Transcript KEI Sentinel Architecture

Kesler Engineering
Component Software:
Using ActiveX Object Modeling
with PI
PI System
User’s Conference
March 1999
Introduction to KEI
• KEI Established 1979 by Michael Kesler
– Co-Author of Lee-Kesler Correlation for
Thermodynamic Properties of Mixtures
– Developed First Industry General Process
Simulator (MW Kellogg -1958)
• Main Business
– Process Engineering
– Software Products
• Sophisticated highly focused models
that work on-line with PI and
ProcessBook
• KEI is Partly Owned by OSI Software
Real-Time and Relational
Databases
• All data is accessible to everybody - all
the time
– Configuration
– Results
– No proprietary application file formats
• Application data is scripted, reported,
and manipulated using real, existing,
well-established, robust tools
Specialized GUIs
• KEI ActiveX controls provide specialized
view of process operations for
operators, engineers, and management
• Configuration Wizards provide ‘notraining’ views of configuration data
• Administrative control panel allows
monitoring and administration of KEI
engine on a per model basis
Embed Application GUI in PB
• Allow end user to customize view of
data
• Use ProcessBook to provide
customizable navigation to data views
• Use ProcessBook trends and other
widgets
KEI Component Software
• KEI ActiveX controls embedded in
ProcessBook
• KEI models are COM objects that are
instantiated when needed
• KEI engine uses KEI models to
generate results
The KEI Sentinel Engine
Creates and Uses KEI Models
Open-Equation modeling system,
calculated and known parameters
are specified in the RDB by the engineer.
Component Based Modeling
Models implement the
IKEIModel COM interface
Parameters are the
‘knowns’ and the
‘unknowns’ of the model
Parameters and Models
• Parameter values come from different
sources...
– Specified constant
• Value is stored in RDB
– PI Tag
• Live (conditioned) value from PI tag; PI tag
specification is stored in RDB
– Calculated Value
• Value is calculated by the KEI engine
Models: COM Objects
Fired Heater Model
• Parameters: Stack Temperature, Stack O2,
Thermal Efficiency, Process Duty,…
• Equations: Overall mass balance, Reaction
stoichiometry, Overall energy balance,…
• Submodels: FuelGas, Pass 1, Pass 2,...
Process Pass Model
• Parameters: Tube length, Mean Metal
Temperature, Mean Flux, Duty…
• Equations: Equations of radiant heat
transfer,…
Models: COM Objects
• When a model is created, its
parameters and submodels are read
from the RDB.
• All model configurations are stored in
the RDB.
Models: COM Objects
• Use the same model for off-line what-if
analysis
– Excel spreadsheets, VB applications
• Open equation model format permits
flexible case studies
– “How much fuel would be required if stack
temperature was 550ºF?”
• On-line case studies
KEI ActiveX Controls Also Create
and Use KEI Models
KEI Model
Explorer
• Developed to persistently store COM objects used in PB
applications
• Containment hierarchy (Company  Refinery  Crude Unit
 Fired Heater  Pass  Duty)
• Possible to trend or get a pump’s inlet and outlet pressure,
vibration, etc., with one click
• Call methods on a heater pass to calculate its duty
KEI Software Philosophy
Products are sophisticated highly focused
models that work on-line
• Software must be robust!
• All data, including configuration and
results, must be accessible to
everyone, all the time
• Software must be easy-to-install and
maintain
Fired Heater Sentinel - Value
• Correct Fired Heater Operation of a 200 MM
Btu/hr Heater Results in an 2% Increase in
Efficiency, Saving US$70,000/yr
• If Fired Heater Is the Plant Constraint, 1%
Increase in Heat Transfer (by Preventing
Coking) Will Increase Throughput by 1%
– Up to US$1 Million/yr in Refining Margins for a
100,000 BPD Refinery
• Pass Balancing Can Increase Plant
Throughput If Plant Is Heater Constrained
Inferential Fractionator
Sentinel - Value
• Estimate $2-5 Million Increased Profit * Due to
Improved Crude Switch Operations
– Allows Control System to Shorten Crude Switch
Time
– Reduce Product Give-away
• Eliminates 1+ Hour Analyzer Lag Time
• Reduces Number of On-line Analyzers
~$100,000 Per Stream Per Property (~5
Analyzers)
*“Mining Gold From Black Oil Using Virtual Picks and Shovels”, P. Nick, Process
Systems Engineering; W. Cheng TOSCO Los Angeles Refinery, NPRA
Computer Conference Nov. 1997.