You are on page 1of 23

Computers & Industrial Engineering 48 (2005) 765787

www.elsevier.com/locate/dsw

A complete cellular manufacturing system design methodology based on axiomatic design principles
O. Kulaka,1, M.B. Durmusoglua,*, S. Tufekcib,2
a

Industrial Engineering Department, Istanbul Technical University, Macka, Istanbul 34367, Turkey b Department of ISE, University of Florida, 303 Weil Hall, Gainesville, FL 32611, USA Available online 8 January 2005

Abstract This paper provides a framework and a road map for people who are ready to transform their traditional production system from process orientation to cellular orientation, based on Axiomatic Design (AD) principles. A feedback mechanism for continuous improvement is also suggested for evaluating and improving the cellular design against pre-selected performance criteria. A complete implementation of the proposed methodology at a manufacturing company and resulting performance improvements are also provided. q 2004 Elsevier Ltd. All rights reserved.
Keywords: Cellular manufacturing; Axiomatic design; Implementation guide

1. Introduction With the increasing popularity of lean thinking, cellular manufacturing has become a signicant focal point in manufacturing. In concert with this development, there have been numerous publications on design and improvement of cellular manufacturing systems (CMSs). Among these, we can list part family and machine group determination (Nancy & Wemmerlov, 2002), evaluation of grouping efciency measures (Sarher & Mondal, 1999), group scheduling (Wemmerlov, 1992), machine layout in cells (Aneke & Carrie, 1986) and capacity planning in cellular manufacturing (Sule, 1991). Numerous modeling techniques have also been proposed for part family and machine group determination, including visual analysis (Burbridge, 1969; Groover, 1987), coding and
* Corresponding author. Tel.: C90 212 293 1300x2666. E-mail addresses: okulak@itu.edu.tr (O. Kulak), durmusoglum@itu.edu.tr (M.B. Durmusoglu), tufekci@ise.u.edu (S. Tufekci). 1 Tel.: C90 212 293 1300x2746. 2 Tel.: C1 352 392 6753.
0360-8352/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.cie.2004.12.006

766

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

classication (Choi, 1992; Hyer & Wemmerlow, 1987), clustering algorithms (Singh & Rajamani, 1996), mathematical methods (Heragu & Chen, 1998; Soanopoulou, 1999) and articial intelligence approaches (Soleymanpour, Vrat, & Shankar, 2002; Zhao & Wu, 2000). All these approaches are developed to satisfy only one or limited functional requirements of the CMS design. Approaches, which include all aspects of cellular system design, are very limited. Silveira (1999) provides one such approach. This approach intends to integrate concepts and techniques into an integrated system in a logical sequence. This approach, however, is mostly based on his past experience and lacks detailed principles for implementation. With its scientic basis, the authors believe that axiomatic design (AD) approach (Suh, 1990) will provide a sound and systematic basis to cellular manufacturing design. Many AD applications in designing products, systems, organizations and software have appeared in the literature in the last 10 years. AD theory and principles have been introduced rst time by Suh (1990). Black and Schroer (1988) provide denition and use of decouplers for achieving exibility in CMSs. Gunasekera and Ali (1995) have provided a three-stage approach to metal forming process. The stages are composed of conceptual stage, initial stage and nal stage, respectively. The conceptual stage is designed using AD approach. Suh (1997) provided a conceptual approach for dening, classifying and design of systems using AD methodology. Suh, Cochran & Paulo (1988) provided an AD-based model for an ideal production system in line with lean principles. Babic (1999) provides a decision support system for arrangement of exible manufacturing systems. This approach uses AD design principles together with FLEXY intelligent system. Cochran, Eversheim, Kubin, and Sesterhenn (2000) convert complex production system into small, exible and decentralized production segments. In this approach they use lean principles in conjunction with segmentation and AD principles. Cochran, Kim & Kim (2000) provide a performance evaluation system for production system. Chen, Chen, and Lin (2000) proposed a knowledge-based decision support system using independence axiom of AD in order to improve cell performance. Houshmand and Jamshidnezhad (2002) also provide a lean manufacturing based production system design model using AD approach. In this model organizational capabilities, technological capabilities and value stream analysis are used as the basis. In addition to the design of manufacturing systems, AD approach has also been used in software design (Kim, Suh, & Kim, 1991), product design (Tseng & Jiao, 1997) and quality system design (Suh, 1995) areas. These studies have convincingly shown the applicability and benets of AD in solving industrial problems. Considering the literature mentioned above, a road map including all functional requirements of CMS design is not found. In this study, a methodology is developed using AD principles in order to ll this void in the design process.

2. Methodology With this work, a road map for people who are ready to transform their traditional production system from process orientation to cellular orientation, based on AD principles is provided. In addition, a feedback mechanism for continuous improvement is also provided for evaluating and improving the cellular design against pre-selected performance criteria (Fig. 1). Selection of project team, ensuring broad-based participation, analysis of current conversion process, and determining conversion strategy for transition to cellular manufacturing constitute the rst stage of our proposed methodology. The design of the CMS starts following the preliminary stage. At this stage the AD approach to cellular manufacturing is presented to the transition and design team in a systematic

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

767

Fig. 1. Cellular manufacturing design process with feedback mechanism based on AD principles.

