Axens proposes S², its State
Space APC suite, designed to efficiently implement and maintain
plant-wide APC applications. We provide robust control and
safe capture of plant benefits.
Main S² toolsDynamic
Modelling PackageISIAC Multivariable
State-Space controllerMVAC
Other S² componentsOff-line Configuration
MVAC Configure Simulation
MVAC Simulation
On-line
Automated Step Tests
MVAC AutoStep InferentialMVAC
Inferential Web
AccessMVAC
Web Maintenance
& ReportingMVAC
Report
Off-line simulation
of complete on-line APC
configuration MVAC
Full Test
S² AccessoriesDCS
connection and S.C.A.D.A. (Supervisory Control and Data Acquisition)
Overview of the interconnection
of the different S² components.
ISIAC:
DYNAMIC MODELLING PACKAGE
ISIAC is a powerful modelling platform,
used to identify the dynamic behaviour of the plant.
ISIAC"Industrial System
Identification for Advanced Control"
Features:
•
Wide choice of model formats: State-Space, FIR, ARX,
etc
•
Model Validity checking
•
Model reduction to first or second order
•
Raw data import from flat file or spreadsheet
•
Powerful data plot tools
•
State-Space format for MVAC use and for plant simulation
•
PID tuning tool
•
PRBS step test sequences generation (for use in automated
step tests)
ISIAC processes step test data, and derives
the control models. This data processing tool allows engineers
to perform custumized calculations, transformations and operations
such as de-noise the signals, collected data slicing and data
batches concatenations. The modelling tool functionality allows
the users to pre-process the chosen data and to apply different
types of models (state space, FIR, ARX). The tool includes
features to help the user evaluates the quality of his modelling
: models response display, confidence indicator... It is also
easy to perform some post-processing operations as models
concatenation and parameters adjustment.
MVAC:ENGINE FOR ON-LINE CONTROL
MVAC is Axens' proprietary State-Space
multivariable controller, providing superior performances, thanks
to the use of latest generation control algorithm.
The
on-line APC application includes a Supervisory Control and
Data Acquisition (SCADA) software, to connect MVAC to any
DCS system. The SCADA includes features such as historian,
log files, DCS drivers (OPC,…), data validation, watch
dog logic and scheduling.
The MVAC Difference
.MVAC provides:
•
All the classical control functions
•
A complete set of robustness features
•
Specific features allowed by the state space technology embedded
in the MVAC kernel:
o Use of intermediate variables (grey box modelling)
o Use of calculated variables (combination of MV, CV
& DV)
o Estimation of non measured disturbances (better predictions
& better control)
Grey Box Modeling: the
State Space approach
The traditional black box modelling
has only inputs (MV) and outputs (CV)
The relationship between
MVs and CVs is extended to describe how unmeasured disturbance
affects the process.
Measurements
of process variables that are not explicitly controlled, provide
valuable information about the internal state of the process.
‘Grey Box’ model adds known
internal causes and effect relationships into the model.
The use of intermediate variables
enhances controller performances:
•
early and accurate disturbance detection is a prerequisite
for superior control
• extended
model provides the means to determine the dynamic state of
the process and its future trajectory
•
the controller is less sensitive to individual model inaccuracies
• control
actions are more robust, less aggressive, and better timely
coordinated
Unmeasured Disturbance
Model
The MVAC tool includes the modelling of the unmeasured disturbances.
Observers (Kalman filters) can be configured for each controlled
variable and for each intermediate controlled variable.
Benefits:
•
High robustness of the control.
• On-line
model inaccuracies correction (for example: static gain).
Overview of
the MVAC engine technology
The main components integrated in the control engine are presented
below.
Real Time Optimization The optimizer embedded in the MVAC engine
is able to operate in 3 different modes:
•
Linear, for economical
optimization, when LP cost can be associated with MVs or CVs. •
Quadratic, typically
convenient to enhance the weight of some MVs or CVs •
External, used
to import External Target (for MVs or CVs) from a rigorous
optimizer
Below is an example of linear optimization, where LP costs
(in k€/hour) are associated with CVs.
Below is another example, where a rigorous reactor model,
associated with a solver, is able to find, in real time, the
best reactor inlet temperature making the trade-off between
kinetic and thermodynamic reaction limitation. Here the optimizer
is configured to accept external targets.
MVAC Configure:
DATA BASE AND CONTROLLER
CONFIGURATION
MVAC Configure provides user friendly tools for Data Base and
Controller configuration.
•
Data Base Configuration,
consisting mainly in:
o DCS tags configuration
o
SCADA customization
o Pre-Calculation •
Filtering •
Analyzer validation
o Post-Calculation
•MVAC
Controller GUI
o used for Configuration and for Tuning
o same GUI used OFF-Line and ON-Line
o HELP available with full documentation and guidelines.
MVAC Simulation:CLOSED LOOP SIMULATION
MVAC Simulation is a ready to use off-line simulation package,
mainly used to validate the robustness of the controller tuning.
The tool allows the user to create
different simulation scenarios, and to introduce different
levels of plant-model mismatch for controller testing.
Following are some of the key features
of the simulation tool:
•
Configuration of
predefined scenarios
o changes in MV or CV limits or targets,
o drop out of measurements,...
o replay with different tunings
•
Plant
model used for simulation
o same or different from control model
o detailed process models can be used
o noise and intermittent signals can be added anywhere
in the loop
HDT unit simulation, with noise,
analyzer and operator changes on CV target and on MV limits.
MVAC AutoStep: CLOSED LOOP AUTOMATED PLANT TEST
MVAC AutoStep (patent in progress) is an automated tool for
on-line, closed loop step testing.
