journal of Materials Processing Technology 139 (2003) 267–272

Fig. 6. The architecture of a CIPP prototype system.

and the system is currently undergoing further refinementand evaluation.

6. Conclusions

A problem in the traditional product development cycleis that the communication between different perspectives islinear in nature. This would lead to long and costly devel-opment cycles as design mistakes are often not discovereduntil the downstream stages of product development. Thispaper proposes a CIPP system. An initial testing phase ofthe CIPP prototype system has been completed in close col-laboration with an industrial partner with very encouragingresults. It is shown, first, to provide a tool for CAD by in-tegrating downstream constraints into the design phase sothat the need for redesign (due to design mistakes) later canbe reduced in the product development cycle. Secondly, itintegrates PP and production scheduling so that the CIPPSprovides a process plan that optimizes due-date performancewhile minimizing manufacturing cost or lead time by takinginto account alternative process plans and the real-time re-source status in shop floor. Thirdly, it responds continuouslyto manufacturing conditions and production task changeswith the help of these holons that can be organized dynami-cally, and cooperate with each other to perform an appointedtask flexibly. Therefore, the development of holonic archi-tecture for CIPPS satisfies the requirements of agile manu-facturing: rapid response to changing requirement; reductionin both time and cost of the product realization process; and

integration within a heterogeneous, wide-area networkedenterprise.

Acknowledgements

This work was supported by the National Nature ScienceFoundation of China under grants 59885004, 59990470 and59889505.

References

[1] P. Gu, S. Balasubramnian, D.H. Norrie, Bidding-based process plan-ning and scheduling in a multi-agent system, Comp. Ind. Eng. 32 (2)(1997) 477–496.

[2] T.W. Liao, E.R. Coates, F. Aghazadeh, et al., Modification of CAPPsystems for CAPP/scheduling integration, Comp. Ind. Eng. 26 (3)(1994) 451–463.

[3] H.-C. Zhang, IPPM—a prototype to integrate process planning andjob shop scheduling functions, Ann. CIRP 42 (1) (1993) 513–518.

[4] H.-C. Zhang, S.H. Huang, A fuzzy approach to process plan selection,Int. J. Prod. Res. 32 (6) (1994) 1265–1279.

[5] H.-C. Zhang, S. Mallur, An integrated model of process planningand production scheduling, Int. J. Comp. Integr. Manuf. 7 (6) (1994)356–364.

[6] R.T. Yeh, Note on concurrent engineering, IEEE Trans. Know. DataEng. 4 (5) (1992) 407–414.

[7] B.N. Yetukuri, N.V. Yetukuri, G.W. Fischer, SPAW: a design toolfor planning a manufacturing process in a concurrent engineeringenvironment, Comp. Ind. 32 (1) (1996) 79–93.

[8] P. Sousa, C. Ramos, A dynamic scheduling holon for manufacturingorders, J. Intell. Manuf. 9 (2) (1998) 107–112.

[9] A. Koestler, The Ghost in the Machine, Arkana Books, 1989.