Overview and Motivation
Incorporating wireless sensor networks in an industrial process-control
setting promises to increase the number of process variables that can
be monitored and controlled. This is due to the improved ability
to add, upgrade and/or reconfigure process variable wireless
sensors. The potentially large quantity of raw data from
distributed oil and/or gas production facilities that may be supplied
by a network of wireless sensors will have increased value if it is
managed, interpreted and utilized effectively. The main
functions of an ICAM system, from the control/IT viewpoint, can be
organized in terms of Intelligent Agents, which automate and
monitor/diagnose their activities as follows:
Data reconciliation: processing raw data to detect and
remove outliers ("gross errors"), correct inconsistencies (e.g., material and energy
imbalance) and reduce the effects of error sources (e.g., noise);
Process understanding: use the results of
data reconciliation to assess the current dynamic behaviour and
efficiency of operations (e.g., by linearized model identification
to characterize dynamic behavior near the present set-point), and to be
prepared for future operational changes, including changing and/or
adding processes;
Data interpretation: derive higher-level
information related to the operational status of the facility and the
"health" of the system and subsystems (e.g., fault detection,
isolation and accommodation);
Wireless sensor network liaison: collaborate
with the WSN to detect faults (e.g., node failures) and collaborate in
WSN energy management under the constraint that operating control
loops must remain stable; and
Process optimization (improvement): use the
processed data and models supplied by the previous layers, to (1) continually
update an "internal model" (virtual model) of the distributed oil
and/or gas production facilities, so the behaviour of the virtual model
tracks that of the actual process, (2) apply optimization routines to
the virtual model to make the best use of available equipment and raw
materials (where the performance index is maximizing the overall
economic performance of the system), and (3) support management
decision-making in the areas of plant modification and/or expansion.
This last very ambitious functionality has not been addressed.
The ICAM control/IT environment uses an innovative and synergistic
combination of artificial intelligence techniques and control systems
technology. The Phase I version of ICAM, which is nearing
completion, incorporates agents for system identification based on
time-series analysis, dynamic data reconciliation, parity vector
techniques for fault diagnosis and accommodation, and coordination with
the management of the wireless sensor network. A prototype supervisor
utilizing a rule-based systems approach was also developed and
demonstrated. The agent-based architecture was chosen so that other
advanced agents (e.g., using optimization for process understanding and
improvement) can be added in subsequent ICAM implementations.
Continue to the ICAM Architecture Page
Return to the UNB PAWS Home Page
Information supplied by: Jim Taylor
Last update: 3 February 2010
Email requests for further information to:
Jim Taylor (jtaylor@unb.ca)
|