Integrated Supervisory KBS for Online Process Supervision

The integration of various tasks for fault-tolerant optimal operation of batch processes is closer to realization because of the availability of software environments for efficient real-time operations and powerful computer hardware. The first integration problem (horizontal integration) focuses on the coordination of monitoring, diagnosis, and control tasks, and their supervision for enhanced decision making and intervention. The second integra tion problem (vertical integration) focuses on the coordination of technical tasks related to process operations (monitoring, diagnosis and control) with higher level management tasks (supply chain management and plantwide optimization).

Horizontal integration can be formulated as a fault-tolerant control problem where appropriate control policies are formulated and implemented in real time in response to equipment failures and disturbances to prevent operational losses or process shutdown. Consider as an example the failure of a sensor that provides critical information to the control system. The monitoring system will detect an abnormality in process operation either because the change in sensor readings will be significantly different from their expected values or the controller acting on erroneous information will cause significant changes in some process variables. This will trigger diagnostic activities either by an automated fault diagnosis system or by plant personnel. Actuator faults or disturbances will also follow this detect-diagnose-decide-intervene sequence. In a fault-tolerant environment, these activities will be carried out automatically under the supervision of a supervisory real-time knowledge-based system.

Basila and co-workers have developed such a supervisory real-time KBS (MOBECS) for retuning or restructuring multivariable feedback controllers for a tubular packed-bed reactor system [43, 274, 275]. MOBECS (Model-Object Based Expert Control System) was developed initially for a single-input single-output control system [43], then extended to multivariable processes to provide fault-tolerant, minimally conservative robust control by using advanced multivariable control techniques. MOBECS is capable of emulating the steps typically carried out in redesigning the multivariable control system and bumplessly implementing the new control law with the entire control system remaining under automatic control. Control system redesign efforts can be initiated by plant personnel or MOBECS can be instructed to assess process performance automatically and take the appropriate controller redesign actions. If controller restructuring is necessary, MOBECS initiates the development of a new process model by using closed-loop data collected at the current operating point. The new model is used in developing a new controller with improved performance.

Powerful real-time KBS development environments such as G2 of Gen-sym enable the development of more sophisticated supervisory systems for automating and integrating all process supervision and control activities. The building blocks for a real-time supervisory KBS for monitoring and fault diagnosis of multivariable batch fermentation processes are discussed in Section 8.4.2. They illustrate how the KBS coordinates and enhances the interface between detection of abnormality in process operation and fault diagnosis. Additional modules for model development and control system design/modification can be added to implement fault-tolerant control of batch processes.

The real-time supervisory KBS can be vertically interfaced with various software tools that fulfill supply chain management, tracking of process equipment maintenance and repair, and plantwide optimization. Supply chain management interface would transmit consumption levels of raw materials, specific properties (supplier, impurity level, critical characteristics) of feed materials, production schedules and forecasts on product availability. Updating of equipment maintenance and repair records would reduce the surprise and cost of emergency repairs. Automated logging of equipment faults discovered, repairs made, and parts replaced would provide a health record for various equipment, forecast the parts that may be replaced in the near future and restock them to minimize emergency ordering and downtimes caused by waiting for their shipment. Interface with plantwide optimization and planning software will reduce the constraints and time losses in running batch campaigns with many fermentation and storage vessels, increase effective use of every process operation from raw materials to final products, and prevent raw material, resource, and product shortages. This interface will also provide appropriate financial data to management for higher level decision making.

Was this article helpful?

0 0

Post a comment