[Coin-rpx] discussing the plan for the "real problem exchange" at the INFORMS Practice Meeting
Robin Lougee-Heimer
robinlh at us.ibm.com
Mon Mar 20 09:28:14 EST 2006
Dear RPX-members:
Tomorrow 2/21 at 11 EST, there is going to be a conference call to plan
activities to develop the "Real Problem Exchange" at the INFORMS Practice
Meeting. See below. I'll post the call-in information to the list. (To
subscribe to the list, visit http://www.coin-or.org/RPX)
Robin
----------------------------------------------------------------------------------
Robin Lougee-Heimer
IBM TJ Watson Research Center
1101 Kitchawan Road, Yorktown Heights, NY 10598
ph: 914-945-3032 fax: 914-945-3434
robinlh at us.ibm.com
http://www.coin-or.org
----- Forwarded by Robin Lougee-Heimer/Watson/IBM on 03/20/2006 09:19 AM
-----
"Kempf, Karl G" <karl.g.kempf at intel.com>
03/20/2006 07:31 AM
To
"Kempf, Karl G" <karl.g.kempf at intel.com>,
<david.heltne at lakesideassociates.com>, <hpc at acm.org>, Robin
Lougee-Heimer/Watson/IBM at IBMUS
cc
Subject
RE: time to finsih the plan for the "real problem exchange" ...
Thanks for your replies - Tuesday 8 AM PST it is - I'll have my admin get
us a bridge as soon as I get to my office this morning ...
From: Kempf, Karl G
Sent: Thursday, March 16, 2006 7:47 PM
To: david.heltne at lakesideassociates.com; hpc at acm.org; robinlh at us.ibm.com
Cc: Kempf, Karl G
Subject: time to finsih the plan for the "real problem exchange" ...
Team - below is what we committed to in OR/MS Today ......
lunch Monday - facilitated breakout
late Monday - birds of a feather
lunch Tuesday - facilitated breakout
topic 1: form and characteristics of problems
facilitators from practitioner community
example problem statements, models, data sets
distribute questions before session / post on the web
result 1: guidelines
topic 2: policies, procedures, systems
facilitators
(straw schemes ????)
distribute questions before session / post on web
result 2: guidelines
What is "done" ...
Terry Cryan has the two birds of a feather sessions scheduled and has
rooms etc.
I am getting ready to co-facilitate the birds of a feather on form and
characteristics of problems - intend to use one problem set from Willems
(see attached - this is a fixed set of 20 problems of varying sizes
including data for each) and a problem structure that I use in my
university collaborations (see below for a partial description - this I
consider to be a problem generator - it comes with some standard parameter
sets but can be re-parameterized and expanded along many axes).
All the rest we have to nail down. I am free Mon 20th or Tues 21st or Wed
22nd from 9-10 MST (that is 8-9 PST, 10-11 CST, 11-12 EST). Please tell me
if any of these work and I will set up a phone bridge. THANKS - Karl PS
- you might remind me if we had anyone else interested in this that I have
forgotten (in fact you can forward this message to them !!!)
This is a discrete time model with a state description at each time, and a
set of stochastic transitions that move the model forward in time.
Assume there is one factory that produces 2 products named A and B. The
linear manufacturing process is 5 steps long. There is some complexity
associated with the steps - a raw equipment and operator capacity (there
is some number of each with an associated production rate), the minute to
minute availability of the equipment and operators (equipment needs
periodic maintenance, operators need periodic breaks), batching and setup
criteria, quality checks requiring engineering staff, and so on. But
there are only three things that can happen to an individual product A or
B at each step in each time period - it can be successfully processed
through the step and at the end of the period be passed on to the next
step, it can not finish processing at the step and remain at the step
through the next time period, it can be processed through the step but
fail its quality test and be scrapped at the end of the time period
precluding it from ever moving to the next step. Assuming that the factory
will always be heavily loaded (and to avoid the necessity of
"simulation"), each step has associated a probability for each of these
possibilities, and those probabilities are applied to each product during
each time period to decide the state of the factory at the start of the
next time period. There is an overall maximum load that the factory can
support (simply an upper bound on the number of products in the factory).
At the beginning of each time period a decision must be made concerning
how many raw As and Bs to release into the line.
- there is a supply process that is 5 manufacturing stages long
representing one factory that manufactures products A and B
- at each manufacturing stage, for each product, the following things can
happen to the entities in that stage during one time period
- there is a probability that the product will advance one
stage
- there is a probability that the product will wait at the
stage
- there is a probability that the product will be misprocessed
and scraped
- prior to the distinguished first manufacturing stage, there is a
decision node where a decision has to be made each time period for each
product on how much raw material to release (from an infinite supply) into
the first manufacturing stage
- there is an upper bound on the total Work-in-Progress of products A and
B that can be accommodated in the factory at any time
Assume there is a demand for products A and B. There is an aggregate
forecast of order quantities for each product 10 time periods prior to
their desired delivery data. For the first 8 time periods, in each time
period the forecasts can be updated. Due to the variability in the
customers' markets, individual orders can be requested to be delivered one
or two time periods earlier or later. They can also request that
individual orders be increased or decreased in quantity by 5% or 10%.
-there is a demand process that is 10 ordering stages long that consumes
products A and B
- at each stage, for each product, the following things can happen to the
orders in that stage during one time period
- there is a probability that the order will advance one or
two stages
- there is a probability that the order will wait at the stage
or digress one stage
- there is a probability that the order will be decreased
(perhaps to zero) or increased in quantity
- prior to the distinguished first demand stage, there is a decision node
where a decision has to be made each time period for each product on how
many new orders to release into the first ordering stage
- there is no upper bound on the total Orders-in-Progress of products A
and B that can be accommodated in the ordering system at any time
- there is an inventory position into which flows products A and B that
successfully advance out of the last manufacturing stage
- the materials in this position perish after 10 time periods
- there is an inventory position into which flows orders for products A
and B that successfully advance out of the last ordering stage
- the orders in this position perish after 5 time periods
- there is a decision node between these two positions where a decision is
taken at the end of each time period as to which orders to fill with which
products
- each time a product enters the supply process, a charge is applied - $W
for A, $X for B
- each time a product in inventory is matched with an order in inventory,
and both are withdrawn, a credit is applied - $Y for A, $Z for B
- the goal is to maximize profit over some horizon as the difference
between total credits and total charges
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/rpx/attachments/20060320/4a13126c/attachment.html
More information about the Coin-rpx
mailing list