Printer Friendly

Production flow analysis for planning group technology.


Group technology (GT) is a method of factory organization in which organizational units known as groups complete all the products or parts they make and are equipped with all the processing facilities they need to do so.

A recent (20 years) synonym for GT is "cellular production." I do not use this term myself, because may research is done mainly in factories and telling workers that they are going to be put into cells (prison) is likely to be counterproductive. This word "cell" is mainly used today to describe a part of a group or sub-group.

GT is used in batch and jobbing production, leaving "continuous line flow" (CLF) as the preferred method of organization for the simple process industries and for high volume mass production. Where continuous line flow cannot be used, and where companies are not so small--say 20 direct workers of fewer--that they are in effect already single groups, it is generally possible to find a total division of any factory into GT "groups" of machines and associated "families" of parts, with no backflow and no crossflow between groups. Organized in this way, companies will normally be much more productive and profitable than is ever possible with the traditional process organization. Figure 1 shows the effect of this change on the material flow system. Figure 1A is typical of the flow found in batch or jobbing production factories, using traditional "process organization" with organizational units which specialize in different processes. Because different parts use different combinations of processes in different sequences, a process organization always produces very complex material flow systems. Figure 1B shows the material flow system in the same factory after changing to GT. Using PFA, I and close associates have now found a total division into groups and families in 33 different factories with no backflow and no crossflow between groups. Our findings have convinced me that it is extremely unlikely that we will ever find a case where product organization (CLF or GT) could not be substituted for process organization. Because this change is always possible and is always profitable, it is submitted that process organization is obsolete (Burbidge (forthcoming)).

Most of the early examples of group technology were based on the classification and coding (C & C) of component drawings using either the Brisch or Opitz C & C systems. As will be shown later, this is an inefficient method for planning GT groups and families. It has been replaced in recent years by production flow analysis (PFA) (Burbidge (1989)) which finds the division into groups and families by analyzing the information in the "process routes" for parts.


The term "group technology" was first used by Professor Mitrofanov of Leningrad University as the title for his research into the relationship between component shape and processing methods. One of his findings was that it is possible to set up a lathe to make a "group" of similar parts one after the other at the same setup (Mitrofanov (1946)).

Later, an engineering company in Alsace took a section of lathes set up in the Mitrofanov mode and added two milling machines and two drilling machines to form a "group" which completed all the parts it made (Sidders (1962)). At this stage, the meaning of the word "group" changed from "a set of parts" to a "set of machines," and a new word "family" (famille des pieces) was introduced to describe the set of parts.

Work in England--notably by Graham Edwards at UMIST--(Edwards (1968)); Joseph Gombinski of EG Brisch and partners (Gombinski (1964)); Gordon Ranson (Ranson and Toms (1966)); and C. Allen of Ferranti, Edinburgh, showed that the same principles also applied to other processes, and that a total division of bath and jobbing processing departments into groups was generally possible with ony a few easily accommodated exceptions. The effect of this change on the material flow system in a factory has already been illustrated in Figure 1.

This view of group technology as a method of organization of general utility in batch and jobbing production was confirmed by two international conferences run by the Turin International Center at Turin in Italy (Burbidge (1970) and Burbidge (1976)). The idea of GT had a strong following worldwide in the 1960s and 70s. A world survey (Burbidge (1975)) found 59 companies in 13 countries which had completed the change, out of a total of 483 companies in 32 countries which reported that they had started and not yet finished implementation.

The advantages of GT in comparison with traditional process organization are tabulated in Figure 2. From experience in 33 companies mentioned earlier, I believe that these advantages, while they are only possible with GT, are not automatically achieved. One needs to recognize the new possibilities and then take the necessary steps to achieve them.

GT went into a decline in Britain at the beginning of the 1980s from which it has only recently started to recover. The decline was mainly due to difficulties in planning the division into groups and families for GT but also in part to difficulties in implementation. The recovery came partly with the adoption of production flow analysis (PFA) for planning GT groups and families, partly with the realization that GT is the essential starting point for the automation of batch and jobbing production, and partly from the success of Japan in manufacturing, which has been largely due to the Japanese use of simple material flow systems like those based on GT.

The term group technology has been mainly associated with component processing. Very similar groups are also used for assembly. Assembly groups complete assembled products, or major stages of assembly in the case of large complex products. They are based on the objective of minimum throughput time and normally make most of their own subassemblies. As the design of assembly groups is based mainly on this one simple objective and does not require the use of "production flow analysis," it need not be considered in the present paper.


