(115c) Expert Systems for Process Unit Troubleshooting

The need to document and disperse expert knowledge and experience throughout a large organization is an important challenge in most industries. At UOP we were challenged with an expanding and increasingly globalized Technical Services function. This group of people provides support services for thousands of our licensed process units in the worldwide refining and petrochemical industry. They diagnose and solve customer operating problems while relying on a comparatively smaller group of process specialists with deep knowledge of selected process technologies. Our objective was to broaden 24/7 access to the expertise of these specialists more widely across our organization. In addition, it was important to increase the troubleshooting and diagnostic capability of our less-experienced field personnel and improve our ability to respond to customer needs. We decided to develop a series of computer diagnostic tools using an artificial intelligence technique commonly called Expert Systems. These programs are designed to address troubleshooting needs for some of our licensed technologies. Over seven years, we have completed seventeen expert systems and implemented them for internal and customer use. These systems address process areas such as hydrocarbon separations, FCC, reforming, hydrogen purification, HF alkylation, propane dehydrogenation and catalyst selection.

Expert systems are interactive programs designed to prompt the user through a systematic decision making process, where the same decision choices must be made repeatedly.  An expert system can be effectively applied to situations requiring diagnosis and problem solving.  The interactive nature of the system simulates a conversation between a technology expert and the user.  The system assists the user in troubleshooting the process unit operating problems and provides the necessary background information to make informed decisions. During a session with the expert system the user benefits by having the problem narrowed down, relevant information has been gathered and the problem may be solved outright.  In addition, the user has learned from the interactive experience. The expert system is also a valuable teaching and learning tool.  

Our approach to development of an expert system requires one or more technical experts to work in concert with a development and knowledge acquisition team. Through the knowledge acquisition process consideration is given to all relevant operational situations, generating discussion of special cases that might otherwise be overlooked.  The expert systems approach allows the full complexity of an expert’s decision-making process to be described and documented. It is especially helpful in complex situations involving balancing many interrelated factors or judgment-based decisions.  The rule-based expert system approach, using heuristic rules processed by the expert system software inference engine, allows even very complex logic to be captured completely.   

Providing 24/7 access to this level of troubleshooting expertise to our staff has provided benefits as mentioned above.  We now also offer these expert systems to our customers on a subscription basis.  Refinery process engineers or managers can directly access the expert systems online and begin immediate troubleshooting of a process unit upset.  This initial effort can facilitate further discussion with our Technical Services staff, making the UOP-customer consultation more efficient. 

The expert systems approach is an effective way to document organizational expertise and extend it to a broad user group who can exploit it to increase their own knowledge and capability.  This approach can be used in any area that would benefit from expert guidance in decision-making.  In this paper we will review the background, development principles, and benefits of the UOP Expert Systems project.