Intelligent Manufacturing € February € 1996 € Vol. 2 € No. 2
Intelligent Manufacturing € February € 1996 € Vol. 2
€ No. 2
Sandia National Laboratories (Livermore, Calif.) has developed an
intelligent agent-based architecture to support agile manufacturing
at its Integrated Manufacturing Technology Laboratory (IMTL).
Machines linked by Sandia's agent software can request and receive
information from each other, communicating laterally as "peers"
rather than through a hierarchical arrangement.
In the future, the software will give people running the system
recommendations, advice and options for various trade-offs that may
be critical, such as fabrication deadlines, cost and precision
tolerances. This information will enable better-informed decisions
during manufacturing. The intelligent agent software is part of a
joint project involving Sandia, Stanford University's Center for
Design Research and the University of California at Berkeley's
Mechanical Engineering Department.
Intelligent agents are the foundation for a software information
infrastructure that enables the integration of heterogeneous
business, engineering and manufacturing software components that can
be distributed either within a corporation or across corporate
bounds. These agents can communicate with other agents as well as
humans, and the information is presented in a format appropriate to
each. Necessary information technologies to support an agent
architecture for manufacturing include distributed object
technologies, networks and communication protocols, security and
authentication, integrated data and knowledge bases, and human and
agent interfaces.
The system permits planning "on the fly" to take advantage of
opportunities or to correct problems that may arise, explained
UC-Berkeley's Paul Wright, a collaborator on the project. The system
also allows choosing between options before the manufacturing process
begins.
At Sandia's Agile Manufacturing Cell, several machine tools are
coordinated to work in unison. The software under development has
been linked to a lathe, robot and coordinate measuring machine. A
demonstration of the software shows the robot pick up a stock piece
of metal and transport it to the lathe, where it is turned into a
small vessel and probed for on-machine acceptance, then carried to a
cleaning station, next to the coordinate measuring machine, and
finally deposited in a finished part area.
During this process, information from the inspection can be
transmitted via the Internet to other organizations. The software
also works with different computer operating systems. Eventually,
manufacturing with this type of intelligent agent control can take
place in a virtual enterprise across organizational and geographic
boundaries.
"The notion of agile manufacturing has been around for a while now,"
noted Sandia's Hisup Park, coordinator of the project. "It so far has
just been a vision, and what we intend to do with the Agile
Manufacturing Cell is apply flesh to the vision. The product
development process is not a set of discrete steps; it is really a
continuous cycle of events that feed one to another."
The discrete steps, Park said, have been imposed by traditional
approaches to manufacturing in which specialists at each step have
had to guess or assume capabilities of the other parts of the cycle.
Instead of requiring these specialized groups to simply take what the
preceding group has "thrown over the wall" to them during product
development, the new software can provide more thorough information
about the manufacturing operation, which can be used to enhance
product reliability, Park explained, or to build up a database of
knowledge.
Eventually, he envisions a network of responsive, well-characterized
manufacturing centers that are connected through high-speed data
links and can work as a single entity if necessary for either
commercial or defense purposes. Within this network, intelligent
agent software modules can function as well-defined building blocks
that easily fit together to control the process.
The ultimate goal is to help Sandia reduce cost and time for precise
production of defense components. The Sandia agile manufacturing
scheme will also include a virtual reality component in which
operators can test their control software before machining a part,
communicate between dispersed groups and train users.
Sandia is a multi-program Department of Energy laboratory, operated
by aerospace and defense contractor Lockheed Martin Corp. (Bethesda,
Md.).