Magazine Nov/Dec 99 pgs.
52-53
Artificial Intelligence in the HPI
Software Requirements and Functionality
Artificial Intelligence applications in the Hydrocarbon Processing Industries (HPI), typically classified as "Intelligent Systems," represent the next level of computer based automation tools assisting management with the safe, reliable, and efficient operation of process equipment and systems.
Government mandated regulations such as Process Safety Management (PSM) and the industry-initiated emphasis on Abnormal Situation Management (ASM) require a higher level of sophistication in industrial software applications. Process surveillance, operations monitoring, and performance reporting must be available to console operators, operations supervisors, and engineering professionals. These groups demand information relating to process alarms, operating stability, and process trends. These needs cover the whole time line from strategic issues to instantaneous events.
Intelligent Systems are typically classified within the categories of Sensor Validation and Smart Objects. Sensor Validation is the category that attempts to determine if the sensor is reading correctly. Although independent multiple sensors can be installed to obtain this information, costs and maintenance of redundant sensors are usually prohibitive. The force driving the process data and instrument reading validation is process safety and reliability. Alarm management, auto shutdown systems and government regulatory standards all rely on timely and accurate data acquisition. As a distinct software product, sensor validation must provide the functionality to audit, assess, and validate process measurements generated in the process units and utility systems.
For example, assume it is 11:30 P.M. when the DCS console operator slops benzene from the Aromatics Extraction Unit because the analyzer incorrectly indicates the product is out of specification. The operator, who has been cautioned to minimize call-outs for lab technicians, diverts the product to the slop tank until the lab analysis is available at 7:30 A.M. Since the Unit is typically operated at full capacity, there is no opportunity to reprocess the slopped benzene, causing lost incremental revenue in excess of $16,500. This common occurrence in the HPI can be avoided with judicious use of validation software.
Sensor Validation must provide the user with analytical tools to perform the following steps:
• Sensor Input Conditioning
• Data Validation Logic
• Message Handling System
Input conditioning routines allow the user to screen the data, filter process noise, and develop multiple inferential models (soft sensors). Validation should provide several voting protocols with the proper choice depending upon the specific application. Finally, the software must provide a flexible, user friendly message handling system so the computer can communicate with the outside world.
Smart Objects are a collection of software components to evaluate the operability, process conditions, and mechanical integrity of process equipment and systems. By the way, Sensor Validation should be applied as an integral part of Smart Object development. If properly applied, a set of Smart Objects provides process and mechanical surveillance analogous to having a process consultant on a full-time basis to assist with troubleshooting activities.
A recent audit of several extensive databases relating to industrial accidents revealed that tube leaks or ruptures in the process furnace is a very common event leading to equipment failure and employee injuries. A computer based system to monitor process heaters for tube leaks and other potential safety or operating problems provides the DCS operator with a valuable tool to enhance equipment monitoring. A simplified example of a heater Smart Object is illustrated in Figure 1.
Smart Object software includes a set of generic modules for the most common equipment classifications such as pumps or compressors. Each generic object includes:
• Simulator (1ST principles)
• Standard Rules Set
• Rule Builder (for customized rules)
• Report Writer
The primary focus of the individual Smart Object is a component that the user can bring up and run with minimal effort and one that can be easily modified or tailored to meet site specific situations. Smart Objects indicates if your Advanced Process Control strategies are working properly or which dynamic model is contributing to the instability.
Modeling and simulation give the Smart Object additional intelligence for the developer to use in building customized rules. The Rule Builder is intuitive in nature with new rules easy and quick to construct. The Report Writer provides the DCS operator or operations supervisor with valuable and accurate information required to solve operational problems.
The Hydrocarbon Processing Industries is very competitive, and successful companies are required to operate at the margin. As experienced in other industries, firms that properly structure and apply advanced computer technologies will remain viable and competitive. Intelligent systems offer operations management an effective software tool to optimize their marginal return.