Transcript Presentation title here

July 9, 2014
Using Simulation Approach to move from Manual to
Real-Time Autonomous Scheduling for a Batch Heat
Treatment Process
Steve Thornton
Scientific Fellow - Through Process Integration
Tata Steel Research and Development
Agenda
• Tata Group
• Tata Steel
• Tata Steel Speciality
• Through Process Integration
• Knowledge Engineering
• Data Mining
• Simulation
• TSB AUTOPLAN Project
• Deployment into Real Time Operations
• Summary and Questions
2
A Part of the Tata Group
• In 2007 Tata Steel Limited acquired Corus
Group plc
• On 27 September 2010 Corus rebranded to Tata
Steel
• Tata Group is one of the world’s fastest growing
and most respected corporations
• Tata’s businesses span seven major industry
sectors: engineering, communications and IT,
materials, services, energy, consumer products
and chemicals
• Tata is India’s largest private sector employer
and has over 540,000 employees in over 100
countries.
3
4
Tata Steel
• One of the world’s top 500 companies
• A top 10 global steel producer and the
second most geographically-diverse
steel company
• Annual crude steel capacity of 28 million
tonnes
• Tata Steel employs more than 80,000
people across five continents
• Manufacturing operations in 26 countries
and customers in more than 50 markets
worldwide
• Turnover in 2012-13 was $22.5 billion
• A major part of the Tata Group
5
Key Facts – Deliveries, Turnover and EBITDA
Expansion at
Jamshedpur + New
Steel Works at
Kalinganagar in
Odisha will add
6Mtpa to Tata Steel
India by 2015
[1 Rs. crore = Rs. 10,000,000 = $166,852 - 134,712 Crore = $22.5bn]
6
Through Process Integration - Tools
Knowledge
Engineering
Data
Mining
Discrete
Event
Simulation
7
Through Process Integration
Example – Manufacturing Process for Rail
– Prediction of Final Product Internal Quality and Application to Process Improvement
Liquid
Solidify
Shape
HM
HM
DeS
Conv
STIR
•
•
•
•
DeS
Conv STIR
Hot Metal (95% Fe) from LAF
Blast Furnace
Desulphurisation Process VDG
Convert to Steel (99% Fe)
Stir and Add Alloys
CC4
LAF
VDG
CC4
Electrical Heating and
fine tuning of composition
PEN
Vacuum Degas to
remove Hydrogen
Continuous casting to
solid form of ‘Blooms’
PEN
RHT
RHT
Slow Cooling for
solid state
dehydrogenisation ROLL
(if required)
ROLL
FIN
Reheat to 1250°C for
rolling process
Rolling to elongate and
change profile to
required section shape
FIN
TEST
Time scales – between 2 and 8 weeks
Decoupled Processes
Different Technologies (Manufacturing and IT)
Generally individual processes not well integrated (People and Data)
TEST
Finishing
Processes, e.g.
straightening
Pre-dispatch
product
assessment of
internal and
surface quality
8
Integration Technologies
Knowledge
Engineering
Data
Mining
Discrete
Event
Simulation
9
Knowledge Engineering – Mapping the Landscape
• What is Knowledge ?
(Often intangible) is what is applied to go from data and information, to
a decision and/or action.
• What do we need to achieve (value to the business) ?
• What Knowledge do we need ?
• What Knowledge do we have ?
• Where is it ?
• Is the Knowledge Base secure ?
• Are we applying our Knowledge effectively ?
• Could we do it differently ?
• Could we do something else ?
• Knowledge about knowledge is the raw material for
business improvement
10
Knowledge Guided Data Mining – Workshop Templates
What is the Current Level of Performance ?
KPI #1
What are Aspirational Levels of Performance ?
Concepts
Influences
Measures
Data
Tasks
Process 1
Process 2
Process 3
Process 4
Process 5
Process 6
Process 7
Process N
•
•
•
•
Template to Support Knowledge Capture from different process perspectives
Opportunity to pose cross-process questions and concerns
Identify needed data rather than starting with ‘what have we got’
Develop collaborative knowledge framework integrating KM with Analytics
11
Integration Technologies – Tools in the Box
Knowledge
Engineering
Data
Mining
Discrete
Event
Simulation
12
Data Mining – Is it Cheating to Ask ?