Benefits: •
the testing time is significantly reduced •
the plant variations are controlled and kept into a safe predefined
zone •
the resulting data are easy to process, and
produce high quality dynamic models
MVAC AutoStep takes advantage of
other S² components to run automatically the step tests:
• ISIAC
generates preliminary models from the pre-tests, and a step
sequence designed
o to cover all control relevant frequencies with PRBS
o for multivariable application
o to avoid correlation
• MVAC
Simulation is used
o to simulate the sequence and predict effect on CVs.
o to define step move size for the MVs
o to validate MVAC tuning, if MVAC AutoStep is used
in closed loop
•
MVAC applies
automatically the sequence on the plant
o the sequence is injected in the MV external targets
o if MVAC AutoStep is used in closed loop, the CV limits
are protected.
•
MVAC Web
is used
o to monitor the sequence
o to operate changes on CV limits, MV targets, MV step
move size, etc…
HDT
unit test: Feed Flow and Reactor Temp. are stepped, while
Sulfur in product is protected.
MVAC Inferential:ON-LINE PREDICTION OF
NON-MEASURED PROPERTY
MVAC Inferential permits to estimate on-line current value
of properties not measured on-line, or available only with
significant delay.
Two families of inferential are available •
Axens Proprietary inferential, based
on rigorous kinetic models •
State Space inferential,
based on State Space dynamic models
Axens Proprietary Inferential
For Axens licensed processes, rigorous models are available.
Axens’ expertise in processes and catalysts allow to
build model based, robust inferential structures with physical
sense.
Those kinetic based inferential are embedded
in a structure, where the use of a powerful solver and additional
on-line modules, allows several modes of inferential operation:
•
Direct Mode -
product quality inference
-
feed quality inference
-
interactive what-if studies
-
static gain estimation (useful for APC in case of a non linear
process)
•Update
Mode:
- catalyst activity estimation
Prime-G+ inferential screen,
with access to What-if, Catalyst_Activity and Static_Gains
screens.
State Space Inferential
for Qualities
This inferential
takes full advantage of the ability of the State Space technology,
to make use of the supplemental information available, to
describe the internal states of the process.
State Space Inferential are typically used
to predict a product quality measured either by •
on-line analyzer,
with significant delay, •
or by laboratory analysis, at
variable frequencies.
The goal is to provide APC with real time
quality information.
Off-line, ISIAC produces the State
Space models used for static (or dynamic) modelling of the
quality to be predicted from on-line real-time available data.
On-line, as described
above (Hydrotreatment example), MVAC State Space Inferential
is structured in 2 parts:
•
State Space Model: producing unbiased, open loop predictions
•
Bias update routine, using available analyzer information
coming from
o on-line analyzer •
after analyzer raw signal validation •
using Intermittent signal logic
o or, from laboratory analysis •
after analyzer raw signal validation •
using SPC techniques, to reject statistically questionable
analysis •
applying bias filtering, based on time synchronized inferential
outputs
MVAC State Space Inferential delivers
delay free quality prediction, useful for enhanced APC performance
and better control of product quality.
MVAC WEB: REMOTE
WEB ACCESS FOR MONITORING AND TUNING
MVAC Web provides the facility mandatory for remote monitoring
and remote tuning.
Web Access is
available for: •
MVAC •
APC views •
Trends •
MVAC Inferential •
MVAC AutoStep
MVAC Web is password protected, and allows
secured remote access.
No client software installation is required, except standard
web browser.
Below is an example of MVAC Web used
for MVAC AutoStep sequence monitoring, where CV1 dynamic model
accuracy is checked on-line, using UBP plots (UnBiased Predictions).
MVAC Report:MAINTENANCE AND REPORTING
MVAC Report provides the facility mandatory for maintenance
support:
The logs generated by
the SCADA, are easily extracted and transferred, for routine
maintenance activities, troubleshooting or to document some
particular process control behaviour.
These data can then be directly imported
into a standard curve trending tool, preconfigured
(during the APC project) with templates designed for quick
analysis of the multivariable controller performance (control
curves, statistical analysis, etc…)
Below is an example of a performance
analysis of an HDS unit, where the Sulfur controller has to
reject the disturbances from Feed Flow variations, and from
Feed Sulfur Content variations.
Statistical analysis report is
available to document the performance of the controller.
If needed, the log data can be
used, for detailed analysis, in replay mode, by MVAC Full
Test: see next section.
MVAC Full Test:
SIMULATION OF COMPLETE APC
APPLICATION
This unique component of S² offers the possibility to test
off-line the complete APC application
before implementation, reducing significantly the on-site commissioning
time and potential customer plant disturbance.
The
goal targeted by using this component, is to implement on
site, a fully tested APC application, not only a tuned controller.
To achieve that, the complete on-line APC
application (MVAC, S.C.A.D.A., Addins, Web Access,…)
is installed on a computer where a "dummy" OPC Server
is available.
The OPC Server is connected on both sides: •
on the APC side, to the tested software •
on the DCS side, to a dynamic simulator
A Dynamic Simulator is configured
to simulate
•the interactive DCS screen dedicated
to APC
•the DCS protection logic for APC (watch
dog,…)
• the plant behaviour, using a model
file generated by ISIAC
Once all the components
are started and running, the APC application can be tested in
2 modes
•Interactive mode the
changes are manually made from the simulated DCS screen, from
the MVAC GUI, or from the Web-Access
•Replay mode a
log file generated by the S.C.A.D.A. (from a previous simulation,
or from plant operation), is directly injected in the
dynamic simulator, to replay a previous simulation, or real
data imported from the plant.
This tool is intensively used during
FAT (Factory Acceptance Test) and during training.
