Page 33 - How To Implement Lean Manufacturing

14 Cha pte r T w o

What triggers production is customer consumption. To implement pull, the business
must make the huge cultural change of responding to what the customer does, rather
than what the customer wants. It is the “take one, make one” concept and this production
mode is called “replenishment”. It is virtually impossible to have a pure pull system in
most manufacturing environments, but all pull system have three characteristics.

• First, no production is initiated until actual consumption has occurred.
• Second, since the production process sometimes has mismatched rates or
machines which require changeovers, storehouses are often needed to maintain
flow. Hence, the production process must feed the storehouse, at a rate greater than
the consumption of the customer. However, to maintain control, all storehouses
must have a maximum upper limit on inventory. In fact the entire production
process must have an upper limit on inventory to make pull a reality.
• And third, the production is triggered only when the next downstream customer
in the production process comes to pick up the product. This aspect has gotten
lots of press from the GM supervisor who said, “Yeah, I understand pull, cuz in
a pull system we ain’t sendin nuttin nowhere, somebody comes to git it”.
MassProd systems use “push” production systems. They are the opposite of pull
systems. They rely on schedules and forecasts and try to “push” the product to the next
work station. They do not have controls on inventory so it is possible to have work in
process (WIP) explosions with highly variant lead times and inventory problems galore.
But the real killer problem solved by pull production is variation caused by scheduling
changes, often initiated by customer order changes.
The downside of a pure pull system, is that it must be a “make to stock” system. You
must have inventory to make it work, and recall inventory is one of the named seven
wastes. Sometimes, holding this inventory is not practical so the inventory buffer is
handled with a time buffer or approximated with a FIFO (first in first out) lane.
Finally I only know of two pull systems that have proven themselves in industry.
These two are kanban and CONWIP.

Changeover Times
The issue of changeover times is generally an item that is largely ignored in MassProd.
Long changeover times are often just simply accepted. Consequently, you have two
options when employing model changeovers. First, you can shut down the entire line
when the machinery undergoes a changeover. Of course, this causes a loss of produc-
tion. In this case, all equipment needs to be oversized to account for the downtime. This
then requires a greater initial investment. Alternatively, what is most often done is to
oversize only the equipment that needs the changeover, to account for the downtime,
and design inventory buffers in front of and behind the machines so the rest of the line
can continue producing during the changeover. This helps keep the investment in the
rest of the line down, at the expense of work in process (WIP) inventory. In addition, the
conventional wisdom is that if the changeovers are done less frequently, the time to do
the changeover, which is a form of downtime, can be distributed over a larger volume
of parts, thereby reducing the per-part cost. This conventional wisdom leads to long
runs and extremely large inventories before and after the machinery requiring the
changeover. The Lean solution is a large paradigm shift. That is, in Lean we do not
accept that the changeovers will take a lot of time. The technique employed is SMED