You are on page 1of 4

Hybrid Automatic-man ual Assembly Systems

T. K. Lien Department of Production and Quality Engineering, Norwegian University of Science and Technology, Trondheim, Norway Submitted by F.O. Rasch(l), Trondheim, Norway

Abstract Hybrid assembly systems consist of an automatic assembly section that feeds a manual finishing line. They are competitive for medium volume assembly of products with standard basic structures and many variants that require manual finishing. This paper describes two basic principles for the manual assembly: The parallel arrangement of stations, and the sequential arrangement. A theoretical model for the possible output as function of production volume, number of assembly operations, number of product variants, and station operation time imbalance is presented. This model enables prediction of the performance and selection of the optimal system for a given set of conditions. Keywords: Assembly, Flexibility, Productivity

1 INTRODUCTION The assembly process is the most complex process in modern manufacturing. While automation has made most of the part manufacturing very efficient and less dependant on manual labour, the manual processes still are the most used in assembly. This situation is apparently due to the complexity of the products at the final assembly stage. It is also a function of the modern trend towards customer-specific assembly. Most of the manufacturing is tailored to the individual requirements of the customer. Even highly standardised products are offered in a suite of variants that leads to only small batches of similar products at the assembly department. Studies of automation possibilities in assembly have been made more than two decades ago. Nevins et. al. [ I ] demonstrated in a study in 1978 that robotic assembly would be viable for yearly production volumes starting in the range of 100.000 to 200.000. According to this study a flexible robot system would be more efficient than dedicated high output assembly machines for volumes up to 0,5 to 1 million units per year. One would believe that these limits would be drastically reduced over the years due to the dramatic reduction in computer and electronics prices. A recent study performed by this author reveals that there has been a change, but it is less dramatic than expected. Assuming present days costs, it can be shown that robotic assembly is viable for yearly volumes starting in the range 75.000 to 120.000. The reason for this moderate reduction in the starting limit seems to be that the cost of feeder systems has not followed the general trend of cost reduction. For many companies, the yearly production volume for the individual products falls far below this limit. The only possibility for reaching volumes that would seem economical in assembly is to combine products of the same family where differences are moderate. In some cases this is easy. The flexibility of the feeder systems and robots allow such systems to be designed at competitive costs. But a large number of cases show that the individual differences between product variants are so large that only manual assembly is viable for the complete process.

THE COMBINATION AUTO MATlC ASS EM BLY

OF

MANUAL

AND

2.1 Traditional line thinking When combining several similar products into a group to form an assembly family it is quite common to observe that the similarities in structure are large in the first stages of the assembly process. It then becomes less pronounced as the products come closer to completion. Therefore it is fairly easy to set up an automatic assembly system to cover the first part of the assembly process, and let the final assembly operations be done manually. This concept put certain requirements on the structure of the product to assemble. The individual products that shall be assembled can have different components, but they must have similar structures and assembly methods. If the similarity in structure and assembly tasks is sufficiently high, the assembly process can be considered equal for all variants. A redesign of the product is often needed to obtain sufficient similarity. Then automatic assembly can be applied for the first part of the assembly process [2]. To obtain maximum efficiency, the finishing manual assembly operation should be coupled directly to the automatic stages. This will ensure high efficiency. But the drawback of this approach is normally a very tight coupling to machine pacing for the assembly workers. Such pacing is considered to be out of line with modern production philosophy for several reasons:
0

Machine pacing of manual tasks gives low worker satisfaction and thus low productivity. The normal idea of a line with a series of short cycle time manual operation will be highly vulnerable to stoppage because of worker absenteeism. To prevent production loss for the previously mentioned reason, a backup staff has to be maintained. This increases the operational cost.

2.2 Parallel finishing assembly stations The author of this paper proposed an alternative solution to the challenge of combined automatic and manual operations around 1980. The idea was to let each worker

complete the assembly of the product at his station. This leads to a system where one normally needs a number of workers to operate in parallel at the end of the automatic assembly section. By the use modern PLS control technology it is possible to set up a conveyor system that feeds each worker with assembled basic, or core components of the product. These basic components are feed at a rate that is equal to the actual working speed of each individual worker. Thus, machine pacing is eliminated [3]. This would seem to be an inefficient solution since individual workers operational speed will determine the overall output. But experience with the system has shown that it will not necessarily give a lower overall efficiency. It is true that individual worker speed will determine the overall production rate. But provided that the workstations are laid out properly according to ergonomic principles and good assembly technology standards, normal workers will perform just as well as in machine-paced ope rations. The parallel solution provides two other major benefits:
0

1
I

8
utomatic kembly section Buffer storage Second manual sequential stage Buffer storage Third manual sequential stage

Worker absenteeism affects only the workstation of the absentee; it does not stop the line. Thus the production rate is reduced when workers are absent, but production can be maintained even without a backup staff.

