Posted by ""John Gunkler"" <
jgunkler@sprintmail.com>
First, I would think, you need an operational definition of ""bottleneck.""
My work in Lean Six Sigma (process improvement) has taught me that there are
(at least) two kinds of bottlenecks:
1. Capacity constraints (where an operational step is unable to process
enough items, or to process them fast enough, to meet the underlying
customer demand rate); and
2. ""Time Traps"" (steps where there is an excessive amount of time wasted in
non-value-adding activities.)
Operational steps in a process correspond to our flow equations in SD. Both
kinds of bottlenecks result in delays. But to analyze bottlenecks you will
probably have to de-construct your flow equations -- i.e., understand the
underlying processes that create the equation.
Bottlenecks can sometimes be identified by comparing flow rates of upstream
and downstream flows -- when one flow has a rate that is lower than its
neighbors (either upstream or downstream), you may have identified that flow
as a capacity constraint. More tricky is when more than one flow within the
process cannot meet the exit rate required by customer demand. Then you may
have consecutive flows that have equal or similar rates, but all of them are
capacity constraints.
Your model needs to have an explicit rate equation for customer demand --
where ""customer"" means whatever person or process requires the output of
your model. This rate (called ""Takt rate"" in Lean Six Sigma) tells you how
fast the process must produce its outputs. To take a simple example, if
customers are purchasing 600 widgets per day, the Takt rate is 600 widgets
per day, and the Takt time = 1/600 days per widget (or, assuming a
productive work day of 7.5 hours, you have just (7.5 * 60)/600 = 0.75
minutes, or 45 seconds, to produce each widget.)
Producing below Takt rate creates an order backlog; producing above Takt
rate creates a finished-goods inventory. Both are expensive.
Time traps are usually created by policy decisions -- which includes how the
processes that lie beneath the rate equation are designed. Capacity
constraints may also be due to policies (particularly if they have to do
with human capacity), but if they are due to machine or technology capacity
you have a different kind of problem to solve.
Bottleneck analysis (in Lean Six Sigma) is fairly simple:
1. Compute the amount of time each ""flow"" requires to produce one item.
For example, in a port operation, unloading a ship proceeds at a certain
rate for a certain kind of goods. If the unloading process takes one hour
to unload 600 widgets, then (on average) the unloading process has a Takt
time of 45 seconds per widget. If you can only unload 450 widgets in an
hour, then the Takt time is 60 seconds per widget.
2. Create a vertical bar chart of the Takt times of each step in the
process (each flow.)
3. Draw a horizontal line at the Takt time of customer demand (for widgets,
45 seconds.)
4. Look for any steps (flows) whose bar extends above the Takt time line.
Now, test your vertical bars against future (changing) demand rates by
drawing different Takt time lines and seeing which steps will be too large.
To identify the worst bottlenecks, it is useful to use an equation for the
Work Turnaround Time (WTT) for each step. This is defined as:
WTT = Sum (over all the different things processed at that step) of [Setup
Time (for each different thing) + (Processing time for each thing)*(Batch
Size for each thing)]
The steps (flows) with the highest WTT will be the worst bottlenecks.
The book, The Lean Six Sigma Pocket Toolbook, by Michael George et al.,
published 2005 by McGraw-Hill, will be helpful. You may also want to study
Goldratt's ""theory of constraints.""
Posted by ""John Gunkler"" <
jgunkler@sprintmail.com>
posting date Thu, 14 Jul 2005 10:01:58 -0400