Most of the early applications of GT were based on classification and coding (C & C) of the information in component drawings. It was reasoned that parts that were similar in shape or function could be made by the same set of processes and could therefore be made in the same group (Gombinski (1964)). While this is true for the majority of parts, there are too many exceptions for reliability. Parts may, for example, be similar in shape, but have to be made in different groups because they differ greatly in size, tolerances, requirements quantities or materials.

Another problem with C & C is that it does not bring together the many parts which differ greatly in shape and function, but ought to be made in the same group because they can only be made on the same limited set of machines. Mainly for this reason, C & C only find families of parts. Because it only looks at product design, it gives no help at all in finding the groups of machines which will be needed to make the parts.

Classification and coding (C & C) may have value in some companies as a route towards standardization or as a technique for reducing drawing office or tooling costs. It is, however, a very inefficient method for finding groups and families for GT. Even if a company already has a C & C system, it would be unwise to use it for planning GT, because it does not find a complete division into groups.


Production flow analysis (PFA) (Burbidge (1989)) is a technique for finding both GT groups and their associated "families" by analyzing the information in component process routes which show the operations needed to make each part and the machines to be used for each operation.

Where C & C says: "It is probable that parts which are similar in shape or function can be made by the same group (set) of machines," PFA takes a more direct approach, saying: "Parts which are made using the same set of machines can be made in the same group." Another difference is that PFA is a technique for simplifying material flow systems. C & C ignores this aspect of the problem. PFA consists of five sub-techniques used progressively to simplify that material flow system in an enterprise. Three are illustrated in Figure 3.

Company flow analysis (CFA). The first analyzes the existing flow of materials between the different factories in a large company and develops a new, simpler and therefore more efficient system in which each factory completes all the parts it makes.

Factory Flow Analysis (FFA). The second studies each factory in turn. It plans the division of the factory into major groups or departments each of which completes all the parts it makes, and it plans a simple unidirectional flow system joining these departments.

Group analysis (GA). The third uses matrix resolution to divide each department in turn into groups, each of which completes all the parts it makes. Providing that one starts with departments which complete parts, GA can, inside certain limits of group size, and with very few exceptional parts, always find groups which complete parts, with no backflow, no crossflow (between groups) and no need to buy any additional equipment.

Line analysis (LA). The fourth analyses the flow of materials between the machines in each group to find the information needed for plant layout.

Tooling analysis (TA). The final technique returns to matrix resolution--in this case matrices of parts and the tools they use. It studies each machine in each group in turn, in order to find "tooling families" of parts which can all be made on the machine with the same set of tools at the same setup and also to find the sequence of loading which will minimize setup times.

PFA is normally done on the computer. As the main requirement is for sorting and listing, the software needed is relatively simple.


There is not sufficient space here to describe all the PFA techniques in detail. FFA and GA are the main techniques used to find groups and families. FFA will be examined first. One starts FFA by recording the existing flow system, as illustrated for a simple case in Figure 4. The first step is to choose a code for the processes. Each part is then given a "process route number" (PRN), showing the codes for the processes used to make it in usage sequences, and a PRN Frequency Chart is drawn showing the number of different parts with each PRN number. A FROM/TO chart is drawn to show the number of parts which move between each pair of processes in both directions, as well as the totals for each row (from) and column (into). This chart also shows the number of parts which "start" and "end" with each process, and also a checking total for each process. Finally, a network diagram is drawn showing all the processes and all the movements between them.

Simplification of this system starts by finding the "primary network" or the network which includes the smallest number of PRNs covering the largest number of parts and which includes some flow into and/or out of every process. The next step is to simplify the primary network, first by combining closely associated processes--or processes shown in the FROM/TO chart as often used together to make parts--to form departments. At the end of this step (known as Method 1), most of the remaining PRNs which were not used to form the primary network can now be handled by the new simple flow system.

