More Related Content Similar to ABB Scheduling.pdf Similar to ABB Scheduling.pdf (20) ABB Scheduling.pdf1. © ABB
| Slide 1
Industrial Perspectives on the
Deployment of Scheduling Solutions
Iiro Harjunkoski, ABB Corporate Research Fellow, PSE’15 / ESCAPE 25, Copenhagen, 1.6.2015
June 8, 2015
2. © ABB
| Slide 2
A global leader in power and automation technologies
Leading market positions in main businesses
~140,000
employees
Present
in
countries
~100
Formed
in
1988
merger of Swiss (BBC, 1891)
and Swedish (ASEA, 1883)
engineering companies
In revenue
(2014)
billion
40
$
June 8, 2015
3. © ABB
| Slide 3
Power and productivity for a better world
ABB’s vision
Update
A leader in addressing power
infrastructure and control needs
for utilities, industry and transport
& infrastructure
A leader in operational asset
effectiveness – uptime, speed,
yield – and efficiency
Contributing to decoupling growth from environmental impact
Less energy per unit GDP
Less pollution per unit energy
June 8, 2015
4. © ABB
| Slide 4
Power and automation are all around us
You will find ABB technology…
crossing oceans and on the sea bed,
orbiting the earth and working beneath it,
on the trains we ride and in the facilities that
process our water,
in the fields that grow our crops and packing the
food we eat,
in the plants that generate our power and in our
homes, offices and factories
June 8, 2015
5. © ABB
| Slide 5
Innovation is key to ABB’s competitive advantage
Leadership built on consistent R&D investment
More than $1.5 billion invested annually in R&D
8,500 scientists and engineers
Collaboration with 70 universities
June 8, 2015
6. © ABB
| Slide 6
Agenda
What is scheduling
Its role in a hierarchy in transition
Current hypes and trends
Main technical challenges
Added value from scheduling
Lessons learned
Requirements for deployment / information exchange
Conclusions
June 8, 2015
7. © ABB
| Slide 8
Starting Point … The Process
What do you need know to schedule it?
Producing a Schedule Difficult!
Process flow (what can be produced)
Available resources (equipment,
material, personnel, …)
Dependencies & rules (processing steps,
cleaning, change-over times)
Current state and capacity of resources
(including running production situation)
Current orders (due dates, priorities)
Target for the scheduling activity
(throughput, inventory, cost, profit)
The process Resources
Targets Information
Scheduler (short-term)
Experience
Static
≈ Static
Dynamic
Dynamic
≈ Static
Information
typically needs to
be collected from
various
spreadsheets,
documents and per
telephone
June 8, 2015
8. © ABB
| Slide 9
Mathematical Model
Most Information Subject to Frequent Changes
30.6
38.6
34.6
31.6
12
8
part 1 part 2 part 3 part 4 part 5 part 6
Process
Rules
Resources
States
Orders
Targets
Normally no chances
needed
Dynamic data
Must be adapted
regularly
Typically changes
at least daily
Changes are frequent
User must be able
to change these to
adapt to situations
June 8, 2015
9. © ABB
| Slide 10
World
User Interfaces
System
Integration
P&S
Planning and Scheduling
The Big Picture
June 8, 2015
10. © ABB
| Slide 11
World
P&S
Planning and Scheduling
How academics like to think
June 8, 2015
11. © ABB
| Slide 12
World
User Interfaces
System
Integration
Planning and Scheduling
Industrial view
June 8, 2015
12. © ABB
| Slide 13
What Happens in the Surrounding World?
Typically Scheduling Found on Level 3
ERP
(Level 4)
MES / CPM
(Level 3)
Supervisory control
(Level 2)
Regulatory control
(Level 1)
Process
(Level 0)
June 8, 2015
13. Traditional Automation Pyramid Dissolving
More Data than Ever!