Action
Knowledge
Information
Data
75% of an Organisation’s Knowledge is ‘hidden’
in Data and People !
13
Knowledge Guided Data Mining – Workshop Templates
What is the Current Level of Performance ?
KPI #1
Reheating
What are Aspirational Levels of Performance ?
Concepts
Influences
Measures
Data
Tasks
Differentials in heating
rates at surface and in
centre results in
stresses which could
open up voids,
especially towards the
ends of the blooms
where heat input
through ends could
also be an issue
Reheating rates and
delays during
processing.
Furnace residence
times
Level 2 data from
furnace control model,
harvesting into
MSMLive imminent
Harvest MSMLive
tables
Voids formed due to
oxidation of silicates
and/or excessive
porosity in centre of
bloom
Rates of progressions
through different
sections of the
furnace
Centring of bloom on
the beams in the
furnace
Transient casting
speeds during final
solidification
Delays in different
zones
Temperature
differentials centre to
surface
Visual records of
bloom positioning in
furnace
Speed profile from
mould to unbending
point
Arc_bloom
Arc_bloom_hist
Furnace charge and
discharge times and
thus furnace residence
Align snapshot
timestamps with
discharge times to find
matches with time
based data
Casting speed from PI
data repository at
Steelplant.
Analyse Speed signals
for each strand to
compute speed profile
through machine and
derive parameter to
represent.
14
Data Mining – Making Data Earn Its Keep
Multi-Process, Multi-Variable, Time shifted
• Finding Patterns in Your Data
• Which you can Use
• To do Business Better
• Tools and Techniques for effective combination of business
knowledge and data
• Hunching not Crunching
• Highly collaborative approach requiring knowledge engineering as
well as data analysis skills
• BIG DATA – Volume, Velocity, Variety
• Real-Time Analytics – Pattern Recognition, Rule Induction etc.
15
Functionality
Data Exploitation Maturity Curves
Where
Next ?
time
Data Capture / Measurement
Data Integration / Product Tracking
Business / Supply Chain Integration
16
Functionality
Data Exploitation Maturity Curves
time
Data Capture / Measurement
Data Integration / Product Tracking
Business / Supply Chain Integration
17
Integration Technologies
Knowledge
Engineering
Data
Mining
Discrete
Event
Simulation
18
Discrete Event Simulation (DES) - Characteristics
• Computer based technique for building models of real-life systems
• Which exhibit behaviour approximating that of the real system which can
incorporate natural variability
• Can deal with complex systems which are difficult or impossible to model
rigorously [But cannot recreate “reality”]
• Allow possible outcomes of a scenario to be investigated and thus
assessment of risk and robustness
• Stimulates knowledge capture and ideas generation
• Achieve shared insight and good decisions, more quickly
• Permit simplification of existing situation to provide opportunity to throw
away knowledge legacy
Experiment with your business in the safe
virtual world of the computer
19
Simulation Project – Why ?
Verification
Iteration
Interviews and
Data Collection
Model
Building and
Validation
Experimentation and
Recommendations
Extension
(If required and have time)
25%
25%
50%
Start with Objectives !
What do we want to achieve ?
Why do we need a model ?
What will we do with it ?
20
Simulation Project Process – Questions to Insight
Verification
Iteration
Interviews and
Data Collection
Model
Building and
Validation
Experimentation and
Recommendations
Extension
(If required and have time)
25%
25%
50%
21
How Real Does it Need to Feel ?
