Page 308 - How To Implement Lean Manufacturing

CHAPTER 18

An Experiment in Variation,

Dependent Events, and

Inventory

Background
This is a simple experiment that can be done at your desk or in a small group. All that
is needed are some dice and a simple form (shown later in this chapter). The experiment
is a factory simulation using pull production while studying the effects of variation and
dependent events on factory performance.
This is a phenomenon that is understood by only a few.
In short, when we have variation, as we do in any process, coupled with dependent
events, as we do in a multistep process, then the process will not produce to the average
rate of the processing steps, unless we have inventory between the dependent steps.
Plus, as the variation is increased, the inventory levels must increase to maintain pro-
duction. In addition, as the number of sequential steps are increased, the inventory
increases by an exponential factor. It is for these reasons that most factories cannot produce
at the nameplate average rate of the equipment, unless inventory levels are extremely large or
variation is reduced to very low levels.
Read that sentence again.
It gives tremendous insight into why, when projects or production schedules are planned,
halfway through the project, overtime is needed, and in the end more overtime is needed, and yet
we still often need to pay to expedite the shipment.
This experiment explains this phenomenon and more.
Variation and dependent events are everywhere in a factory. Take a simple cell, for
example. Let’s say we have a six-station cell and all work stations have 60 seconds of
work, which is also takt. Also, there is one piece at the workstation and there is no
inventory between stations—true one-piece-flow. When station 1 finishes a piece, so do
stations 2 thru 6, and in unison, all six pieces of in-process work are simultaneously
pulled to the next work station every 60 seconds—the perfect synchronization of
process flow: the ideal state.
But, for the moment, let’s imagine that the cycle time for station 4, although it aver-
ages 60 seconds, varies from 50 to 70 seconds. When station 4 performs at 50 seconds, it
finishes its process, and then station 4, has a ten-second wait time before its product is
pulled by station 5. Ten seconds of waiting time elapse, which is a waste for station 4,

285