Page 225 - How To Implement Lean Manufacturing

Cellular Manufacturing 203

56 seconds of work. Consequently, we decided we would design multiple cells using
the work from the first 16 stations, (which totaled 172 seconds of work) and these cells
would then feed into our final cell, which would be testing and packaging. Now, what
would those multiple cells that feed test and packaging look like?
We knew we wanted to increase the cycle time. The short cycle times were a root
cause for much of the waiting time, and the increase in cycle time was causing the low
production. We have found that, for repetitive small work of this nature, cycle times of
30 to 90 seconds seem to work best. So we did some rough calculations.
Each cell will now have 172 seconds of work, if we use four work stations and can
balance perfectly, each station will have 172/4 = 43 seconds of work plus 2 seconds for
transportation. With three cells, operating in parallel, that means we would produce
3 units per 45 seconds, or a theoretical cycle time for overall production of 15 seconds
compared to our design cycle time of 16.5 seconds. On the surface that looks good, but
we know we would not be able to balance the work stations perfectly and since we were
not really worried about overproduction, we decided upon four cells. At this point,
Greg chimed in and mentioned they had already approved some design improvements
in final packaging and test, which would reduce the required work on the fifth cell.
So our final design was four cells of four stations, each working in parallel, and
feeding a fifth cell. We decided to lay out all four new cells in the U design layout and
leave cell 5 in a straight flow line with five work stations.
We now need to calculate the approximate cycle time. We have 43 seconds of work
for the cell, and if we allow two seconds to pass the unit to the next station in the cell, we
would have a cycle time of 45 seconds. If we use four cells in parallel, we could produce
four units in 45 seconds, or about an 11-second cycle time. We are well below takt, were
producing more with the same staff and so we knew we were on the right track.
Wow! Were we happy with that! And we had just begun!

Why Not More Cells?
First, when the restraint of no available capital was placed on us that set the lower limit
of the cycle time at about 12 seconds. This is the current cycle time for the two expensive
testers at the end of the line, hence it made them the de facto bottleneck and the limit on
our rate. Now, in all designs of this nature—single station in series—the theoretical
number of work stations will be the work time divided by the cycle time, or in this case
172/12 = 14. If we add a little for OEE losses, it is easy to see we need 15 or 16 stations.
Next, at this point it becomes a matter of style if we want three five-station cells,
four four-station cells or five three-station cells. This calculation can be refined in addi-
tional ways, but we did not feel our data were accurate enough to draw these conclu-
sions. Somewhat arbitrarily, we selected four, four-station cells, which allows production
modulation in 25 percent increments yet are easier for material supply than five cells of
three each. Quite frankly, any of the three would have worked well in the beginning.

Balancing the Work within a Cell
Now, just how do we balance the work in the four stations of any cell? In our new cell,
we will have four work stations, which we will call stations A, B, C, and D. As a first
pass, we will try to combine the work from the existing stations 1 through 4 into the
new workstation A in each of the four new cells. Likewise, existing work stations 5
through 8 were combined into new work station B; 9 through 12 into C; and 13 through
16 into D, respectively.