As real as it needs to be (to gain confidence of target audience)
22
Applications of Simulation Approach
Lean Manufacturing
•
Explore impact on coupled processes and develop business logic
Restructuring
•
New Plant design and Business configuration, investment and operations decisions
Optimising Logistics in line with configuration changes
•
To ensure continued service for example in conjunction with throughput changes
Scheduling Applications
•
Capturing scheduling knowledge and application for real-time decision support
Product Application Strategy
•
Alternative product application strategy, move decisions downstream
Supply Chain Optimisation
•
Changing scheduling and stock policies to reduce supply chain inventory for major
supply chains
Overall, develop simulation models for experimentation in safe virtual world,
and routine application for decision support
Achieve Shared Insights about the ‘As-Is’ and ‘Could-Be’
23
TPI – Development Directions
Knowledge Engineering > Remove Risk > Improve
Processes > Exploit Capability > Knowledge Systems
Data Integration > Data Mining > Supply Chain
Monitors > Order Fulfilment > Decision Support Tools
Simulation > Operational Strategy > Information and
technology Needs > Scheduling Support > Supply
Chain Development > Real-Time Application
24
Tata Steel in Europe
• Second largest steel producer in Europe
• Crude steel capacity of 20 mtpa
• Approximately 35,000 employees worldwide
• Major manufacturing sites in the UK, the
Netherlands, Germany, France and Belgium
• Supplier to the most demanding markets:
•
•
•
•
•
•
•
•
•
•
•
Automotive
Construction
Packaging
Energy & Power
Material Handling
Consumer Goods
Engineering
Rail
Shipbuilding
Aerospace
Defence & Security
25
Tata Steel in Europe
• Second largest steel producer in Europe
• Crude steel capacity of 20 mtpa
• Approximately 35,000 employees worldwide
• Major manufacturing sites in the UK, the
Netherlands, Germany, France and Belgium
• Supplier to the most demanding markets:
•
•
•
•
•
•
•
•
•
•
•
Automotive
Construction
Packaging
Energy & Power
Material Handling
Consumer Goods
Engineering
Rail
Shipbuilding
Aerospace
Defence & Security
Tata Speciality
Main Sectors
Carbon Steels
Alloyed
Stainless
Heat Treated
26
Qualities of Steel Offered by Tata Steel Speciality
• Alloy through-hardening, casehardening and nitriding steels
• Carbon and carbon-manganese steels
• Micro-alloyed steels
• Stainless steels
• High quality aerospace grades
27
Manufacturing processes
• Melting
• Continuous bloom casting
• Ingot Casting
• Re-melting
• Primary rolling
• Re-rolling
• Finishing
• Inspection and Testing
http://www.tatasteeleurope.com/en/products_and_services/products/long/speciality_steels_and_bar/manufacturing_processes/
28
Process Route
£15M project to commission unit at Stocksbridge in early 2015
29
Continuous Heat Treatment Line
No 7
Hardening
Furnace
“Available”
Stock
Conveyor
Quench
If Forming
Bed 1
Occupied
Forming
Bed 1
Forming
Bed 2
Enhancements
Extra Temper F10
Temper
F8
Temper
F9
Temper
F10
Downstream Brinell & Saw
Out1
Cooling time required
dependent on bar diameter
Out2
19 Pitch Walking Cooling Bank
• Heating to 800°C or 880°C in Continuous Furnace
Brinell
• In-Line Quencher to harden in homogeneous manner
• Tempering at sub-critical temperature (450°C – 650°C) to soften/modify
• Cool to ambient for Brinell Hardness Testing (approx 5 hours)
Saw
• Saw to remove ends and cut required test pieces (3 – 5 cuts per bar)
Exit
30
Run Furnace Without Gaps and Use F10 to avoid
blocking – example Schedule ‘A’
Furnace 10 relatively well utilised for this scenario but some
spare capacity evident.
31
Run Furnace Without Gaps and Use F10 to avoid
blocking – example Schedule ‘206’
Matching of Hardening Furnace to Original Design with 2
Temper Furnaces is very good when charges are ‘filled’ to max
32
Batch Heat Treatment Process
“Supply”
No Oil Quenched Charges
SL Access
CB7
CB6
F6
F5
OQ1
F4
WQ2
CB5
F3
CB4
TC2
TC1
CB3
WQ1
CB2
F2
F1
CB1
Charges
delivered (and
removed) by
Sideloader
Batch Furnaces
(Note 2 shorter than others)
Water
Quench
Oil
Quench
Air
Cool
Transfer Car
‘Temporary store)
Gantry
Crane
Charger
No Hot bars; OK ex water Quench
CB also used as ‘parking’ places, e.g. waiting for removal or
to start treatment, or pause in non-critical timed treatment
Pit - Charges can be placed here by crane prior to
removal from compound ? Also air cooling here ?
