|
All names and companies are fictitious,
however the scenarios are not. Products employed are bolded in each scenario
Industrial
Process Optimization
Scenario #1:
Off-Line Problem Resolution
Tom is a Product Engineer at
SigmaMax, where he is responsible for the production of T3140 high contrast
photographic paper. It's 3 AM and he's been called out of bed to come in
to the factory to figure out what the problem is with his product. Since Midnight,
3 consecutive rolls of photo paper were rejected for high DMin, the whiteness
of unexposed material, and the line was shut down. Each roll has a
factory cost of over $37,500 for a total reject material cost of about
$112,500 in
the last 3 hours. Bleary-eyed, Tom stares at the Distributed Control
System (DCS) trend plots for the variables he knows are important.
Everything seems okay. Zone 2 oven temperature seems a bit high, maybe he
should turn it down. The line speed was at 1,250 feet per minute,
a little low. Checking the materials charges on the batches of
photo-sensitive solutions they are applying, he notices that the charges are
all correct, except a few that are abnormal but in specification.
Hummm... It all seems a bit confusing, especially at 3 AM. What to do?
Tom swings his chair around to the
BioComp console and starts up iImprove. He loads a product
improvement system that employs predictive models of his 12 key product
characteristics. These characteristics were modeled using iModel
with cleansed data from iManageData that came out of their on-line
process database, the OSIsoft PI
Data Archive. He knows that he cannot change the batch characteristics
without rejecting 8,000 gallons of silver-based solution, a costly decision, so he tells iImprove
to use the current values for those characteristics. He checks the
specs on the 12 product characteristics that he wants "on target"
and clicks the "Go" button. iImprove starts a search of how to
run this line to get all of Tom's specs on target using this batch.
Within just a few seconds, iImprove has located a good solution, then a better
one, then one that's nearly perfect. Tom reviews the solution summary
report. "Makes sense, speed up the line to 1,375 FPM, drop Zone 1
by 5 degrees, Zone 2 by 10 degrees and increase the solution feed pump rate by
10% and set dampers to OPEN. We must have been over-drying the
product".
Tom instructs the operators to
apply the new set points and machine damper configuration and they start the
line. He waits for 20 minutes for the first roll to come off. QC
tests show all properties within specification. Time for bed.
Again.
Total BioComp Software Cost for
this Solution: Less than $8,500
Scenario #2:
On-Line Problem Pre-Identification and Consultation
Susan is a production operator on
a series of production units. Feed material flows in from the main
pipeline and comes through her column to remove impurities. She
has no control over the feed rate or composition coming to her units. As
she's watching her process, an alert comes over the speakers,
"Warning: Impurity concentration on Unit 4 is anticipated to be high in 30 minutes. Corrective action suggested". Susan walks over to
her console to see a window that has appeared from the BioComp Process
Intellect system. It seems the BioComp Interpretive Performance
Predictor sub-system has predicted that under current trends, the product
will be going out of specification. Susan looks at the DCS trend plots
to see that about 5 out of about 30 variables have drifted different
directions, but it's not clear what should be done, as these conditions offer
conflicting control actions to resolve. She reviews the Process Intellect system advisory, which indicates
that the variance of feed composition, flow rate plus the fact that
it's a particularly warm day is changing the dynamics of her unit. Process Intellect's iImprove sub-system already
formulated an optimized solution and recommends reducing the steam rate on the
heater to correct the situation. Susan makes the
recommended changes and in a few seconds, the alert vanishes as the
anticipated problem is resolved. Downstream, potentially serious storage
problems have been averted and customer satisfaction with product quality have
been assured.
Note: Solutions can be provided that
automatically make corrections
Strategic
Product Design
Determining
Product Characteristics For Increased Market Share
Bill is a product manager for a
Global 5000 corporation. He needs to tune his product design to maximize
market share. He has segmented marketing survey data that matches
alternative product features with customer likelihood to purchase as well as
switch brands. Bill is challenged to come up with a set of features that
will give him the greatest market share but will not explode his manufacturing
costs.
Bill uses iModel to create
a system of models of product features vs. likelihood to purchase and another
system of models of customer willingness to switch brands. He starts iImprove
and loads these two systems. He sets his objectives: Maximize both
likelihood to purchase and switch brands while minimizing costs. He then
constrains his product features within reasonable limits and assigns a cost
function to each. He presses the "Go" button. iImprove
finds superior combinations and degrees of product characteristics that
achieve his objectives. He prints out a solutions report and emails it
to his product team to review before their next meeting.
Discrete
Parts Assembly Optimization
Reducing
Rework to Maximize Capacity and Minimize Factory Costs
(Note: This scenario uses a variant of our iImprove
technologies: the Mix-n-Match Optimizer)
Lisa is a product engineer for an
electronics company. A part of her responsibilities is to assure the
effective production of high-tech electronics assemblies. On her lines,
a variety of sub-components are assembled to create a final assembly which is
tested for over 60 functional properties. Historically, she has been
challenged with the success rate of her first-pass testing. She knows
there are interactions between the characteristics of the sub-assemblies and
her final results, because if they assemble and test, then disassemble and use
other sub-assemblies, they get different results. But she doesn't know
what those final results will be until they assemble the product, an expensive
proposition that is limiting her total capacity and the ability to serve a new
contract that will double their demand.
Lisa uses BioComp's Process
Intellect to create predictive models of sub-assembly characteristics vs.
key assembled product performance measures. She then uses Process
Intellect's Mix-n-Match Optimizer to search through selected
alternative combinations of available sub-assemblies to maximize the overall
performance of all resulting assemblies. The Mix-n-Match
Optimizer does this in just a few seconds by virtually assembling alternative
combinations of sub-assemblies, estimating product performance using the
models created and then providing a list of the best combinations to
Lisa. Lisa uses the system daily, printing out the list of the
sub-assemblies to combine to create the products.
The results of the system have
been remarkable. Assemble-and-test cycles have been substantially cut
with first pass testing rates increasing considerably.
She is now able to serve the larger contract with existing resources,
effectively doubling her capacity.
|