The final step, known as Method 2, studies the few exceptional parts which do not fit the system at the end of Method 1. In the mechanical industry, as an example, these exceptions have never yet exceeds 10% of the parts. A typical example might find four departments at the end of Method 1:

1. Cold forge and welding.

2. Machining.

3. Finishing (painting and electroplate).

4. Assembly.

Analysis of the exceptional PRN's which do not fit the new simple material flow system might find some which go from "cold forge and welding" to the "machine shop" for intermediate milling and drilling operations, and then back to the cold forge for further operations. These interdepartmental moves can be eliminated by transferring one or two machines to the cold forge department from the machine shop. Some parts may be found which go outside for intermediate subcontract operations. As far as possible, all such operations should be brought back into the factory. Some parts may be found which go from machining to assembly for intermediate fitting operations and then back again. These transfers can be eliminated by training machinists to do the work. It is generally possible to eliminate all exceptional transers involving backflow, by minor re-deployment of machines and operations in this manner, or, as a last resort, by buying a part instead of making it. An example of FFA (before and after) is shown in Figure 5.


Providing FFA has found departments which complete all the parts they make, GA can always divide them into groups of machines and families of parts, in such a way that each group completes all the parts it makes. As shown in Figure 6, this task can be described as the resolution of a matrix.

In most companies the part/machine matrix is much too large for manual resolution and--for reasons concerned with the nature of the data--it is also very difficult to solve the problem mathematically even with the help of a computer. The solution used with GA is to divide the total matrix into a smaller number of sub-matrices known as "Modules," in such a way that groups can be formed by combining modules.

a. Forming Modules

Each module is a part/machine matrix based on a "key" machine from the plant list (i.e., list of machine tools). Each module includes all the remaining parts--not already included in previous modules--which have operations on the key machine and shows all the other machines used with it to make these parts. An example is shown in Figure 7. The top portion of the figure gives information about the machine tools used, the middle portion shows the module matrix, and the bottom portion gives additional information about the parts.

To find modules which can be combined to form groups, the solution lies in the sequence in which machines are selected as key machines from the plant list. To find this sequence, the machines are first classified into:

S = Special category. There is only one of each type, and it would be very difficult to transfer the work it does onto any other machine type (e.g., bar lathe, crankshaft grinding machine, gear tooth rounding machine, etc.)

I = Intermediate category. Same as S, but there is more than one of each type.

C = Common category. There are several of each type and it is easy, if necessary, to transfer the work they do to other related machine types (e.g., lathes, mills and drills in machine shops).

G = General category. There are few machines of each type. They are used for a high proportion of the parts, or for many different types of part. They are unlikely to be suitable for inclusion in groups (e.g., saws for cutting blanks, x-ray machines, painting and electroplating equipment).

E = Equipment category. Items used to assist manual operations (e.g., benches, vices, surface plates and manual power tools).

The plant list is now re-ordered in SICGE sequence with the machines in each code block in ascending order of Big F, where Big F is the number of different parts with operations on a machine type. This re-ordered plant list--known as the "special plant list" or "SPL"--indicates the sequence in which the machines must be selected as key machines for forming modules. Modules are then formed by programming the computer to:

1. Select key machines in turn from the SPL.

2. Find all the remaining parts with operations on the key machines.

3. Find all the machine types used to make the parts.

4. Print out the module.

The modules are formed with a successively reduced data base. Each module removes at least one (the key) machine and a number of parts from the data base. Each part is listed in one module only. Machines may be included in several different modules. The modules are given the same identifier as their key machine, to facilitate later analysis. When all parts and machines have been included in modules, a "Module Summary" is produced. This is illustrated in Figure 8, with data taken from an actual firm. The individual "modules" and the "Module Summary" provide most of the information needed to select groups and to assin the modules to these groups.

b. Number of Groups

The first step in choosing the groups is to decide how many groups one wishes to form. A scale effect chart is shown in Table 1. It starts with the present number of workers, machines, and parts in a department, and then gives the average numbers for different numbers of groups. It provides a rough guide to the number of groups needed. With the present generation of mixed manually operated and semi-automatic machines, the author prefers groups in the range of 12 to 15 workers with perhaps one or two larger groups for complex parts. These groups are a convenient size for supervision, and it has been found in practice that groups of this size can be formed with the existing mchines without having to buy any new equipment. This size of group is tending to fall with increases in the number of NC machines. The size of group depends partly on the complexity of the product. If the most complex part requires operations on say 25 machine tools, there must be at least one group containing 25 machines. With simpler products, groups can be much smaller.
No. of Group Workers Machines Parts
 1 200 260 1020
 5 40 52 204
 10 20 26 102
 11 18 24 93
 13 15 20 78
 14 14 19 73
 15 13 17 68
 16 12 16 64
 17 12 15 60

c. Choosing Groups

In selecting groups, the main objectives are, first, to find a total division of the machines in a department into "groups," and of the parts it makes into associated "familes"; and second, to form groups which complete all the parts they make without backflow or crossflow between groups. Few of the advantages listed in Figure 2 are obtained if one fails to achieve these objectives.