The production rate equals the average rate of the workers. In a traditional sequential line, the rate of the slowest worker will determine the overall rate. The structure of a hybrid system with parallel manual assembly is shown in figure 1. In this figure the automatic first section of the line is indicated as 3 stations in sequence. But it can of course be any number of automatic stations combined to feed the manual assembly line following. Each manual station is set up to allow the operator to assemble any of the products that will be run on the system. The number of operations will of course vary with the product. With the parallel set-up this will not cause any problems, since the operator at any time will have the necessary tools and equipment available. This principle does on the other hand reduce the range of product variants that can be assembled on the system without stop for change of equipment set-up.
0

Figure 2: Structure of a system with sequential manual stations One could then argue that this would be more efficient. In particular the effect of learning will be faster since each operator would get more repetitions of a limited number of subtasks than in the parallel system. This concept has however, some limitations for a flexible assembly process:
0

The variability of the products assembly sequence would put restrictions on how many different products that can be run in this serial system The division into subtasks should not be made too fine in order to obtain fully utilised process stages.

2.3 Sequential finishing assembly line One major argument against the parallel arrangement for manual finish assembly is the assumption that it does not provide the same opportunity for worker training as a more specialised arrangement would do. A specialisation would imply that each assembly station would do only a limited number of finishing assembly operations. This would then enable better arrangements for more specific tasks.

The operations cannot be balanced for optimal efficiency because of the variability in products to assemble. Figure 2 shows the structure of a 3-stage sequential manual assembly system. The actual assembly process in a particular case determines the number of stages. At each stage the operators perform similar assembly tasks in parallel. The number of operators is determined by the assembly rate needed for the system. Since there will always be short cycled variations in the operation time between each of the stages, the system needs substantial buffer storage between each stage. The buffers will increase the volume of work in progress, increase throughput time and make a conveying system between the stages more complex. 3 EFFICIENCY CALCULATIONS In order to determine the relative efficiency of these two concepts some figures of merit has to be determined. These figures should not reflect the capability of the individual worker, but they should indicate the systems ability to efficiently utilise the work force effort. On such figure of merit is the inherent time loss, or unproductive time imposed by the system structure.
0

Figure 1: Structure of a system with parallel manual stations

3.1 The theoretical production capacity for parallel assembly stations In a parallel system the following basic parameters are used: M : The number of assembly workers

f p : The time to assemble a product of type p np :The number of products of type p to assemble N : The number of different products to assemble fpmm : The maximum fpamong N products. T : The man-hours required for one worker to assemble all of the products. TM:The man-hours required for Mworkers to assemble all of the products Dw: Length of the working day It is easily seen that:
N

(M;-1) to M;. Assuming that all values of T, are equally probable, the average idle time of the station will be: 1 Li=-(Mi-(Mi-l)).& 2 1 L. = -Dw I 2 The relative loss at station iwill be:

T = c n p .tp
p=l

4= I
2.Mi

(7)

The time TMwill be bonded by:


m m

For a line with S stages, the average relative loss time will be:

Assuming that the assembly system is running the complete working day, the system will be busy up to the time when there is just not enough time to complete one more product for each worker. Assuming that each worker then will stop and not do more that day, the maximum unproductive time for each worker will be bounded by fpmm.. The relative loss will be

<tpmax
Dw

This calculation of loss time assumes that a defined number of products have to be assembled, and the computed M; gives the minimum number of operators required at each stage. To compare the performance of a sequential versus a parallel system, a similar loss figure has to be applied to the parallel system. 3.3 The theoretical loss figure for parallel assembly stations for a given production volume The loss figure for sequential stages given in the previous section is applicable to a system with parallel manual stations as well. In the parallel case, the number of stages, S,is 1, and M; = M. Then the relative loss time for parallel systems, given a defined daily production volume, Ap, is given by:

(3)

p -

The size of the relative loss is therefore dependent on the assembly time of the products at any particular time. No specific figure can be given. Experiences from several systems installed in Norwegian industry show that the manual assembly normally time lie in the region of 10 minutes or less. A typical loss figure for a 7.5-hour working day would then be:

6 <-. l o
-

100%

7.5.60 SP < 2.2%

(4)

/z =- 1 2M
4

(9)

This loss figure does only reflect the loss due to the system structure. It assumes that the assembly is running at full capacity all the time, and that there is no waiting for parts to assemble. Other losses that normally appear in production are not system specific and are as a consequence not considered here. 3.2 The theoretical loss figure for sequential assembly stations for a given production volume In this analysis, the situation at each assembly stage is considered. The following parameters are used: i : Assembly stage number S : The number of stages M; :The number of workers at stage i 6: Man hours required to complete one days batch at stage i T,will be determined by the demand for each day. With a given demand, the required number of operators at each stage will be given by:

M , = i n - +1
Using equation (5)one realises that the need for whole workers dictate that the efficiency always will be less than unity except for the rare cases where the ratio of T, to DW is an integer value. The real capacity need is in the range

