The purpose of this work is to help operators controlling large chemical processes to better understand potential abnormal situations and help them during non-steady state processes. This work employees an application developed with Gensym Corporation's G2 Expert System. We work with the process engineers, machine mechanics, and machine operators to help them understand the machine and then record the knowledge into the expert system. The expert system is then used to help the operator's and mechanics diagnose potential problems and provide corrective action guidance.
The particular machines this discussion will focus on make motion picture and x-ray film base. These machines are each approximately 30 years old, with varying sensor capabilities. While one machine may have over a thousand sensors another may only have a few hundred. Also, over the years, engineers have made "improvements" to the machines that then cause the process to vary across the machines. Currently the sensor values are read from the machine and stored into a historian. Additionally, there is a control program that the operators use for monitoring and controlling the process. G2 accesses the sensory information via an off the shelf data bridge. G2 does not currently control the machine.
The traditional path for developing such a tool has been for a group of software engineers to develop the application in a tradition programming language in a "one size fits all" manner. Given the variances between these machines, this type of application does not work well. Additionally, changes to the control program require that the entire machine is shut down, this is difficult to accomplish when some machines are required to run 365 days a year.
The application we are developing using the G2 product permits the user to modify the application while the application is running. Additionally, the expert knowledge is recorded in a graphical representation that the process engineers can easily understand and maintain. This permits the process engineers to add new information or change existing knowledge as it becomes available.
This presentation will discuss the machine interactions we are working with, the process for eliciting knowledge, and the implementation of the expert system.
Colloquia Series page.