Simulations make the case for uneven buffering, placed appropriately between workstations.

Designing production lines with no form of mechanical pacing is not an easy affair. For instance, where to place operators who work at different rates, and where to keep unfinished items along the production line, are just some of the problems facing the line manager.

One important decision is determining the size of the storage space in between workstations where partly finished products are kept, awaiting the next step of the process. This storage space, or buffer, has been the subject of numerous studies, such as buffer capacity allocation and placement.

The underlying problem with buffer capacity decisions arises from two scenarios: In one, an operator temporarily works faster than their predecessor, so buffer stocks dwindle. In this case, the succeeding station will suffer from “starving” delays; the other possibility is if an operator temporarily works faster than their successor. In this case, the buffer soon fills up completely, and the preceding station will suffer from “blocking” delays.

Figure 1 shows a five-station serial line with four buffers, where squares depict stations and diamonds represent buffers. So why not just put in place buffers with equal capacity between each workstation (i.e., a balanced-buffer line)? There are often technical considerations to overcome. In some cases, the total buffer capacity needed for efficient working of the line has to be spaced unevenly. This is called an “unbalanced-buffer” line; i.e., the buffers are not all the same size. And based on our research, it may be beneficial to deliberately unbalance the buffers.


It often has been assumed a line with equal buffer sizes gives the best performance, and anything else is going to lead to losses in efficiency. The work we have done shows this is not necessarily so; buffers of unequal size placed appropriately between workstations can sometimes even lead to increases in effectiveness.

One way to measure line efficiency is to calculate the average buffer level (ABL) for the whole line; evidently, the ideal is as few work-in-process pieces in storage as possible, so ABL needs to be kept low. Another way is to calculate the time the line is inactive (idle time, or IT) as a percentage of total working time. This, too, needs to be kept as low as possible, to reduce labor costs.

Buffer Placement Decisions

We ran computer simulations on lines with five and eight workstations, with total buffer capacities (TB) of eight and 24 units for the shorter line, and 14 and 42 units for the longer (eight-station) line, giving average buffer capacities of two and six units, respectively, for both line lengths. The buffer capacities were then assigned unevenly along the lines. Figure 2 illustrates the general idea for some of the simpler patterns, the five-station line with a total buffer capacity of eight units.


The patterns can be described in five general policies:

• Ascending order: Buffer capacity concentrated at end of line.
  
  
• Descending order: Buffer capacity concentrated at beginning of line.
  
  
• Inverted bowl shape: Buffer capacity concentrated in middle of line.
  
  
• Bowl shape: Buffer capacity lowest in middle of line.
  
  
• General: Buffer capacity not concentrated at one area of the line. It can be either no particular pattern, or zigzag, in which buffer capacity alternates between high and low.

Following the simulations, we were able to see that none of these five general policies was noticeably better or worse than any of the others in broad terms, but particular patterns within the five general policies showed substantial improvements of performance, either in IT or ABL, when compared to the balanced-buffer line.

In each policy, we found that the less extreme the buffer capacity allocation – i.e., the more evenly spread it was – the better the results insofar as IT was concerned. The two best patterns obtained from the simulations fell into the “general” policy and are illustrated below.

Pattern 2 (Figure 3) shows a reduction in idle time (-16.14%) compared to the balanced-buffer line. The other lines were all worse than a line with equal buffer sizes would be, and showed an increase in IT (Table 1).


In terms of ABL, the best two patterns had their buffer capacities concentrated toward the end of the line (ascending order policy) (Figure 4). We can see from Table 2 that the savings obtained are considerable.


All four patterns consistently show great improvements in ABL over the balanced-buffer line. It seems to be well worth unbalancing the buffers to improve stockholding performance.

Conclusions

One of the main conclusions is that how to allocate different sized buffers between workstations will depend on the particular conditions of the production facility.

It may be important to keep as few units of partly completed products in storage as possible; for example, we could imagine fresh produce where hygiene and safety issues are foremost. In this case, it might be advantageous to opt for reductions in ABLs; i.e., placing more buffer capacity toward the end of the line. This is especially the case where just-in-time and Lean buffering strategies are in place.

On the other hand, if we are considering a sector where labor costs are high – automotive, for example – it may be better to move toward reductions in IT, and distributing buffer capacity evenly along the line.

We should remember, however, these patterns are specific patterns among numerous possibilities, and imbalance directed in the wrong way could lead to the opposite effect: increases in ABL or IT.

Finally, the sheer size of the potential savings in IT (16%) and ABL (56%), multiplied over the lifespan of a production line, means purposely unbalancing the buffer sizes the right way may be a policy worth pursuing.

Dr. Sabry Shaaban and Dr. Sarah Hudson are at the Rennes International Business School (school-business.com); sabry.shaaban@esc-rennes.fr.

• Prev
• Next