and scientically sound order. These guidelines, which are developed, based on the Independence Axiom, provide necessary steps in transforming an existing process oriented system into a CMS. With this systematic approach the designer will be guided by our methodology for appropriate analysis techniques. Transition to cellular manufacturing follows after all cellular manufacturing steps are successfully completed. At this stage, the production is achieved through CMS. System is now generating necessary databases and information for comparing the system performance with set target goals on some business metrics. Based on target values and achievements, new target values are established and appropriate system modications and changes are affected through CMS improvement procedure provided in our proposed methodology. This procedure is also based on AD Principles. This continuous feedback and improvement procedure is in agreement with the spirit of lean thinking and Kaizen activities. In this paper, a complete set of Functional Requirement-Design Parameter (FR-DP) hierarchy for the design of a CMS through AD principles is provided. This corresponds to the second box (Design of CMS Based on AD Principles) in Fig. 1. Since AD based continuous improvement procedure for feedback mechanism enlarges the paper in scope, this procedure is not included in this study. The following section explains the principles of AD. Section 4 provides the proposed FR-DP hierarchy. The design matrices corresponding to the FR-DP pairs are also provided in this section. In Section 5, a case study of converting a conventional metal ramp/stairwell manufacturer from process-oriented layout to cellular manufacturing based on the design methodology is presented. Finally, Section 6 provides conclusions of our work.

3. Principles of axiomatic design The ultimate goal of AD is to establish a scientic basis for design and to improve design activities by providing the designer with a theoretical foundation based on logical and rational thought processes and tools (Suh, 2001). In accomplishing this goal, the AD provides a systematic search process through the design space to minimize the random search process and determine the best design solution among many alternatives. The most important concept in axiomatic design is the existence of the design axioms. The rst design axiom is known as the Independence Axiom and the second axiom is known as the Information axiom. They are stated as follows (Suh, 1990).

768

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

Axiom 1. The Independence Axiom. Maintain the independence of functional requirements. Axiom 2. The Information Axiom. Minimize the information content. Mathematically, the relationship between the FRs and DPs are expressed as fFRg Z jAjfDPg Here, {FR} is the functional requirement vector {DP} is the design parameter vector, and jAj is the design matrix that characterizes the design. In general each entry aij of jAj relates the ith FR to the jth DP. The structure of jAj matrix denes the type of design being considered. In order to satisfy the independence axiom, jAj matrix should be uncoupled or decoupled design. jAj matrix is classied into three categories as dened below: Uncoupled Design (most preferred). In this design, the jAj matrix is a diagonal matrix indicating the independence of FR-DP pairs. So each FR can be satised by simply considering the corresponding DP. Decoupled Design (second choice). In this design the corresponding jAj matrix is triangular. Therefore the FRs can be answered systematically FR1 to FRn by only considering the rst n DPs. This design appears most frequently in real life. Coupled Design (undesirable). In this design the jAj matrix is has no special structure. Therefore a change in any DP may inuence all FRs, simultaneously. In designing systems with AD principles we try to avoid coupled design as much as possible.

4. Cellular manufacturing system design through AD principles Step 1. Choose FRs in the Functional Domain. The rst step in designing CMS is to dene the functional requirements (FRs) of the system at the highest level of its hierarchy in the functional domain (Suh, Cochran, & Paulo, 1998). At this stage many functional requirements may be established. Each functional requirement established at this stage may lead to a completely different cellular manufacturing design. Therefore, extreme care should be given to all functional requirements before a single functional requirement is adopted at this highest level. In this work the following has been selected as the highest FR. FR Z Provide customized production More product variety, in smaller batch sizes with highest quality and more frequent deliveries at lower costs summarize the needs of customers in the 21st century (Steudel & Desruelle, 1992). These requirements are forcing companies to re-evaluate their classical manufacturing systems for more exibility in response to these customer needs. The exibility of a manufacturing system is measured by its speed and ability to respond to rapidly changing customer needs. Step 2. Mapping of FRs in the Physical Domain. Design parameters (DPs), which satisfy the FRs established in the previous step, are selected through a mapping process between the functional domain and the physical domain. In order to make the correct DP selection, the DP set corresponding to the FR

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

769

set established before, must be exhaustively generated. The following DP has been selected to satisfy the FR provided above. DP Z CMS Design The production system, which can answer customers needs in an efcient way through elimination of waste, reduction of lead time and improved quality is a CMS designed with lean manufacturing principles in mind. Step 3. Decompose FR in the Functional Domain-Zigzagging between the domains. If the DPs proposed for satisfying the FRs dened in steps above can not be implemented without further clarication, the AD principles recommends returning to the functional domain for decomposing the FRs into their lower functional requirement set (Fig. 2). The following lower functional requirements set is dened for decomposing the FR determined in Step 1 above. FR1ZClassify and group products/parts and machines for simple material ow FR2ZDevelop resources capability based on product specications FR3ZRearrange resources to minimize waste FR4ZProvide production based on customer demand. Step 4. Find the Corresponding DPxs by Mapping FRxs in the Physical Domain. In satisfying the four FRs dened above, we move to the physical domain from the functional domain. The following DPs are in response to the FRs listed above. DP1ZProcedure for dening product/part families and machine groups DP2ZProcedure for developing production resources capability DP3ZProduct oriented layout DP4ZPull production control system. Step 5. Determine the Design Matrix. Once the FR-DP sets are dened in Steps 3 and 4, the corresponding Design Matrix (DM) provides the relationships between the FR and DP elements. It is important to insure that the DM as established satises the Independence Axiom (IA) of the AD principles. If the DM matrix is uncoupled or decoupled, then it satises the Independence Axiom of AD principles (see Suh (2001)). The design equation and the DM corresponding to the FR-DP sets are as follows. 2 FR1 3 2 X X X X X X X 32 DP1 3

7 6 6 6 FR2 7 6 X 7 6 6 6 FR3 7 Z 6 X 5 4 4 FR4 X

7 76 76 DP2 7 7 76 76 DP3 7 5 54 DP4

(1)

This design is a decoupled design, and thus, satises the IA. In the DM above, a symbol X represents a strong relationship between the corresponding FR-DP pair. Step 6. Decompose FR1, FR2, FR3 and FR4 by going from the Physical to the Functional Domain again and determine the corresponding DPs.

