Chemical & Engineering News,
March 27, 1995

Copyright © 1995 by the American Chemical Society.

n-Dimensional modeling manages corporate information

Information management within a company is becoming increasingly important because of the growth in corporate information flow and the process modeling that goes along with it. Various software systems for managing information exist, each with its advantages and disadvantages. But at Carnegie Mellon University in Pittsburgh, researchers are developing a system specifically aimed at supporting design, marketing, and other corporate activities - the full life cycle of a project.

This system is an information modeling environment called n -dim (for n -dimensional information modeling). The " n -dim group" at Carnegie Mellon has been working on the system for four years. The project has progressed through three generations of development to a current multimachine, multiuser demonstration system.

With close industrial collaboration, the Carnegie Mellon group is testing n -dim as a support environment for both software engineering and electromechanical component design. In collaboration with a joint project of the Department of Defense's Advanced Research Projects Agency and industry, the group is developing a prototype system on the Internet to support searching for and contracting with vendors for engineering services such as analysis, design, catalog access, and physical prototyping.

Any information management system for corporate use, including design, should have a number of wide-ranging characteristics. Among those considered desirable by the Carnegie Mellon team are keeping and linking information on people and their activities, on formal and informal organizational structures relating those people, on numerical and symbolic data describing physical artifacts (such as data for a flowsheet computation), on group development of arguments, on drawings, on correspondence and e-mail files, and on many more sources of information.

The system should also have the capability to store tremendous amounts of information; to find things in that information by browsing, by formal searching, or by being directed to it; to respond very rapidly; to record enough history of the data creation process so a user can study the process or reproduce the arguments behind the data and not just the values; and to support exploration among alternatives for deciding, especially for a new activity, what information to gather, how to structure it, and how to share it.

The directory/subdirectory tree used by all computer operating systems is one simple form of information management. Databases and document management systems are other ways. World Wide Web (WWW) on the Internet takes a hypertext approach, which links text or graphics documents through embedded labels (activated by a mouse) in one document that point to the addresses or locators of other documents. The n -dim system is somewhat analogous to WWW but is more adapted to corporate and design activities.

An object-oriented approach designed to have the characteristics deemed desirable by the research team, n -dim is a more extensively developed system than can be detailed here. But a few of its features help to capture its essence.

Conceptually, n -dim is based on "models" made up of "objects" and "links." Links tell how files are interrelated and are themselves objects that point from one object to another and always have an associated label. At the lowest level, model object types in n -dim include integers, real numbers, complex numbers, frames (structures of data that, for example, can become the input for an external program such as a simulator), file pointers, and the like. From these and other previously generated models, new n -dim models are constructed, each simply a list of links. At a higher level, grouping supplements this overall structure so that it can be known, for example, which "data file" leads to which "printout." A given file can exist in several groups.

Special "has-part" links point from the identifier of the model itself to the identifier of an object the model "contains." The object may exist separately elsewhere than in the model, but when the model is displayed on a monitor, it is the display that gives the impression that the object is "in" the model.

Each n -dim object has several attributes such as a unique n -dim identification string, the name of an owner, a time of creation, a time of what is termed publishing, and a language used to create it. The object also has its contents, which for a model is a list of has-part and labeled links. The system stores all attributes except the contents for large objects (such as text, graphics, audio, or animation files) in any of a number of databases throughout its web. In this way, a query can be constructed by using a model's attributes to find the model, which may be stored anywhere in the web.

One of the concepts in n -dim is that of "publishing." Publishing is a mechanism for maintaining history in n -dim to make an object persist. It is similar metaphorically to the operation of a library. That is, once an article is published, it will always exist and cannot be altered. Any two people claiming to have read the same article actually will have read different copies, but ones that are guaranteed to be the same.

Objects to which n -dim can point can be partitioned into two types. In one type, objects are given to n -dim to store and manage; in the other, the users' own files or external programs belong to the system as a whole, and are managed outside n -dim by the user.

An editing process is used to create objects in n -dim. For a model managed by n -dim to be shared, its creator must publish all the parts of the model. A published object is guaranteed to be immutable if n -dim manages everything the object knows about either directly or indirectly through its parts. n -Dim will not allow a published object to be modified by anyone, even the creator.

However, since a published object may not be considered finally correct, n -dim provides a way to manage revisions. Anyone can relate a revised object to an original by creating a separate "revision model" that contains the two objects with a revision link that n -dim uses to locate revisions rapidly.

In this way, an original object will continue to exist along with the revision. But someone can open the latest revision rather than the original by means of what is called a prescription model. This model has two parts, one of which is the original object and the other an n -dim model representing a person or group of persons. The n -dim system doesn't actually open a prescription model if a user clicks on it but locates the latest revision of the object made by anyone in the group and opens that.

These and other provisions of n -dim combine to provide the benefits of greatly expanded information management. But the developers note there is another side to the information management coin that technology alone can't resolve. As one example, maintenance of a more complete employment history may make some employees nervous about how their supervisor will use the information when annual review time comes around. Or, in another case, with more information retained, a company might expose itself to litigation if evidence needed by the opposing party is residing in the information management system.

Making up the n -dim group within the university's Engineering Design Research Center (EDRC), which is funded by the National Science Foundation, are chemical engineering professor Arthur W. Westerberg, and EDRC research faculty Robert Coyne, Eswaran Subrahmanian, and Michael Terk. Also in the group are Ph.D. students Douglas Cunningham, Allen Dutoit, Robert Patrick, and Mark E. Thomas; research staff Ira Monarch and Suresh Konda; staff Sean Levy and Eric Gardner; and affiliated faculty Yoram Reich of the department of soil mechanics at Tel Aviv University, Israel.

[ACS Home Page] [ACS Publications Division Page]