Side Loader access, introduction of charges into
compound and removal of some (some ?)
Supply Assumptions
•
•
•
•
•
Assume scheduled material all available,i.e.
No Overlength Bars
Dot Matrix stamps replaced by Hard Stamps
All necessary ultrasonic testing done
Correct number of bars
33
CHT & BHT - Scheduling
• Manual Scheduling approach is the norm
• Until recently, process team leaders were responsible for detailed scheduling
‘on the job’
• ‘Hard’ scheduling recently introduced
• Scheduler prepares detailed 24-36 hour schedule each morning from ‘charge’
information on Heat Treatment system
• Considers
–
–
–
–
–
–
Hard rules and constraints, e.g. next slide
Prioritising ‘late’ or ‘current’ orders – Customer focus
Minimising step changes in temperature between charges – Energy focus
Maximising Utilisation – Throughput focus
‘Availability’ of material
Commercial interventions and requests
• In general, heat treatment is not the final process so some flexibility
• 12-24 hours published to the Heat Treatment System so material can be
assembled and delivered
• Usually, refreshing and/or repairing of schedules only done once per day.
34
BHT - Examples of Rules and Constraints
Confirmed/Determined through Trials (Experiments), Experience and Measurements
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
Some orders must temper within x hours (rare)
Standard practice is to allocate some furnaces to hardening and some to tempering (because of
temperature setting differences) – choice driven by job sheets and experience
Aerospace material – tempering restricted to furnaces F1 and F2 or F5 and F6
For high temperature hardening charges (e.g. 1000+), use ‘suitable’ furnace next to quench station (if
possible)
F1 – only furnace which can be controlled to 450°C for low temperature quenching
F2 – Can be controlled accurately at temperatures down to 500°C
F3, F4 & F5 tend to favour for higher hardening temperatures
Can harden in any furnace however
F6 ‘always’ reserved for tempering
Some charges specify quench method, others give option but times are specified.
Time from hardening to quenching should be ‘as quickly as possible’
If a furnace is empty it is set to ‘minimum fire’.
Furnace temperature losses and ramp rates need to be accounted for.
Example, if a furnace was used to harden at 850°C and then is used for tempering at 650°C, would
take 2 to 3 hours to cool sufficiently. Generally consider a cooling rate of R°C per minute applies.
During removal of charges from a furnace, the control is set to manual whilst thermocouples are
removed.
Ramp up rate is based on is based on volume of steel in furnace.
See next slide for further consideration of heating and cooling rates.
35
BHT – Calculating Required Ramp Up Time
Furnace
Temp
Y – Ramp Up Time
X – Soak Time
°C
Y mins
X mins
Minutes
Regression modelling based on
• charge weight
• Surface area
• Volume
• Time since last use
• Temperature last use
• Last temp – This temp
Charge weight most influential, note that the data
incorporates a lot of unrecorded actions by operators.
A lot of scatter
Best result so far shown right. Up to 10 hours red line is mean of ramp up hours
for ranges of batch weight shown. Afterwards no evidence to suggest other
than flat 4.25 hours should be possible.
Red Line
Hours = 0.325*ChargeWeight + 1
From the data however it does seem feasible that this could be accelerated,
maybe taking the bottom 25 percentile as the envelope…
36
Autoplan – Advanced scheduling algorithms to autonomously
produce and update schedules with minimal human intervention
• TSB Research Project led by Preactor International
• To Develop an Autonomous Systems Development Tool
(ASDT)
• To Assess the effectiveness of Autonomous Scheduling
using end users in real applications
• Project due to be completed in February 2015
• Collaborators
•
•
•
•
•
DeMontfort University
C4FF - Centre for Factories of the Future
Tata Steel (end user)
Plessey Semiconductors (end user)
TDK-Lambda (end user)
http://www.preactor.com/Home.aspx
37
Autoplan – What is it?
