To start with the conclusions, we thoroughly enjoyed this model-based real-time application. Main administrative designer Kristian Lien (Hardy) and technical designer Sigurd Skogestad (Laurel) had done an excellent preparation job. For example, assistants were assigned tasks according to gender, and we enjoyed receiving our application identification and authorization code from female PROST members. Practical and computational organization by Tove Krokstad, Steinar Hauan and other PROST people will probably and hopefully set the basis and standard for forthcoming applications like ESCAPE-8 and PSE'2000.
Basic hardware for the application was 1000 years old, but additional modules in terms of Royal (Electric) Gardens gave the 400 end users concurrent access. Despite a broad range of modeling languages, most of the 200 software modules that were run in parallel and sequential sessions, contributed to the overall goal. The goal may be formulated as: How to transform data about physical plant equipment and materials, procedures and controls, and resources, functionality and products into useful knowledge for designers and operators of chemical processing plants? Results from chemical, mechanical and electrical engineering must be glued together with basic laws of chemistry, physics and mathematics as well as recent achievements in computer technology.
Strict synchronization of software modules for increased efficiency functioned well. Intense and very long running periods were strenuous to the end users, but recuperation periods were well organized. To add audible effects like Cantus gave the application added value.
The primary software modules were:
Art Westerberg's module clarified end user resources and needs in this kind of applications. The most interesting module tagname (according to its inventor Kristian Lien).
From Howard Rosenbrock's module we especially remember the dramatic test performed by an SAS end user. Next time he'd have some better supporting tools. The most demanding module for the end-users?
From preliminary textual prototyping results, the Manfred Morari module seemed untested, but it turned out to be full of interesting and valuable information processing. The most loved (by industry-people) and hated (by academic collegues) module?
Knut Harg's module concerned the industrial interface. The annual efficiency improvement requirement of 2-3 per cent is quite a challenge. The most important module for subsequent discussions?
All modules included routines for making the right algorithm or method available to the end user in the appropriate manner at the right time. Fortunately, it seems that these algorithms and methods already exist. It is just a matter of implementation, but implementation must be given more academic weight and status, and much more industrial effort. Important constraints are people/organization, Distributed Control Systems (DCS) and Enterprise Resource Planning (ERP) or business systems.
From plenary discussions, we found the two main topics very well chosen:
On open systems it become clear that we miss a general, common terminology and data model for continuous processing plants. The batch ditto SP88 has simplied integration and interactions considerably (ref.: Sandro Macchietto). The CAPE-OPEN project including main simulator companies like SimSci, AspenTech and Hyprotech seemed very promising, but difficult problems with conflicting business interests must be resolved.
To initiate, develop and improve industrial and academic interactions, joint research and development projects and permanent communication channels remains to be the most popular solution. Still, generally accepted guidelines for how to organize and finance the projects and select their topics are yet to be found. It is important that governments continue to fund a considerable part of the research being carried out by academia. In the opposite case, the universities may lose their independence and ability to support young faculty members.
Otherwise, what did we learn from the application? Well, it's getting very large, and it is difficult to handle all process systems engineering problems in one unit. A careful specification and operation of smaller modules and sessions handling part of the overall problem is required. The desired plug-and-play concept is not yet in place, but we are getting there. Forthcoming applications of this kind should put more effort into tailor-made modules giving all end-users a good overview of problem sections.
This report has been prepared by Knut Wiig Mathisen from Norsk Hydro ASA. Complaints about probable misconceptions, errors and ambiguous statements are rejected.