ERP
(Level 4)
MES / CPM
(Level 3)
Supervisory control
(Level 2)
Regulatory control
(Level 1)
Process
(Level 0)
ERP
(Level 4)
MES / CPM
(Level 3)
Supervisory control
(Level 2)
Regulatory control
(Level 1)
Process
(Level 0)
June 8, 2015
14. © ABB
| Slide 15
Dissolution of the Automation Pyramid
Opportunity and Challenge
Production systems lose their old stiff hierarchy – enabler
of cross-benefits
By full connectivity basically anything is possible!
Countless innovation opportunities – how to create
added value to customers?
Different question: What actually makes sense?
How to integrate and aggregate data from a large number
– and wide range – of data sources?
Information security and privacy is key!
How to build-in the flexibility and agility necessary to deal
with future changes – and unexpected questions?
Design must support cross-benefit optimization
ERP
(Level 4)
MES / CPM
(Level 3)
Supervisory control
(Level 2)
Regulatory control
(Level 1)
Process
(Level 0)
June 8, 2015
15. © ABB
| Slide 16
Mobility
Market Trends
Five Major Trends that Manufacturers Must Follow
The hype
Analytics
Cloud
Computing
Big
Data
Internet
of Things
What the customer really needs
Safety
Lower cost
and
simplified
operations
Production
efficiency
Better
asset
utilitization
/ ROA
More
effective
decisions
June 8, 2015
16. © ABB
| Slide 17
Real Solution Realizing the Vision
Marine fleet advisory – all data at hand
What the customer really needs
Safety
Lower cost
and
simplified
operations
Production
efficiency
Better
asset
utilitization
/ ROA
More
effective
decisions
June 8, 2015
17. © ABB
| Slide 18
Relevant Technical Hypes / Trends / Enablers
Effect on production planning & scheduling
Automation Cloud Big Data
Renewable energy
Smart Grids
Internet of Things
Mobility
Remote operations
Service
Unmanned sites
June 8, 2015
18. © ABB
| Slide 19
End of Isolated Solutions
Scheduling must balance between dynamic systems
Importance of optimization & scheduling is increasing along the complexity
Amount of information is increasing
Frequency of changes explodes due to direct connectivity
New modeling aspects must be included
Resource availability
and pricing
(ERP, smart grid)
Process variations,
e.g. quality, yield,
disturbances (DCS)
Production
scheduling
June 8, 2015
19. © ABB
| Slide 20
What Value can Planning and Scheduling Create?
Value to an End User
Calculating the benefits of scheduling is easily “too subjective”
The true value of a P&S solution may be difficult to prove
How much difference to earlier manual procedure?
‒ Clear metrics must be defined and agreed upon – in advance!
Most natural to implement a test period and compare production with/without
scheduling
‒ No two periods are exactly the same!
‒ Real production is dynamic, it is difficult to exactly follow a given schedule due to
disturbances, uncertainties in process and times or involved manual decision
steps
‒ Important to isolate the effect of other simultaneous improvements
‒ How to compare reality (what happened) with a theoretical option (what could have
happened)?
NPV = �
𝑛𝑛=0
𝑁𝑁
𝐶𝐶𝑛𝑛
1 + 𝑟𝑟 𝑛𝑛
NPV
Costs for
deployment,
licensing, training
hardware, etc.
Savings
Increased sales
June 8, 2015
20. © ABB
| Slide 21
Success Stories from Real Industrial Implementations
A number of successful examples on scheduling
Source: Computers and Chemical Engineering, 62, pp. 161-193
Example from the dairy industry (Unilever)
Scheduling optimization in the petrochemical industry
(Mitsubishi Chemical Corporation, Braskem)
Production optimization in a pulp & paper plant (ABB)
Crude-oil blend scheduling optimization (Honeywell)
Scheduling of drumming facility (The Dow Chemical
Company)
Scheduling in an integrated chemical complex
(The Dow Chemical Company)
Medium-term scheduling of large-scale chemical
plants (ATOFINA Chemicals, BASF)
30%
2%
2% 10-20 MUSD/year
$ 2.8 MUSD/year
75% Shutdowns
$
?