There are four main rules to remember when choosing groups:

1. There is only one of each "S" type machine, so they can only be in one group each. If the same "S" type machine is found in two or more different modules, they must be combined.

2. If the module summary shows a number of modules which use mainly the same machines, they will probably be together in the same group.

3. Although there are exceptions, most parts which are similar in shape or function will probably be made by the same set of machines.

4. In any mechanical product, about 50% of the parts are simple items made on a few "C" class machines only.

Looking at the module summary in Figure 8, for example, and starting with Rule 1, it was found that:

Group 1. Modules AA, AC, AF, AH and AL combined to form a group nucleus which included all parts made on "S" class machines: AA, AC, AF, AH and AL. Although this group made some other parts, its main product was large gears and it was called the "Large Gear Group."

Group 2. Module AE (S class) combined with Module BG, to bring together all parts (gear boxes) made on machines AE (S class) and BH and BG (C class) to form a Gearbox Group nucleus. This group is now running with an unstaffed night shift.

Group 3. Modules AL and AK (S class) and AN, AR and AT (I class) brought together most parts made on machines AI, AK, AN, AQ and AU to form a strong nucleus for a Small Gear Group.

At this stage, all "S" class machines have been found a home, except AB, AD, AG and AJ. As these machines have no links with other S machines, they are simple to place. There are some obvious candidates among the "I" class modules for adding to Group 3 (e.g., AR, AT, AV, AW and AX), but it was decided not to add them to the nucleus at that stage, but to leave them for the later consolidation stage.

It will be seen that some machine types will be needed in more than one group. There are, for example, already one or two cases where "I" and "C" class machines are going to be needed in both Group 1 and Group 3. These will be reconsidered at a later "reallocation of operations" stage. Another three simple groups were then formed at the pointed end of the Module Summary (Figure 8; BG and following modules). The choice in this case was influenced by the materials used and by the types of part produced. This information came from the module printouts (Figure 7).

At this stage the Module Summary (Figure 8) was marked up with a highlight pen to show which modules had already been allocated to nuclei. Working with a reduced number of unallocated modules, three more group nuclei were quickly found, using Rule 2, and the remaining modules were then added in the consolidation stage to groups which used mainly the same machines. A perfect fit is seldom found. In selecting the "best fit," it will often be necessary to add a module which contains one or more additional C class machines, not previously used in a group, to that group.

d. Reallocation of Operations to Machines

An example of reallocation of operations is illustrated in Figure 9, using data from another company. When all modules had been allocated to groups, a "Group Summary" was produced in the form of a matrix, with group numbers as column heads and machine type numbers identifying the lines. The symbols used are:

F = Total number of parts with operations on a machine type (also called Big F).

f = Number of parts in a module with operations on a machine type.

[Sigma] f = Number of parts in a group with operations on a machine type.

N = Number of machines of each type.

Cd = SICGE category.

Unlike the Module summary (Figure 8), where the machines are listed in "special plant list" sequence, the machines in the "group summary" are grouped by general type, for example: lathes, borers, mills, etc., as shown in Figure 9A. The machine tools were allocated to the groups in proportion to the number of machines of each type (N) and to the number of parts ([Sigma]f) using the machine type in each group. Any values of "[Sigma]f" which are too small to warrant the allocation of a machine are put in parentheses to indicate that they are exceptions (Figure 9A).

A meeting of the company process planning engineers then considered the group summary and attempted to reallocate the exceptional operations to similar types of machine which were already in the groups. For example, in Group 1, 23 exceptions on machines 010105, 010119, 010123, and 010125 (see Figure 9A) were all reallocated to 010126, bringing its total for [Sigma]f up to 26 (see Figure 9B). This type of reallocation will normally eliminate most of the exceptions. The remaining (say) 0.5 to 2% of exceptions are then eliminated by: the further transfer of underloaded machine tool types between groups, by replanning methods, by minor product design changes, by bringing sub-contracted operations back into the factory, or in the last resort, by buying instead of making the part.

It will be realized tat a simple part made in five operations on machine tools of which there are (say) four different sub-types of each in the factory (any of which might be employed for an operation) will have over 1,000 different possible routes. Possible changes in the sequence of operations may increase this figure. It is this enormous variability in the route data which makes a total division into group technology groups possible and which also explains why more elegant mathematical methods such as cluster analysis, rank order clustering, and Boolean algebra have been unsuccessful in practice up to now.