• Looking at different ways of production scheduling that:─ Produces schedules autonomously without the need for manual input
─ Produces schedules more often and at fixed times (minutes, hours)
─ Uses different scheduling rules and objectives, compares current status
to performance measures and selects the best rule for the next
scheduling run
• Using three companies in different industries to develop the tools and
assess the benefits and effectiveness of autonomous scheduling
• Also looking at the capability of Genetic Algorithms (GAs) in
scheduling applications (DMU and C4FF)
38
Case Studies – Proof of Concept
Three collaborators/plants have been selected to provide a proof of concept of
autonomous scheduling.
Plessey Semiconductors
TDK-Lambda
Tata Steel
Plymouth – Wafer Fabrication
Ilfracombe – Power Supplies
Sheffield, Steel Bars
Scheduling of 6” wafer fab production.
Scheduling of PCB component
insertion lines.
Scheduling of heat treatment furnaces
39
Process Route
£15M project to commission unit at Stocksbridge in early 2015
40
Autonomous Agent Approach
Importance
Objectives
Performance
C
E
U
C – Customer
E – Energy
U - Utilisation
Tolerance
• Independent rule sets are associated with
each objective
• Positioning on grid determines which
objectives dominate and when to act
• Moving look ahead window evaluates
impact of current rule at each ‘event’ and
changes rule if tolerances are exceeded
• Multiple iterations may be required to test
alternative choices and seek better solutions
Currently evaluating different strategies for Application
41
Consider CHT and BHT as Combined System
Supply as per Heat Treatment System
What is throughput potential of the line for variety of
typical CHT Schedules ?
Utilisations and potential bottlenecks ?
Sideloader from BHT
(Cooled Charges)
Supply
Sideloader
Movements
Charges cool using linear rule, processed
as soon as reach end of prescribed
cooling time. Insert BHT (cooled) charge
in gap of more than 60 minutes available
End (Magnet)
Crane Movements
BHT
Tempers
inserted to
F10 if free
and forming
bed on F8/F9
free
‘Mark and test’
time per bar
before transfer
Brinell
Quench
BUFFER
Sideloader to Cooling Beds
Middle Crane
Movements
F7
Impact on BHT Scheduling ?
F8
F9
F10
F8 and F9 used normally, F10 used if
system blocked or formed charge for
F8/F9 would have to wait more than (30)
minutes
Crane removes bars
one at a time,
between 1.5 and 2.5
mins per cycle
Saw
ReBundle
Straight Through Option ?
(Remove from Process)
42
Run Furnace Without Gaps and Use F10 to avoid
blocking – example Schedule ‘206’
Matching of Hardening Furnace to Original Design with 2
Temper Furnaces is very good when charges are ‘filled’ to max
43
Run Furnace Without Gaps and Use F10 to avoid blocking and BHT
tempering when free + Downstream processing of BHT charges when
gaps available. –Schedule ‘206’
44
Pathway to Intelligent Manufacturing
Deployment
Real-Time Autonomous
Scheduling
Interactive
Workshops
Knowledge
Management
Data
Mining
Rules, Relationships
& Patterns
Process Control
45
Summary
Through Process Integration – a key concept to keep in mind
• Knowledge engineering – understand, secure, deploy
• Data Mining > Analytics
• Discrete Event Simulation
Growth Curves for Data Integration and Exploitation
• Store Everything – Cheaply – (on one platform ?)
• Enable Access – Analysis of Anything (by anyone ?)
• Distill on Demand – Concept of a ‘data ecosystem’
Establish Frameworks and Tools to Support Collaboration
• Attention to Knowledge Management
• “From now on we know it”
Visualisation and Visibility key aspects
• Deliver Shared Insight and Confidence
• Supportive of culture open to automation and process re-engineering
Develop/Identify Options for Real-Time Deployment
46