June 8, 2015
21. © ABB
| Slide 22
What Value can Planning and Scheduling Create?
Supplier versus end user
Supplier will only indirectly benefit from the goodness of optimization
Supplier or solution provider has a different value structure
None of the optimized savings will stay at supplier (unless value-based selling)
‒ Only revenue comes from sales & licensing
Key to minimize deployment costs including associated 3rd party licensing fees
‒ Configurability, re-usability and flexibility of solution is very important
NPV
Costs for
development,
implementation,
deployment,
marketing, etc.
Revenues
from sales
June 8, 2015
22. © ABB
| Slide 23
What Value can Planning and Scheduling Create?
Supplier view
Productization takes long and is often the most costly R&D step (cf. Pharma)
Supplier must make the technology work for the end customer
Integration to other systems, graphical user interfaces, life-cycle support (upgrades),…
Risk: Large investment before any revenues
Multiple sales cases are necessary
Prove the expected value to the end customer (value proposal)
Quick customization and proof-of-concept
€
years
1 2 3 4 5 6 7
costs
revenues
R&D-phase
productization/pilot
sales, deployment
June 8, 2015
23. © ABB
| Slide 24
What is Important for Deployment
Lessons learned
The scheduling model or algorithm is only a part of the overall solution
The scheduler is finally responsible for the schedule and must feel
supported by the tool, allowing him to handle the disruptions in the production
The business opportunity must be well recognized to ensure full
organizational support and alignment. Challenge: estimate true impact
Quick proof-of-concept (weeks): Prove expected value & ability to solve the
problem and availability of data. Define how to integrate the advanced solution
Effective system integration is a must. Communication crucial across systems
Effective user control: Assume the scheduler will adjust the schedule. Do not
try to model all scheduling contingencies!
Automated scheduling must be robust, i.e. the scheduling should always return
a feasible schedule, even when the optimization would fail
Optimal schedule can only bring true benefits when production is aligned to it -
only partially following an optimized schedule may even lead to worse solutions
Source: Computers and Chemical Engineering, 62, pp. 161-193
June 8, 2015
24. © ABB
| Slide 25
Requirements for a Scheduling System
Important aspects to make the change
Automatic data Input/Output – no manual support should be required
Consider running production. Every schedule builds on top of a current
production situation (every schedule is a re-schedule)
Solution algorithms need to be configurable taking into account normal
daily and possibly frequently changing requirements
Enable use of alternative/backup optimization approaches
Optimization results should be returned sufficiently quickly (milliseconds
to minutes)
Allow the results to be manually tuned and adapted by the operator
Objective function i.e. the target is always a simplification of reality
Many decisions are still made manually - an optimization solution may
only serve as a guideline
Ease-of-use is critical to the acceptance of any solution
June 8, 2015
25. © ABB
| Slide 27
Overview ISA-95 Standard (Purdue Model)
Scheduling B2MML Structure
0..n
Is assem bled
from
Personnel
Requirement
Equipment
Requirement
Material
Requirement
Correspondsto
element in the
Correspondsto
element in the
Correspondsto
element in the
May contain
0..n 0..n
0..n
0..n
1..n
Personnel
Requirement
Property
Equipment
Requirement
Property
Material
Requirement
Property
1..n 1..n 1..n
Operations
Schedule
Segment
Requirement
1..n
Is made up of
Is made up of
Process Segment
Work Definition
Requested
Segment
Response
Corresponds
to an <
May correspond
to an <
Segment
Parameter
Operations
Request
0..n
May be made up of
0..n
Physical Asset
Requirement
0..n
Physical Asset
Requirement
Property
1..n
0..n
May contain
May correspond
to an <
Correspondsto
element in the
ANSI/ISA-95
still not perfect
but open to
improvements
June 8, 2015
26. © ABB
| Slide 28
Common Information Platform (ISA-95 / B2MML)
