Focus and Leverage Part 73

In my last posting we talked about Value Stream Mapping and I said that in addition to all of the usual benefits associated with it, the one that truly stands out from the pack is our ability to identify the system constraint.  In this blog I want to shift gears and briefly talk about the potential benefits and rewards of cellular manufacturing if you do it right.  I will however, focus on one benefit that isn’t discussed much and that's the impact it can have on variation reduction.

We’re all familiar with the positive effects of implementing cellular manufacturing in our workplaces such as the improved flow through the process, the overall cycle time improvement, throughput gains as well as others.  But there is one other positive effect that can result from cellular manufacturing that isn’t discussed much, the potential positive impact it can have on variation. When multiple machines performing the same function are used to produce identical products, there are potentially multiple paths that parts can take from beginning to end. There are, therefore, potential multiple paths of variation. These multiple paths of variation can significantly increase the overall variability of the process.

Even with focused reductions in variation, real improvement might not be achieved because of the number of paths of variation that exist within a process. Paths of variation, in this context, are simply the number of potential opportunities for variation to occur within a process because of potential multiple machines processing the parts. And the paths of variation of a process are increased by the number of individual process steps and/or the complexity of the steps (i.e. number of sub-processes within a process).

The answer to reducing the effects of paths of variation should lie in the process and product design stage of manufacturing processes. That is, processes should/must be designed with reduced complexity and products should /mustbe designed that are more robust. The payback for reducing the number of paths of variation is an overall reduction in the amount of process variation and ultimately more consistent and robust products.  Let’s look at a real case study.

Several years ago I consulted for a French pinion manufacturer located in Southern France. When our team arrived at this company, it was very clear that it was being run according to a mass production mindset.  There were multiple, very large containers of various sized pinions stacked everywhere. The process for making one particular size and shape pinion was a series of integrated steps from beginning to end as depicted in Figure 1 below.  The company received metal blanks from an outside supplier which were fabricated in the general shape of the final product and then passed through a series of turning, drilling, hobbing, etc. process steps to finally achieve the finished product.  The process for this particular pinion was automated with two basic process paths, one on each side of this piece of equipment.  There was an automated gating operation that directed each pinion to the next available process step as it traversed the entire process which consisted of 14 steps.  It was not unusual for a pinion to start its path on one side of the machine, move to the other side and back again which meant that the pinion being produced was free to move from side to side in random fashion.  Because of this configuration, the number of possible combinations of machines used to make the pinion, or paths of variation, was very high. Let’s take a look now at the number of paths of variation that existed on this machine as seen in Figure 1.

Figure 1



The first step in the process for making this style pinion was an exterior turning operation with two turning machines available to perform this function (labeled A1 on one side of the machine and A2 on the other side of the machine as shown in Figure 1). This purpose of this first step, like the others to follow, was to shave metal off of the blank to ultimately achieve its final shape and critical dimensions.  The next step in the process is referred to as interior turning and, again, there are two interior turning machines labeled B1 and B2, one on each side.  In the third step there were two possible choices for drilling, C1 and C2.  After the pinion was automatically inspected for cracks (with one, common automated gage), it then progressed to one of two hobbing machines, D1 and D2. The parts were then collected in storage bins and sent as large batches to an outside vendor for heat treatment. Upon return from heat treatment the pinions then proceeded to hard hobbing, E1 and E2 and then on through the remainder of the process as indicated in Figure 1.

The boxes to the right in Figure 1 represent the possible paths that the pinion could take as it makes its way through the process. For example, for the first two process steps, there are 4 possible paths, A1B1, A1B2, A2B1, and A2B2. In Figure 1 the possible paths of variation are listed to the right and as you can see, as the part continues on, the possible paths continue on until all 32 potential paths are seen. Do you think that the pinions produced through these multiple paths will be the same dimensions or will you have multiple distributions? What if we were able to reduce the number of paths of variation from 32 down to 2, do you believe the overall variation would be less and how many distributions would you have now?  Or another way of saying it, do you believe the part to part consistency would be greater?  Let’s check it out.

Figure 2 

In Figure 2 we created what I’ll call a “virtual cell” meaning that we limited the paths of travel that an individual pinion can take by removing the possibility for a part to traverse back and forth from side to side of this machine configuration. In simple terms, the part either went down side 1 or side 2 rather than allowing the gating operation to select the path.  In Figure 2 you can see that pinions passing through turning machine A1 are only permitted to proceed to turning machine B1. Those that pass through turning machine A2 are only permitted to proceed to turning machine B2. In doing so the number of paths of variation for the first two process steps was reduced from 4 to 2. Continuing, the parts that were turned on A1 and B1 can only pass through hobbing machine E1 while those produced on A2 and B2 can only be processed on hobbing machine E2. To this point, the total paths of variation remain at 2 instead of the original number of paths of 16. The part continues to hard hobbing where there are, once again, two machines available. The parts produced on A1B1E1 can only proceed to the G1 hard hobbing machine while those produced on A2B2E2 can only be processed on hard hobbing machine G2. We also instructed the heat-treater to maintain batch integrity and not mix the batches.  So at this point, because we specified and limited the pinion paths, the total paths of variation decreased from 32 to only 2! So what do you think happened to variation when we created our virtual cell?

The key response variables for this process were five, individual diameters measured along the surface of the pinion. As a result of limiting the number of potential paths of variation, the standard deviation for the various diameters was reduced by approximately 50 % on each of the diameters!

But even though we were very successful in reducing variation, there was a problem associated with making this change….sort of an unintended consequence, if you will. Remember in the original configuration pinions could move to the next available machine (i.e. either side) as they proceeded along the flow of the process. With the new configuration, they could no longer do this.  Prior to this change, when there was downtime on one side of the machine, the automation and/or operator simply diverted the pinion to the same machine on the other side so as to keep the parts moving. With the new configuration, when a machine in the cell went down unexpectedly, the parts now had to wait until the machine was repaired.

The immediate short term result of this change was a significant reduction in throughput of pinions because of unplanned downtime. However, in the longer term, it forced the company to develop and implement a preventive maintenance system which eventually reduced the unplanned downtime to nearly zero. When this happened, the new throughput surpassed the original throughput and the variation was reduced by 50 %! In addition, the scrap levels for this process were reduced by 40%!  And the great part was, not a single dollar—or should I say Euro—was spent in doing this, yet the payback was huge.   So as your studying your process for variation reduction, keep the concept of paths of variation foremost in your mind because it can make a huge difference in some circumstances.

Bob Sproull