Project Brief
General Competition (April 1992)Neural Network Control and Sensors for Complex MaterialsDevelop generic sensor and control technology based on neural networks for application to complex materials processing. Sponsor: Honeywell, Inc., Technology CenterSensor & System Development Center3660 Technology Drive Minneapolis, MN 55418
Complex materials--ceramics, semiconductors, metal and polymer-matrix composites, thin films, coatings, and others--are increasingly important in product design and manufacturing. As a result, the efficient control of the manufacturing processes used to make complex materials is a critical arena for industrial productivity and competitiveness. The production processes for complex materials, however, do not lend themselves to traditional manufacturing control technologies. There are few good production models for these materials; their critical properties often can be measured only indirectly from the inputs of several different sensors; exact relationships between process parameters and the finished product are often unknown. To attack this problem, Honeywell, Hercules Aerospace, Sheldahl, and 3M propose to use the relatively new computer technology of neural networks. Based on models of the brain and neurological systems, neural networks can learn and model complex processes even when the rules of those processes are not explicitly understood. They can learn by example, do not need sophisticated programming, and handle unexpected situations far more easily than more traditional computing techniques. The proposed program will develop a generic neural network control system, demonstrate the control and monitoring of critical processes in complex materials manufacture, and prove the system on real-world manufacturing problems faced by team members.
|
ATP website comments: webmaster-atp@nist.gov
Privacy Statement / Security Notice • NIST Disclaimer • NIST Information Quality Standards NIST is an agency of the U.S. Commerce Department |