Share All Information through XML-files
Planning
Scheduling
(also manual)
Recipe
management
Control
(tracking)
Production
resource
management ISA-95
Operations
performance
Personnel
Material
Equipment
Process
segment
Operations
definition
Operations
schedule
Operations
schedule
© ABB Group
June 8, 2015 | Slide 28
27. © ABB
| Slide 29
Example Production Process
Used in Following B2MML-Examples
Mixer 1
Mixer 2
Reactor
Packing 1
Packing 2
Packing 3
Raw
materials
Products
Mixing Reaction Packaging
Product Mixer 1 Mixer 2 Reactor Packing 1 Packing 2 Packing 3
A 60 70 120 30 30 N/A
B N/A N/A 240 45 45 60
C 80 80 150 40 40 40
Product Mixer 1 Mixer 2 Reactor Packing 1 Packing 2 Packing 3
A 1200 1000 50 10 10 N/A
B N/A N/A 120 12 12 10
C 800 750 50 10 10 10
Production times (min)
Electricity consumption (kW)
© ABB Group
June 8, 2015 | Slide 29
28. © ABB
| Slide 31
Development Steps for any Deployment Project
Some steps may be iterated over several times
Problem
understanding
(generalization)
Idea development
(modeling, testing)
Prototype building
(proof-of-concept)
Discussion with
customers (input
from end users)
Data interface
definition (data
exchange)
Solution
implementation
(not trivial)
User interfaces
(customer needs)
Testing
System
documentation,
marketing material
Solution or
product launch
29. © ABB
| Slide 32
Emerging and Future Challenges for Industry
Related to scheduling solutions
Faster and more flexible scheduling solutions are needed
Faster innovation cycles needed in order to keep the lead!
Production needs to be more economic, sustainable, energy efficient, flexible, and agile
Modeling of heterogeneous systems with competing targets, different dynamics and time
scales not trivial
Larger optimization problems: Better & faster algorithms/approaches for solving them
Availability of data and proper extraction of necessary information and knowledge for
decision making
Standards to support full collaboration
Dealing with uncertainty emerging from process operations, supply chain, utilities
Electricity, raw-material pricing & availability impact scheduling: potential new opportunities
Globalization – increased the transportation costs and multi-site planning
Rapid market changes - more flexible processes which may need to be adapted on-the-fly
June 8, 2015
30. © ABB
| Slide 33
General Guidelines
Modeling and solution of industrial problems
Different skills must meet, challenge each other and collaborate
Scheduling is a natural part of a well-orchestrated environment
Focus on how to integrate and compromise between various – potentially
competing – optimization functions
Various research fields & communities need to join to merge methodologies
where certain strengths can lend themselves to collaborative solutions
How to efficiently bring the various disciplines together to raise the state-of-
the-art to the next level is a key question
The end-user role is crucial and should be present in all development steps
Acknowledge new technologies & opportunities: System must live for decades
Industry provides more real-life insights and instances to the Academia
Academia values the importance of practical applicability of methods by
targeting fast solution of large and complicated problems
June 8, 2015
31. © ABB
| Slide 34
Conclusions
Long and Interesting Journey for Academia & Industry
More complex systems should become simpler to manage
• Smartphone a good example of this…
Anything is possible: Prioritization essential
• Technology-business alignment to ensure success
Low-level process data and high-level business data need to be combined
in real-time for optimal decisions
• Neutral standards e.g. ISA-95 can help systematizing data exchange
Measuring the true potential of optimization is not trivial – often given
schedule is not realized to even close to 100% exactness
• Manual interaction is often necessary. Important to know which
decisions needs to be optimized and which taken by an operator
Any solution must be maintainable for at least a 10-20 year time horizon
(update models, adapt to major software updates)
Major work ($$$) starts after the mathematics has been proven (90%)
June 8, 2015