Type of Document Master's Thesis Author Kiel, David H. URN etd-07292009-090316 Title Active damage control using artificial intelligence :initial studies into identification and mitigation Degree Master of Science Department Mechanical Engineering Advisory Committee
Advisor Name Title Robertshaw, Harry H. Committee Chair Rogers, Craig A. Committee Member Wicks, Alfred L. Committee Member Keywords
- Artificial intelligence
Date of Defense 1993-06-05 Availability restricted Abstract
This thesis presents an initial investigation into Active Damage Control (AD C) using Artificial Intelligence (AI). AI can alleviate the sometimes complicated task of modelling the system and also provides an adaptable solution process. The two research areas of ADC, damage identification and damage control, are studied in separate investigations.
An AI technique called "rule induction" is used for the damage identification study. Velocity data from three plates (one without damage, one with damage at the center, and one with damage at the edge) are acquired using a laser data acquisition system. A set of rules is then induced from these data which accurately identifies which plates have damage and where the damage is located. With regard to the damage control, a real-time, machine-learning technique called "BOXES" is used to locally control the vibration of various systems by identifying their vibrational patterns. Using this technique, it is shown that the computer successfully learns an effective control law for various simulations using its trials and failures as the only learning information. It is also seen that the learning algorithm is somewhat less effective when experimentally applying this method to a pin-pin, aluminum beam. A discussion of possible improvements are presented in the future work section.
Filename Size Approximate Download Time (Hours:Minutes:Seconds)
28.8 Modem 56K Modem ISDN (64 Kb) ISDN (128 Kb) Higher-speed Access LD5655.V855_1993.K554.pdf 79.48 Mb 06:07:57 03:09:14 02:45:35 01:22:47 00:07:03
If you have questions or technical problems, please Contact DLA.