Simulation programs have historically been run on PCs as fully self-contained programs. The UI, calculation codes and calculations themselves have all been part of the same program on the PC itself, perhaps reading and writing against databases, file stores on the network or reading process data from plant information management systems. But the computing world is changing and many programs are now hosted in the Cloud. Building and running simulation cases is done through a web browser and calculations run without UIs in wrappers called Containers, accessed through web interfaces. Data lives in cloud data stores. The immense computer power of the Cloud and ability to multi-thread simulation technology for parallel processing has opened up the potential to run tens of thousands of simulations, with hundreds at a time, over short time periods in an automated manner – in doing so creating unprecedented levels of codified asset knowledge able to be applied at scale. This presentation will showcase the huge potential for AI/ML algorithms to be utilized and scaled more widely when coupled/trained by first principles simulation, with major benefits in how simulation is used in: Automated model building, maintenance and continuous deviation analysis; Integrated reliability and predictive equipment availability with planning and optimization; and Equipment fouling propensity predictions.
Time Series Foundation Models - current state and future directions
Manage Crew Change Through Coupling of First Principles Simulation, Digital Platforms, AI and ML
1. Mike Aylott
Chief Technology Officer
KBC, A Yokogawa Company
November 11, 2020
Manage Crew Change
through coupling of First
Principles Simulation,
Digital Platforms, AI and ML
2. TOPICS
How First Principles Simulation powers
Operations Planning
Challenges
Digitalization Opportunity
Summary
1
2
3
4
3. 1 month 1 day Current Time 1 to 15 days 1 to 3 months 1 year 2 to 10 years
Strategy
Supply Chain Optimization
Asset Optimization Process Simulation
Supply Chain
Management
Production
Accounting
Variance Analysis
Supply Chain
Scheduling
Production
Scheduling
Production
Planning
Corporate Business Plans
Strategic Plan
Fiscal
Targets
Feed and
Inventory Plan
Model Update
Operating Plan
Actual
Production
Operating Targets
Supply Chain
Planning
Investment
Planning
Annual
Planning
Process & Utility
Optimization
Profit Improvement Process Unit
Management
AP
C
Debottlenecking
Enterprise Data
Integrated Cloud Solution with AI
Integration Model Update
Feasibility
Study
Process and Off-sites Automation
Procedural Automation
Digital Value Chain and Asset Optimization
Actual
Operation Schedule
Process
Designs
4. Today’s Challenges
▪ Limited integration between tools
▪ Linear models have limited validity
▪ Plan does not reflect logistic constraints
▪ Sub optimal schedule
▪ Updating models is time-consuming and
SME dependent
▪ Largely heuristic data reconciliation
▪ Data siloed, poor quality
▪ Slow recognition of opportunities to open
constraints
▪ Control strategies not updated with schedule
changes
5. Collect &
Validate Data
Deploy
Validate &
tune
responses
Update linked
and surrogate
models
Build or
update model
Calibrate
against data
Specific challenges of first principle models
▪ All models have strengths and weaknesses
◆ Time-consuming to
build
◆ Data quality
dependant
▪ All models need maintenance
◆ Situations change: models have to
keep track
◆ Feedstocks, operating scenarios,
unit & catalyst performance
▪ Very dependent on subject matter
expertise which risks being lost
◆ Updates consequently infrequent
7. Why is DX Now Different From The Past?
▪ New platform technologies create new
opportunities
◆ Cloud-based with more flexible storage,
elastic/scalable compute power, flexible
workflow tools leads to more streamlined
automation
◆ Greater insight from big data
analysis atop data lakes
◆ AI Engine maturity allows
ML to supplement first
principle models
▪ Bigger toolbox supports agility to respond to
change and enables stronger value creation
◆ Helps meet Energy Transition
challenges
◆ Helps responsiveness to market
demand changes
◆ Helps drive toward leaner
organizations
8. ▪ Technology
◆ Migrating all applications
to plug into cloud
platforms
◆ Expanding capability
using AI/ML
◆ Cloud and application
agnostic approach
▪ Consulting
◆ Re-tooling/imagining
consulting solutions to
enable agile value
creation
▪ Part of longer-term drive
toward Industrial Autonomy
Our Response to Digitalization Drivers
9. ▪ Data driven, automated identification
of optimization using AI
▪ Intelligent, automated work processes
▪ Automated data management and
integration using semantics
▪ Cloud enabled to facilitate:
• Scalability
• Collaboration
• Rapid enhancements
• Knowledge management
• Integration
• Visualization
• IIoT
Digital Future
10. Using Simulation with AI
Use first principles and
production accounting
models to reconcile data
Use synthetic data from
calibrated simulation models
to drive Machine Learning
(ML) models for planning,
scheduling and optimization
Use refinery-wide
simulation models to rank
improvement ideas
generated from
knowledge bases
Use simulation models to
support broader-scale AI
operations tools
Automate reporting actuals vs.
simulated vs. plan vs. schedule
vs. optimum
Use AI to analyze KPIs and
deviations over time for
improvement opportunities and
when to regenerate models
12. ▪ COVID-19 & Energy
Transition coupled with
crew change are wake-up
calls for our industry
Taking The Plunge
▪ Our simulation technologies
are up for the challenge
◆ Requires we
become more
responsive which
impacts processes
◆ Requires we
capture knowledge
and retool as staffs
change
▪ Digitalization methods &
technologies are enablers
◆ Allows margin
improvements when
market permits
◆ Provides adjustable
information flows when
change is needed
◆ Supports automation
for leaner
organizations
◆ Initial cloud versions
becoming available
now
13. The names of corporations, organizations, products and logos herein are either registered trademarks or
trademarks of Yokogawa Electric Corporation and their respective holders.
The Time is Now
Today’s Digitalization technologies provide
enablers to modernize and streamline value
chain optimization and first principles
simulation.