Real World DSS POP Process Optimization

1
Real World DSSPOP Process Optimization

• At IMERYS Sandersville, GA, U.S.A.-based Kaolin
  Clay Processing Firm
• Sandersville, GA The Kaolin Capital of the World

2
Kaolin Clay

• Paint
• Wallpaper
• Countertops
• Paper and Cardboard Coating
• Sandpaper
• Aluminum Processing
• Ceramics
• Electricity Superconductors
• Stomach Upset Medication

3
Kaolin Clay

• Billions of Dollars in the Worlds Economy
• Main Deposits are in
• Georgia to Alabama (U.S.A.)
• France
• Wales (U.K.)
• China
• Brazil

4
Process Optimization / DSS

• Processing takes a number of steps
• Different steps for different final clays
• Can use different blends of crudes and recipes
  for same final clays
• Can adjust process rates, cuts, chemicals to
  achieve similar results
• Hundreds of mines
• 10s of crude clays to blend
• Hundreds of final products
• 4 plants (or so)
• Supply Chain Management

5
Clay Processing Involves

• Mining and local shipping
• Bashing
• Grinding
• Separating (Centrifuge)
• Thickening
• Oxidizing
• Purifying
• Magnetizing (to pull impurities out)
• Heating
• Blending
• Drying or Evaporating
• Liquifying (Slurrying)
• Packaging (Bulk, Bags, Noodles, Slurry)
• Shipping to customers

6
IT IS A SUPPLY CHAIN

• ERP/SAP/ Fails in the Process Industry
• Need an Enterprise DSS
• Communication within and across plant boundaries
  is NOT enough
• Must truly Optimize!
• Need a PLANNING TOOL

7
POP (Process OPtimization)

• To determine optimal production at several plants
• To optimize crude blends, recipe blends, rates
  and chemicals
• TO MAXIMIZE NET PROFIT (not throughput)
• Annual, 6 Month, 3 Month, 1 Month, 2 Week, etc.

8
POP (Process OPtimization)

• DSS in
• Microsoft Access to manage data, reports, queries
• Lingo Modeling Language The Linear Programming
  Model
• Model linked to database tables through _at_ODBC

9
POP Development Process

• Formed a Team
• Invited a consultant in
• Got bad advice
• Invited another consultant in
• Got good advice
• He joined the team
• He did almost all of the Model development work

10
The Team

• Consultant Linear Programming and MIS expert
• 2 Programmers
• Small Plant Manager (became team leader)
• Project Leader (left and came back)
• Accountant
• Team evolved as needed

11
Others

• Deepstep Plant Manager
• Deepstep Plant Assistant Manager
• Mine Manager
• Production coordinator
• Others

12
Rapid Prototyping Process - 1

• Consultant (and IS Specialists) had to learn how
  clay was processed.
• The Small Plant Manager knew how her plant
  worked. We had much contact with other plant
  managers.
• Consultant had to learn how to develop a Lingo
  Model.
• Small Plant modeled after 8 Months.
• With one IS specialist, they developed the _at_ODBC
  link between the model and database systems.

13
Rapid Prototyping Process - 2

• Then Eureka Moment The model is a
  near-generalized (multicommodity) network flow
  problem.
• The model fell out easily and the tables were
  designed to match.
• First big plant running after 3 more months
• Minor changes following November 1999.
• New plants added between then and Summer 2002.

14
POP Model Structure

• Simple Processes
• Complex Processes
• Nodes
• Links
• Conversion Nodes
• Supply Nodes
• Demand Nodes
• Transshipment Nodes
• Proportion Constraints (2 Kinds)

15
Pop Structure On Next Few PagesFor the Hydrous
(Deepstep) PlantDoes not IncludeCalcine
PlantDryBranch
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
Sandgrinders Bank Simple
22
Magnet Bank - Complex
23
Birds - Complex
24
Bays with Blends/Proportions
25
POP (Process OPtimization)

• Demo of POP
• Modify the data tables
• Update Demands, Unit Revenue, Clay Costs, etc.
• Mega Users can update the process data
• Create a Case
• Run Case (Lingo)
• Lingo returns solution into POP database
• Can save case to compare to others

26
POP Main Menu
27
POP Reports Menu
28
Chemical Use Report - 1
29
Chemical Use Report - 2
30
Utilization Chart
31
Final Demand Report - 1
32
Final Demand Report - 2
33
Open Market Purchase Report
34
Objective Report - 1
35
Objective Report - 2
36
Objective Report - 3
37
Forecast Analysis By Product
38
Forecast Analysis By Process
39
Process Details
40
Lingo With Model
41
Pop Impacts (Highlights)

• In 1999 Correctly predicted the cost of a new
  process because good cheap clays are being
  replaced by not as good, expensive clays
• Annual Model Increased profit by about U.S. 20
  Million out of U.S. 180 Million by increasing
  throughput
• Correctly set blends, chemicals, etc. when moving
  a major clay product to Deepstep. Determined how
  the plant could massively increase throughput.
• Used annually, quarterly, monthly, and biweekly
  for planning

42
Why Not Optimize?

• Managers incorrectly assume that communication
  and collaboration equals optimization.
• Business Schools have dropped Management Science
  / Operations Research courses and programs.
• Math is hard!
• It is too hard to understand the system!
• It is too hard to understand the model!
• It is too hard to understand the solution
  methods!
• It is too hard to get someone to build such a
  system!
• Get Real!

43
The Future

• Can use in (most) any process industry
• Clay Production
• Pharmaceuticals
• Chemical Processing
• Steel
• Others
• Can probably use in discrete parts industries.
• Can save millions of Dollars or Euros annually.
• Why not develop such systems?