{.a)

COMPARISON OF LOSS FIGURE FOR THE TWO SYSTEM ALTERNATIVES The loss figure given by equation (8) is depending on the arrangement of the assembly system. A system with equal, or near equal number of workers on each stage has the best loss figure. Large differences in the number of workers between the stages give the poorest figure. In the table below, a selection of loss figures is given for system sizes that would be typical for yearly production volumes from 30.000 units up to 300.000, and cycle times less than 10 minutes for a complete manual assembly sequence . The table shows clearly that the parallel system gives a lower loss with the same total workforce. This implies that for a given daily production volume, a lower number of workers is required if the parallel solution is selected. To investigate this further, a simulation was performed before selection of a new assembly system for a Norwegian chair manufacturer. The average operation time for the complete manual assembly cycle was 7.5 min. 10 stations was required to produce the daily quantity using a parallel system. To produce similar quantities in a sequential system several independent simulation runs indicated a need for 12 workers on a 3 stage sequential line. The distribution of the workers was 6, 3, 3 to obtain best fit to the division of subtasks on the line.

Parallel stations

Sequential line, 2 stages

Sequential line, 3 stages

19

13

IO.11

16

13

13

10.14

The calculation is carried out for volumes up to 750 at one station in a parallel system. The most competitive sequential system would have 3 times as many repetitions of the assembly tasks at each station on a stage. The ratio number indicates the relative speed of operation in the two systems. If one compares a parallel system with 12 operators, to the best possible sequential alternative, the 4, 4, 4 system, the parallel system will be better for all volumes. At the 150thproduct assembled, the sequential system will have an average output that is 92% of the output of the parallel system. As the produced volume increases, the assembly time t, approaches rn. t, for both system types. Then the discriminating effect of the learning curve diminishes.

Table 1: Loss figures for alternative assembly system configurations. This result indicates clearly that a parallel solution has the highest efficiency potential. It has less system dependent losses, and is more easily configured. THE LEARNING CURVE EFFECT One major argument for the use of sequential line is that it would allow higher repetition rates and thus let the effect of learning make the assembly operations more efficient. This has been examined briefly. Using the learning curve formula proposed by DeJong [4], the expected reduction in operation time has been investigated. This comparison has to be carried out for each specific line comparison to obtain an overall picture. But using a case that is most favourable for the sequential line will indicate what the potentials are. The learning curve according to DeJong [4]:
5

6 CONCLUSION The hybrid automatic-manual assembly system offers a good opportunity for increasing the efficiency in assembly. In this paper, the merits of sequential versus parallel arrangement of the manual section of such systems have been investigated. It is disproved that the sequential arrangement would be better because it offers better opportunities for training and specialisation. The parallel manual station system offers several advantages:
0

It gives less average loss due to mismatch between required volume and the number of workers. Flexibility and variation in individual products within the product range is very easily obtainable. It is less sensitive to absenteeism. One absent worker will affect the output of his station only. Individual rate variation during the working day will influence the output of the workers station only.

1-m t, = tl . ( m + )
Xr

Here is: x : The number of products assembled tr : Assembly time for the first product t, : Assembly time for the x'th product rn : Incompressibility factor, rn = 0.3 - 0.5 for manual work r : The learning rate exponent The incompressibility factor is the limit value of the fraction of tr that the assembly time t, approaches when the number of products assembled approaches infinity. Table 2 shows calculated time reduction according to the learning curve. The learning rate exponent is set to 0.32 and the incompressibility factor to 0.5, according to industrial experience.

Short leaves from a station does not affect the other stations. Several assembly systems of this kind are used in Norwegian industry. Some of them have been in operation for more than 15 years. The most recent installations have been made after thorough analysis of the serial versus parallel configuration. The conclusion was that the parallel configuration is the best alternative because of the flexibility, and higher overall efficiency.
0

Volume per ope rat0r Parallel

I
Parallel

tx

I
Serial 1

Serial 1

Ratio txse riaI/ txpar 1

7 REFERENCES [ I ] Nevins, J.L., Whitney, D.E., 1978, Robot Assembly Research and its Future Directions, The Charles Stark Draper Laboratory, Inc. [2] Lien, T.K., 1987, Integration of the Product Design and Process Development Activities - A Case Example, 2nd Int. Conf. on Design for Manufacture and Assembly. Newport, R.I., April 5 - 8, 1987. [3]Lien, T.K., 1982, Flexible Control of an Integrated Automatic-Manual Assembly Line, 14th ClRP International Seminar on Manufacturing Systems, Trondheim, 28th - 29th June 1982. [4] DeJong, J.R., 1964, Increasing Skill and Reduction of Work Time - concluded, Time and Motion Study, October 1964: 20-33.

Table 2: Learning curve development of operation time. Comparison between parallel and serial assembly systems. Incompressibility factor rn = 0.5

You might also like