It has been suggested that this change in the allocation of operations to machines, from the choice made by the original planner, represents a downgrading of processing quality. In practice however, the reallocation tends to bring together, on the same machines, operations on parts which are similar in shape. This makes it possible to standardize methods, reduce the variety of tools required, and to greatly reduce setup times.

On one machine in a particular group, there were, as an example, 29 parts with operations on the machine. These operations had been planned over a period of 15 years by 18 different planners, some of whom are now dead. One hundred and fifty-three different lathe tools were used to make these parts. It was found that the parts could be re-planned using only 31 different tools with only four different setup changes. Much as one might dislike the idea of tearing down monuments to the dead, it is obvious in most cases that carefully controlled reallocation will greatly benefit the company.

e. Checking the Balance of Load

There is a high probability that at the same production level of output, the load on the machines after introducing GT will be the same, or less, than it was with process organization. When the groups are known, it is simple to calculate the net load (sum of the operation times) on all the machines. It is very difficult to calculate gross load (including setup time and allowances for down time and other idle time). These adjustments are best treated as capacity reductions and are best found by random observation studies, or by machine monitoring. They will change with the introduction of GT.

The load check after planning GT looks for machine types which are heavily loaded in one group and underloaded in another. In such cases, which are very rare, it may be possible to balance the load by moving simple parts from one group to another.


GT groups have been found in practice to be remarkably stable. Two of the oldest GT applications in Britain--Serck Audco, Newport (valves) and Ferranti of Edinburgh (radar)--have the same groups, plus one additional group each, as they had 26 years ago. With the purchase in recent years of CNC and DNC machines, the groups have changed greatly. They now contain half as many machines and employ half as many people as the original groups but have increased capacity.

The term "flexible manufacturing system" (FMS) is generally reserved for fully automated systems. Groups are in effect, however, FMSs with some remaining manual operations. As machines are introduced with automatic cycles, and as the material transfer between machines is mechanized, the groups gradually move towards full automation. There is some evidence that this evolutionary approach to automation is more profitable than the "great leap forward" approach which has been used with FMSs in the past.


If one uses PFA, it is generally possible to find a complete division of any batch or jobbing production factory into groups which complete all the parts in their associated families of parts, without any backflow or crossflow between groups. These groups are, in effect, flexible manufacturing systems with some manual operations. GT provides, in fact, the foundation for an evolutionary approach to complete automation. Among the other advantages of GT are a reduction in throughput times and also, therefore, in the stock investment, and a better accountability possible when supervisors control all the facilities needed to complete parts and can be made responsible for quality, cost and due-date performance in their groups.


Burbidge, J.L. "The New Approach to Production." Production Engineer, Chesterfield Street, London, December 1961.

Burbidge, J.L. (ed.) "Group Technology." In Proceedings International Seminar. Turin, Italy: Turin International Centre, 1970.

Burbidge, J.L. Final Report of the Study of the Effects of Group Production Methods on the Humanization of Work. Research report. Turin, Italy: Turin International Centre, 1975.

Burbidge, J.L. (ed). "The Effects of Group Production Methods on the Humanization of Work." In Proceedings International Seminar. Turin, Italy: Turin International Centre, 1976.

Burbidge, J.L. Production Flow Analysis. Oxford: Clarendon Press, 1989.

Burbidge, J.L. "Change to GT: Process Organization is Obsolete." International Journal of Production Research (forthcoming).

Edwards, G.A.B. "What is Group Technology?" Conference Paper, PERA Melton Mowbray, November 1968.

Gombinski, J. "Classification for Family Grouping Success." Metalworking Production. London: McGraw-Hill, April 1964.

Mitrofanov, S.P. Scientific Principles of Group Technology. (In Russian.) Leningrad: Leningrad University, 1946.

Ranson, G.M., and Toms, B.H. "Production Planning and Control." Mechanical Engineering EDC, HMSO, London, 1966.

Sidders, P.A. Flow Production of Parts in Small Batches. London: Machinery Publishing Co., 1962.
COPYRIGHT 1991 American Production and Inventory Control Society Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1991 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Special Issue on Group Technology and Cellular Manufacturing
Author:Burbidge, John L.
Publication:Journal of Operations Management
Date:Jan 1, 1991
Next Article:A goal programming approach to the cell formation problem.

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters