Knowledge-loaded system saves time

Manufacturing Engineering, Jul 2000 by Braunstein, Janet

Smart software takes over "clerical engineering" tasks

A wave of new, smart software saves engineers and product designers time by automating redundant tasks, capturing corporate intellectual property and even reading patents, technical reports, and other documents. Knowledge-based programs free engineers and designers to innovate and help manufacturers cut time to market-two competitive edges required to survive in today's marketplace.

The term knowledge-based engineering (KBE) covers a range of interpretations. At Heide Corp. (Medfield, MA), which offers a package called Intent!, it means automating the engineering of custom or semi-custom products. At C-Mold (Louisville, KY, maker of KnowHow!, it's a company Intranet portal that provides customized information crucial to distinct disciplines, and also captures employee knowledge that might be lost to the company because of worker turnover. At Invention Machine Corp. (Boston, MA), maker of CoBrain and Knowledgist, it's semantic processing, software that "reads" through tons of documents rapidly, extracts key concepts, indexes the materials and presents the searcher with the solutions most relevant to his problem.

D.H. Brown Associates Inc. (Port Chester, NY), a research and consulting firm, divides these programs into two categories: knowledge management (KM) and knowledge-based engineering. Knowledgemanagement systems mine the data, usually documents, squirreled away on corporate servers and stuffed into corporate archives. Knowledge-based engineering systems capture rules and then automate product development within them. But programs in both classes take advantage of the Internet, growing bandwidth, increasing integration among internal corporate programs, and higher-powered, faster PCs to expand beyond these relatively narrow tasks.

A recent D.H. Brown report notes: "To date, relatively few organizations have capitalized on this opportunity. Those who have report compression of the time and cost of design tasks by as much as 90%." Few companies that have made the investment-in time and in dollars-want their competitors to find out about it until they've had more time to exploit the new technology. The edge thus gained reaches beyond time and cost into design and manufacturing quality, and enhanced job satisfaction among engineers.

While visiting one company, Heide Corp. President Scott Heide was taken aback when a group of engineers rose as one and applauded. They had just been told that Intent! could halve the number of engineers needed by their firm. Were they delighted at learning half of them might soon be out of work? Not at all, Heide learned. "They'd all been working at least 60-hr weeks. That group was ready, willing, and waiting for something to give them a break." The company sold its products to schools. From May to November engineers routinely worked overtime. Because business was slow from December to April, while classes were in session, the company couldn't afford to hire extra engineers.

Invention Machine Corp. was founded in 1992, and its first product, TechOptimizer, went on the market in 1995. TechOptimizer is a problem-solving tool based on a technical knowledge-base. It suggests solutions to problems as engineers describe their systems and enter their questions into the software. Invention Machine has a staff of scientists from a wide range of disciplines who conduct research, and pour their knowledge into the software. This work led to the development of Invention Machines' semantic processor, the backbone of its two newer products, CoBrain and Knowledgist.

"Information overload was starting to hit us," says Philip George of Invention Machine. "The question was: how are we going to get through all this technical information? Can we automate the reading process?" The answer: yes. With the first program, CoBrain, "we're addressing the information overload problem within large companies, which have Intranets with huge amounts of data stored, anywhere from 100 gigabytes to four terabytes," George says. "On companies' Intranets, there are hundreds of folders containing hundreds of subfolders containing thousands of documents."

Fed all of a company's on-line text documents, CoBrain reads them, extracts the essential concepts, and creates a knowledge index in a problem-solution format using the knowledge amassed over years within the company. This is automatically placed into an Oracle or DB2 database for enterprise-wide sharing. As a very basic example, while reading a document that includes "water cool the engine," it creates a problem folder called "cool engine" and it files "water" as a possible solution. In another document, it finds vibration may also cool the engine, so it puts the "vibration" solution into the "cool engine" problem folder. The semantic processor uses proprietary algorithms, and other technology, to read and understand the meanings of words, the relationships between words, and their context.

The CoBrain index gives engineers quick access to others' solutions, which might otherwise take weeks or months to dig up. It also shows where the company has real expertise that it can leverage in new product design, George says. In addition, it reveals buried technologies that might be licensed to create more revenue, rather than sitting on the shelf.