770

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

Fig. 2. The decomposition of CMS design.

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

771

4.1. Classication and grouping products/parts and machines branch This is a very important step in the success of transition from traditional manufacturing to cellular manufacturing (Irani, 1999). The rst step of this branch is to establish the high volume products through Product-Quantity (Pareto) analysis. Therefore the products for CMS are simply determined. The next step is to group similar parts (in terms of their operation requirements) into parts families. There are several algorithmic procedures to accomplish this task. Most of them use the Machine-Part Incidence Matrix. These algorithms swap rows and the columns of this matrix until suitable block-diagonal submatrices or near block-diagonal sub-matrices are obtained. The parts that fall into the same sub-matrix are candidates to be allocated to a potential cell. Once these potential cell allocations are complete, the next and the nal step of this branch is to decide on how many of these cells to implement based on economic justication principles. The designers perform costbenet analyses on each potential cell formation. In this process, each candidate cells contribution to the companys bottom line in terms of productivity, lead time and protability together with return on investment are calculated. Those cells that satisfy the company internal rate of return are recommended for formation. The functional requirement FR1 (classify and group products/parts and machines for simple material ow) as dened above may be decomposed with DP1 (procedure for dening parts families and machine groups) in mind as: FR11ZDetermine high volume products/parts to group FR12ZDetermine operations and machine types for producing each part family FR13ZCluster the parts/machines based on the similarity of operations FR14ZDetermine nal number of machine groups. The corresponding DPs may be stated as: DP11ZProduct-Quantity Pareto Analysis DP12ZMachine-Part Incidence Matrix DP13ZParts/machines clustering techniques DP14ZCost Analysis and economic justication techniques. The design matrix for the above set of FRs and DPs are 3 32 3 2 2 DP11 FR11 X 7 76 7 6 6 76 DP12 7 6 FR12 7 6 X X 7 76 7 6 6 76 DP13 7 6 FR13 7 Z 6 X X X 5 54 5 4 4 FR14 X X X X DP14

(2)

Once again, this is a decoupled design satisfying IA of AD. The functional requirement FR14 (determine nal number of machine groups) as dened above may be decomposed with DP14 (Cost Analysis and economic justication techniques) in mind as: FR141ZEnsure improvements of the cell efciency FR142ZProve the economical performance of the cell.

772

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

The corresponding DPs may be stated as: DP141ZDimensional cell analysis DP142ZBenets/Costs Analysis. The design matrix for the above set of FRs and DPs are #" # " # " DP141 FR141 X Z X X DP142 FR142

(3)

Once again, this is a decoupled design satisfying IA of AD. The functional requirement FR141 (ensure improvements of the cell efciency) as dened above may be decomposed with DP141 (dimensional cell analysis) in mind as: FR1411ZEliminate inappropriate part assignments FR1412ZEliminate inappropriate machine assignments FR1413ZIncrease capability of bottleneck machine which is required more than one cell FR1414ZRedesign the products for producing the exceptional parts in the related cell FR1415ZRemove the exceptional parts that make the system complex FR1416ZEliminate the bottleneck machines (The bottleneck machine is either a machine which is required more than one cell or a machine which its cycle time is greater than takt time.) FR1417ZEliminate the high-level utilization of common machines. The corresponding DPs may be stated as: DP1411ZParts assigned to proper cell DP1412ZInter-cell machine transfer DP1413ZMethod Engineering DP1414ZRedesigned product DP1415ZParts that have been transferred out of the cellular system DP1416ZMachine duplication between cells and/or new machine acquisition according to the takt time DP1417ZMerging cells. The design matrix for the above set of FRs and DPs are 3 2 FR1411 X 6 FR1412 7 6 X 7 6 6 7 6 6 6 FR1413 7 6 X 7 6 6 6 FR1414 7 Z 6 X 7 6 6 7 6 6 6 FR1415 7 6 X 7 6 6 4 FR1416 5 4 X X FR1417 2 3 DP1411 76 DP1412 7 7 76 7 76 76 DP1413 7 7 76 76 DP1414 7 7 76 7 76 76 DP1415 7 7 76 54 DP1416 5 X DP1417 32

X X X X X X

X X X X X

X X X X

(4)

X X X

X X

Once again, this is a decoupled design satisfying IA of AD.

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

773

In this stage, nal assignments of parts to cells are realized. Particular attention is given on eliminating of exceptional parts for decreasing inter-cell part movements. Furthermore due consideration is also given to reassignment of machines to different cells for elimination inter-cell movements. Also due attention is given to the machines which can be used instead of bottleneck machines for improvement of its ability to eliminate the bottlenecks and thus decrease inter-cell movements. After all these efforts if there are some parts which visit multiple cells, the designs of parts are scrutinized through concurrent engineering for possible design changes which may eliminate or decrease the inter-cell movements of these parts. If all fails, these parts may be considered to be taken out of the cellular system, completely. In the nal analysis, eliminating the bottleneck machines is considered. Machine duplication between cells is an alternative to eliminate the bottleneck and to decrease the rest of the inter-cell movements. In addition, a bottleneck machine may appear in the cell according to the takt time. Additional machine alternative is considered to respond to the desired takt time. Although all these efforts are performed, the high-level utilization of common machines between cells is usual. The designer may consider joining several cells into a larger cell, which require most commonly used machines in their operation. 4.2. Resources capabilities development branch In this stage, activities including waste elimination, for the design of a lean process, rearrangement of resources in agreement with the takt time, xing training needs and motivation of the workforce are accomplished. The functional requirement FR2 (develop resources capability based on product specications) as dened above may be decomposed with DP2 (procedure for developing production resources capability) in mind as: FR21ZDevelop process FR22ZSelect most appropriate process elements FR23ZDetermine required training/education needs FR24ZMotivate labor participations. The corresponding DPs may be stated as: DP21ZSimplied and lean process DP22ZProduction resources selection procedure DP23ZMulti-purpose labor training programs DP24ZGain sharing program. The design matrix for the above set of FRs and DPs are 3 32 3 2 2 DP21 FR21 X 76 DP22 7 6 FR22 7 6 X X 7 76 7 6 6 54 DP23 5 4 FR23 5 Z 4 X X X FR24 X X X X DP24

(5)

Once again, this is a decoupled design satisfying IA of AD.

774

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

The functional requirement FR22 (select most appropriate process elements) as dened above may be decomposed with DP22 (production resources selection procedure) in mind as: FR221ZTime production in concert with customer pull rate FR222ZEnsure the selection of right machines (machine assignment) FR223ZEnsure most efcient material movement FR224ZEnsure labor exibility in the cell FR225ZMinimize indirect labor needs. The corresponding DPs may be stated as: DP221ZDetermined takt time for each family DP222ZSelection of machines according to the information axiom DP223ZSelection of Material Handling System (MHS) based on MHS principles DP224ZAppointing multi-skilled worker to the cell DP225ZEstablishment of self directed work teams. The design matrix for the above set of FRs and DPs are 3 32 3 2 2 DP221 FR221 X 76 DP222 7 6 FR222 7 6 X X 7 76 7 6 6 7 76 7 6 6 76 DP223 7 6 FR223 7 Z 6 X X X 7 76 7 6 6 54 DP224 5 4 FR224 5 4 X X 0 X 0 0 X X X DP225 FR225

(6)

Once again, this is a decoupled design satisfying IA of AD. Appropriate production resources are determined following the development of the master process based on product specications. The cell needs to operate for satisfying customer demand without producing excessive work-in-process inventories and nished goods. In controlling these two measures the cell must operate at the calculated takt time. Once the takt time is at hand, the machines capability and stafng can easily be determined for effective operations. Based on selected machines, it becomes feasible to predetermine on the right material handling equipment. Having the required machines and the material handling equipment, the designer can now determine the worker capability for the line. Afterwards, the designer builds the decentralized organization based on self-directed teams in order to reduce the needs of indirect labors. The functional requirement FR23 (determine required training/education needs) as dened above may be decomposed with DP23 (multi-purpose labor training programs) in mind as: FR231ZDevelop waste elimination focus FR232ZDevelop multi-skilled workers. The corresponding DPs may be stated as: DP231ZLean Manufacturing training and education program DP232ZCross-training program.

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

775

The design matrix for the above set of FRs and DPs are " # " #" # FR231 DP231 X Z X X FR232 DP232

(7)

Once again, this is a decoupled design satisfying IA of AD. Once the resource selection is complete, the education and training requirements of the workers can be established. For ensuring the full participation of workers in the education and training activities followed by transition to cellular manufacturing, appropriate gain sharing programs must be established and announced to the workers for strong buy in. At this stage, internal and/or external resources are recruited to establish the required lean manufacturing training. In addition, labor certication programs are also established for developing multi-skilled workers who can operate all equipment in the cell. The functional requirement FR24 (motivate labor participations) as dened above may be decomposed with DP24 (gain sharing program) in mind as: FR241ZEnsure workers participation for 5S activities FR242ZEnsure worker participation in additional lean manufacturing activities. The corresponding DPs may be stated as: DP241Z5S Reward system DP242ZKaizen teams reward system. The design matrix for the above set of FRs and DPs are " # " #" # FR241 DP241 X Z X X FR242 DP242

(8)

Once again, this is a decoupled design satisfying IA of AD. One of the major reasons for failure in the lean journey is the lack of institutionalization of 5S activities. Inadequate 5S applications always lead into bigger problems during this transition. One way to insure the success of lean transition is to ensure the workers participation in 5S activities. This may be accomplished by preparing appropriate 5S score sheet. With this score sheet, cells with higher scores are awarded additional benets and the winning teams are announced company-wide. In addition, appropriate reward system must also be established to recognize the workers who earn new certication for additional operations in the cell. Kaizen teams are established to insure the worker participation in other lean manufacturing activities. With the established gain sharing programs workers are motivated to participate in these activities. 4.3. Resource rearrangements branch At this stage lean manufacturing principles are the guiding principles of this design step. In this step, the focus is the waste elimination. Therefore, in rearranging the resources waste due to motion, material handling and imbalances between resources is minimized. Without this step the designed cell will not provide the expected performance.

776

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

The functional requirement FR3 (rearrange resources to minimize waste) as dened above may be decomposed with DP3 (product oriented layout) in mind as: FR31ZMinimize material handling FR32ZEliminate wasted motion of operators FR33ZMinimize waste due to imbalance in the cell. The corresponding DPs may be stated as: DP31ZMaterial ow oriented layout DP32ZArrangement of stations to facilitate operator tasks DP33ZBalanced resources in response to takt time The design matrix for the above set of FRs and DPs are 2 3 2 32 3 FR31 DP31 X 4 FR32 5 Z 4 X X 54 DP32 5 X X X FR33 DP33

(9)

Once again, this is a decoupled design satisfying IA of AD. For many products, material-handling cost accounts 3075% of total production cost (Irani, 1999). It is imperative to locate cells with heavy trafc close together. This also helps reducing stocks as stock points (Garza & Smunt, 1994). Furthermore, U-shaped cell design allows entry and exit points of a cell close-by thereby minimizing the burden of material handling. Output stocking point of one cell is used as the inbound stocking point for the downstream cell. Locations of workstations are determined which maximizes the in-sequence parts movements (Aneke & Carrie, 1986). The functional requirement FR31 (minimize material handling) as dened above may be decomposed with DP31 (material ow oriented layout) in mind as: FR311ZReduce the travel distances of parts between cells due to the locations of cells FR312ZReduce the parts movements due to the locations of stock points FR313ZReduce the parts movements due to the locations of machines FR314ZReduce the parts movements due to the material handling system. The corresponding DPs may be stated as: DP311ZCells locations which minimize the travel distances of parts between cells DP312ZAppropriate stock points DP313ZMachine locations which maximize the in-sequence parts movements DP314ZAppropriate material handling system. The design matrix for the above set of FRs and DPs are 32 3 2 3 2 DP311 FR311 X 76 DP312 7 6 FR312 7 6 X X 76 7 6 7 6 54 DP313 5 4 FR313 5 Z 4 X X X X X X X DP314 FR314

(10)

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

777

Once again, this is a decoupled design satisfying IA of AD. In rearranging the resources of the cell, the ultimate goal is to minimize non-value added time of operators. This comes from operator movements between stations and operators time spent on equipment setup. In addition, stations designed without serious consideration to human factors and ergonomics plays a signicant role in these losses due to strenuous working conditions and potential work slow down. The functional requirement FR32 (eliminate wasted motion of operators) as dened above may be decomposed with DP32 (arrangement of stations to facilitate operator tasks) in mind as: FR321ZMinimize operator movements between stations FR321ZMinimize operator time during setup FR321ZMinimize operators wasted time in processing. The corresponding DPs may be stated as: DP321ZLabor allocation to minimize walking distance under the constraint of takt time DP322ZStandard tooling, xtures and auxiliary equipments for each cell DP323ZErgonomic interfaces between workers and equipments. The design matrix for the above set of FRs and DPs are 2 FR321 FR323 3 2 X X 0 X 0 32 DP321 DP323 3

6 7 6 4 FR322 5 Z 4 0

76 7 54 DP322 5

(11)

This is an uncoupled design satisfying IA of AD. An ideally balanced cell should utilize all workers, resources and material handling systems to their fullest capacity. In accomplishing this task, the cell designer must insure that each machines cycle time, operator cycle time and the material handling cycle time are as near to the takt time as possible. This will insure a smooth mixed/leveled production satisfying customer demand in most efcient way. The functional requirement FR33 (minimize waste due to imbalance in the system) as dened above may be decomposed with DP33 (balanced resources in response to takt time) in mind as: FR331ZEnsure that the longest machine cycle time % takt time FR332ZEnsure that each operator cycle time % takt time (Operator cycle time is preferably close to the takt time) FR333ZEnsure that automated material handling systems cycle time % takt time. The corresponding DPs may be stated as: DP331ZFitness control of operations assignment of each machine DP332ZFitness control of tasks assignment of each operator DP333ZSpeed adjustment of material handling system.

778

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

This is an uncoupled design satisfying IA of AD. 4.4. Production control branch

The design matrix for the above set of FRs and DPs are 2 3 2 32 3 FR331 X DP331 4 FR332 5 Z 4 0 X 54 DP332 5 FR333 0 0 X DP333

(12)

Satisfying customers by right amount and just-in-time production can only be accomplished through pull system (Hopp & Spearman, 2000). However, just-in-time systems require a steady pull on all products in the family. In order to ensure a steady pull, a leveled/mixed production schedule must be established. This leads us into developing the appropriate Heijunka schedule and the necessary visual management tools including Kanban system for successful implementation. The functional requirement FR4 (Provide production based on customer demand) as dened above may be decomposed with DP4 (Pull production control system) in mind as: FR41ZEnsure smooth and steady production in assembly line FR42ZProvide continuous material/information ow FR43ZProvide continuous feedback information ow. The corresponding DPs may be stated as: DP41ZLeveled/Mixed production DP42ZKanban System DP43ZInformation/Report System and visual management tools. The design matrix for the above set of FRs and DPs are 32 3 2 3 2 DP41 FR41 X 54 DP42 5 4 FR42 5 Z 4 X X X X X DP43 FR43

(13)

This is a decoupled design satisfying IA of AD. The functional requirement FR41 (Ensure smooth and steady production in assembly line) as dened above may be decomposed with DP41 (Leveled /Mixed production) in mind as: FR411ZAcquire information on desired product mix/volume FR412ZEnsure small batch production FR413ZDevelop smooth assembly sequence. The corresponding DPs may be stated as: DP411ZContinuous information ow from customers DP412ZMinimized setup times (SMED) DP413ZSelected and improved Method for Assembly Sequence.

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

779

The design matrix for the above set of FRs and DPs are 2 FR411 FR413 3 2 X X X X X 32 DP411 DP413 3

7 6 6 4 FR412 5 Z 4 X

7 76 54 DP412 5

(14)

This is a decoupled design satisfying IA of AD. The ultimate goal of cellular manufacturing (and lean manufacturing) is to convert the make-to-stock production system (push production) to make-to-order production system (pull production). In pull production material is pulled through the system starting from the customers all the way to the suppliers. Therefore, appropriate pull systems need to be established for assembly, within-cell material movement, between cells material movement and supplier material movement in the plant. Depending upon the conditions of these different plant zones, proper pull mechanisms (Kanban systems) need to be selected. The functional requirement FR42 (Provide continuous material/information ow) as dened above may be decomposed with DP42 (Kanban System) in mind as: FR421ZEnsure information /material ow between cells and assembly FR422ZEnsure information /material ow within each cell FR423ZEnsure information /material ow between the supplier and manufacturer. The corresponding DPs may be stated as: DP421ZWithdrawal Kanban System DP422ZProduction Kanban System DP423ZSupplier Kanban System. The design matrix for the above set of FRs and DPs are 2 FR421 FR423 3 2 X X X X X 32 DP421 DP423 3

7 6 6 4 FR422 5 Z 4 X

7 76 54 DP422 5

(15)

This is a decoupled design satisfying IA of AD. In order to smooth production, mixed model production schedule at nal assembly cell or pacemaker cell is prepared. According to this schedule, the parts needed by the cell have to be obtained in necessary quantities and at necessary times through the supplier cells. Information also has to be ready in needed cells and at needed times. The functional requirement FR421 (Ensure information /material ow between cells and assembly) as dened above may be decomposed with DP421 (Withdrawal Kanban System) in mind as: FR4211ZTransport consistent quantities between cells and assembly FR4212ZEnsure timely deliveries.

780

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

The corresponding DPs may be stated as: DP4211ZStandard container size from cells to assembly line DP4212ZOptimal number of withdrawal Kanban. The design matrix for the above set of FRs and DPs are      FR4211 DP4211 X Z X X DP4212 FR4212

(16)

This is a decoupled design satisfying IA of AD. In order to obtain needed materials and information ows in CMS, a Kanban system is required. Production Kanban is an order to launch production within cell. On the other hand, production Kanban is activated by withdrawal Kanban. Instead of classical Kanban, decouplers, chutes, conveyor, Kanban square or operator are used within cell. The appropriate one(s) among them is necessary to be chosen. The selection criteria can be determined as simplicity and cost. The functional requirement FR422 (Ensure information/material ow within each cell) as dened above may be decomposed with DP422 (Production Kanban System) in mind as: FR4221ZTransport consistent quantities within the cell FR4222ZEnsure timely deliveries between cell operations. The corresponding DPs may be stated as: DP4221ZStandard container size and one piece ow within the cell DP4222ZAppropriate arrangement within cell (decouplers, chutes, conveyors, Kanban squares). The design matrix for the above set of FRs and DPs are      FR4221 DP4221 X Z X X DP4222 FR4222

(17)

This is a decoupled design satisfying IA of AD. Supplier reliability becomes a very important issue in JIT manufacturing. Once the system operates in pull environment, it is natural to expect the suppliers to also become JIT suppliers. Inadequate lot sizes and/or inadequate deliveries from the suppliers lead into sub-optimal plant performance. Excessive amount of deliveries and large shipments lead into inated raw material stocks and increased lead time without impacting the plant throughput. On the other hand, infrequent and or unreliable deliveries and smaller than requested batch sizes, lead into lost productivity and starvation of some of the equipment within the plant. Therefore, appropriate procedures for supplier Kanban calculations and supplier container size determination must be implemented at this stage. Some analytical techniques exist for estimating these two parameters. The functional requirement FR423 (Ensure information /material ow between the supplier and manufacturer) as dened above may be decomposed with DP423 (Supplier Kanban) in mind as: FR4231ZTransport consistent quantities from the supplier FR4232ZEnsure timely deliveries from the supplier.

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

781

The corresponding DPs may be stated as: DP4231ZStandard container size from the supplier DP4232ZOptimal number of supplier Kanbans. The design matrix for the above set of FRs and DPs are      FR4231 DP4231 X Z X X DP4232 FR4232

(18)

This is a decoupled design satisfying IA of AD. Visual management tools are an important part of a lean manufacturing. They play a signicant role in the success of the lean journey. Therefore, well developed visual management and feedback system will help signicantly in improving the newly designed cells and become an important part of the feedback mechanism as proposed in Fig. 1. These visual tools are simple, yet signicant information tools that help workers to assess the system performance and conditions at a glance. Some visual management tools include reports on production performance, quality performance and TPM related visual tools including OEE measures, to name a few. There are many specic visual tools available in each category listed above. The design team must select an appropriate set of visual management tools and their update frequencies to support the cellular manufacturing activities. Additional visual management tools may be added to the system during the feedback cycle. The functional requirement FR43 (Provide continuous feedback information ow) as dened above may be decomposed with DP43 (Information/Report System and visual management tools) in mind as: FR431ZProvide means to improve system FR432ZEnsure freshness of data. The corresponding DPs may be stated as: DP431ZData system DP432ZData acquisition schedule. The design matrix for the above set of FRs and DPs are      FR431 DP431 X Z X X DP432 FR432 This is a decoupled design satisfying IA of AD.

(19)

5. Implementation and results A manufacturing company producing aluminum walkways, bridges, stairs and ramps in USA was focused on CMS implementation for customized production. The proposed methodology was implemented step by step in transforming its system from the existing classical manufacturing to a CMS. Each stage of implementation based on proposed methodology is summarized in the following sections.

782

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

5.1. Preliminary design stage Before implementation, the facility layout of the system was functionally organized as shown Fig. 3. Following the project team selection, plans have been devised for company-wide participation in the transition to lean and cellular manufacturing. Around this time, 6-day lean manufacturing and TPM workshops were provided to 80% of the factory personnel. 5.2. Classication and grouping products/parts and machines stage (FR1-DP1) The rst task was to dene the product families. Based on the machine-part incidence matrix the team decided that stairs, ramp and landings were ideal candidates for cellular manufacturing. Ensuing economic analysis and justication showed that each cell was economically feasible and desirable. Due to large sizes of the incoming aluminum raw material stock, it was also economically more acceptable to cut incoming stock to smaller sizes later to be resized at each cell by using the 22 00 saw as a common machine at the entrance to the facility. 5.3. Resources capabilities development stage (FR2-DP2) After determining the desired takt time for each cell, the team proceeded with the selection of the process elements. It is at this stage the team concluded that each cell should have its own saw, which is much smaller than the large saw that cater to all cells. Furthermore proper allocation of welding machines have also been established for each cell and several new power outlets needed to be installed to service these welding machines. In order to maintain a smooth product ow, each cell received the appropriate number of punches, milling machines and bufng equipment. The educational and training needs of the workers, supervisors and team leaders were established around this time. This was followed by a series of lean manufacturing and TPM workshops for supervisors, team leaders and majority of the workforce. These team leaders started all out announcement of lean journey, cellular manufacturing and the importance of 5S activities plant-wide. For maintaining the 5S discipline, 5S evaluation worksheets were developed and routinely implemented as a prelude for the gain sharing program. Once the standard work in each cell was dened, proper metrics were developed for continuous improvement monitoring. Based on these metrics, appropriate gain sharing program was developed. This gain sharing program was widely publicized throughout the plant and a copy of this program was made available to all plant personnel. 5.4. Resource rearrangements stage (FR3-DP3) With the participation of cell teams nal acquisition in each cell was determined. Machine locations were xed to maximize the in-sequence parts movements. Fig. 4 provides the new layout of the facility based on three newly formed cells. Stafng of the cell was nalized based on the current cell takt time requirements. Tasks were assigned to workers to minimize idle times. Worker assignments according to the Takt time for ramp cell as an example is shown in Table 1.

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

783

Fig. 3. Old layout before transformation.

784

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

Fig. 4. New cellular layout after the transformation.

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787 Table 1 Worker assignment for ramp cell according to the Takt time Ramp cell Mini cells Vertical Mounting channel Other parts Assembly Operations Milling, drilling Welding, cleaning Punching, bufng Cutting, punching, bufng, bending Welding, cleaning Cycle time (s) 300 840 360 Max. 1200 1620 Takt time: 438 s Number of workers 1 2 1 3 4

785

5.5. Production control stage (FR4-DP4) Based on rm demand and forecasted sales, production volume and eventually mixed leveled assembly plans were established for each cell. Based on these leveled production plan optimal container sizes and number of containers have been established for each cell. For Kanban implementation, new carts have been built to transport batches between stations and also act as Kanban signals. Similarly, the Kanban operations between the cells and the 22 00 saw supermarket were established through transportation carts built in-house. Once again, the sizes of each cart were determined through proper analytical tools. Before the start of the implementation of cellular manufacturing, material suppliers were not very reliable and usually promised once a week deliveries. Usually the lowest cost suppliers had won the bids without paying attention to their performance both on delivery frequencies and reliability. Through new vendor certication and rating system, all potential suppliers were evaluated. Several visits and conferences were held with current and potential vendors. The need for just-in-time deliveries and reliability were emphasized in those meetings. After proper negotiations, nal vendors selected promised daily deliveries with smaller batch sizes responding to companys supplier Kanban signals. In order to maintain the overall equipment effectiveness, total quality control and just-in-time production, appropriate information feedback system and visual management tools were developed. Through posting of this information at suitable places in each cell, workers were transformed into selfmonitoring and continuous improvement teams. With the availability of gain sharing system details, they were able to evaluate the impact of improvements to their pocketbooks. The rst stage of lean transformation started in May 2000. Table 2 provides the comparative measurements of important business metrics as of May 2000 and as of December 2001, for the Ramp
Table 2 Comparison of business metrics Performance criteria Raw material stock (days of inventory) Lead time (days) Scrap rate (%) Throughput (units pairs) Overtime (hours/week) WIP (days inventory) Material move distances (m) Before cellular manufacturing 11 18 3 50 300 6 67 After cellular manufacturing 3.5 7 1.6 70 60 2.5 31

786

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

Cell. The travel distance of parts from 67 to 34 m with cellular system design was reduced. This is a 54% improvement. Similarly, the manufacturing lead time has been reduced by 60% to 7 days from its original 18 days. As expected, we have achieved a solid 58% reduction in WIP. The most striking improvement has been achieved in overtime reduction. An 80% reduction in this metric was achieved. One of the primary reasons for this improvement is explained by our new design elimination the 22 00 saw and allocation smaller dedicated saws to each cell.

6. Conclusions In this paper we provide a complete and concise methodology for transforming a process oriented manufacturing facility into a CMS. The methodology is based on Axiomatic Design principles. The proposed process is also implemented at a company that manufactures aluminum ramp rails, landings and stairwells. The results show that the proposed methodology is sound, easy to follow and implement. Details of the feedback mechanism for continuous improvement of the cellular system under the guidelines of AD principles will be presented at our subsequent publications.

References
Aneke, N. A. G., & Carrie, A. S. (1986). A design technique for the layout of multi-product owlines. International Journal of Production Research, 24(3), 471481. Babic, B. (1999). Axiomatic design of exible manufacturing systems. International Journal of Production Research, 37(5), 11591173. Black, J. T., & Schroer, B. J. (1988). Decouplers in integrated cellular manufacturing systems. Transactions of the ASME Journal of Engineering for Industry . Burbridge, J. L. (1969). An introduction to group technology. Proceedings of an international seminar on group technology (pp. 129), Turin. Chen, S. J., Chen, L. C., & Lin, L. (2000). Knowledge-based support for simulation analysis of manufacturing cells. Computers in Industry , 44334449. Choi, M. (1992). Manufacturing cells design. Production and Inventory Management Journal, 2, 6669. Cochran, D. S., Eversheim, W., Kubin, G., & Sesterhenn, M. L. (2000). The application of AD and lean management principles in the scope of production system segmentation. International Journal of Production Research, 38(6), 11591173. Cochran, D. S., Kim, Y. S., Kim, J. (2000). The alignment of performance measurement with the manufacturing system design. Proceedings of ICAD2000, rst international conference on axiomatic design, Cambridge, MA (pp. 115122). Garza, O., & Smunt, T. L. (1994). Reducing ow between manufacturing cells: A sensitivity analysis. International Journal of Production Research, 32(9), 21312147. Groover, M. (1987). Automation, production systems and computer integrated manufacturing pp. 436437. Gunasekera, J. S., & Ali, A. F. (1995). A three-step approach to designing a metal-forming process. JOM , 2225. Heragu, S. S., & Chen, J. (1998). Optimal solution of cellular manufacturing system design: Benders decomposition approach. European Journal of Operational Research, 107, 175192. Hopp, W. J. & Spearman, M. L. (2000). Factory Physics, Foundations of Manufacturing Management, McGraw-Hill. Houshmand, M., Jamshidnezhad, B. (2002). Conceptual design of lean production systems through an axiomatic design. Proceedings of ICAD2002, second international conference on axiomatic design, Cambridge, MA (CD-Disc). Hyer, N. L., & Wemmerlow, U. (1987). Group technology oriented coding systems: Structures, applications, and implementation, capabilities of group technology. In N. L. Hyer (Ed.), The computer and automated systems association of SME, 6784.

O. Kulak et al. / Computers & Industrial Engineering 48 (2005) 765787

787

Irani, S. A. (1999). Handbook of cellular manufacturing systems. New York: Willey. Kim, S. J., Suh, N. P., & Kim, S. (1991). Design of software systems based on AD. Robotics and Computer-Integrated Manufacturing, 8(4), 243255. Nancy, H., & Wemmerlov, U. (2002). Reorganizing the factory. Productivity Press pp. 95145. Sarher, B. R., & Mondal, S. (1999). Grouping efciency measures in cellular manufacturing: A survey and critical review. International Journal of Production Research, 37(2), 285314. Silveira, G. (1999). A methodology of implementation of cellular manufacturing. International Journal of Production Research, 37(2), 467479. Singh, N., & Rajamani, D. (1996). Cellular manufacturing systems: Design, planning and control. London: Chapman & Hall pp. 34180. Soanopoulou, S. (1999). Manufacturing cells design with alternative process plans and/or replicate machines. International Journal of Production Research, 37(3), 707720. Soleymanpour, M., Vrat, P., & Shankar, R. (2002). A transiently chaotic neural network approach to the design of cellular manufacturing. International Journal of Production Research, 40(10), 22252244. Steudel, H. J., & Desruelle, P. (1992). Manufacturing in the nineties: How to become a mean, lean, world-class competitor. New York: Van Nostrand Reinhold. Suh, N. P. (1990). The principles of design. New York: Oxford University Press. Suh, N. P. (1995). Designing-in of quality through axiomatic design. IEEE Transactions on Reliability, 44(2), 256264. Suh, N. P. (1997). Design of systems. Annals of the CIRP, 46(1), 7580. Suh, N. P. (2001). Axiomatic design: Advances and applications. New York: Oxford University Press. Suh, N. P., Cochran, D. S., & Paulo, C. L. (1998). Manufacturing system design. Annals of the CIRP, 47(2), 627639. Sule, D. R. (1991). Machine capacity planning in group technology. International Journal of Production Research, 29(9), 19042147. Tseng, M. M., & Jiao, J. (1997). A module identication approach to the electrical design of electronic products by clustering analysis of the design matrix. Computers and Industrial Engineering, 33(1), 229233. Wemmerlov, U. (1992). Fundamental insights into part family scheduling: The single machine case. Decision Sciences, 23, 565595. Zhao, C., & Wu, Z. (2000). A genetic algorithm for manufacturing cell formation with multiple routes and multiple objectives. International Journal of Production Research, 38(2), 385395.