Scaling up smart structuresJun. 4, 2002, 4:36 PM
NASHVILLE, Tenn. — A new approach may finally make smart structures scalable.
The early promise of smart structures equipping spacecraft, aircraft, automobiles and ships with networks of sensors and actuators that allow them to respond actively to changing environmental forces, an approach predicted to revolutionize their design, construction and performance has never materialized.
That is because researchers found that when such networks grew beyond a modest size of about 100 nodes, they became too complex for central computers to handle. In addition, the weight, power consumption and cost quickly became prohibitive. In other words, they could not be scaled up to large sizes.
Today, however, recent advances in MEMS (micro-electromechanical systems) and distributed computing appear to be overcoming these limitations, reported Kenneth Frampton, assistant professor of mechanical engineering at Vanderbilt, speaking at the Acoustical Society of America meeting in Pittsburgh on June 6.
Frampton, who is an expert in vibration and acoustics, and his colleagues have incorporated these advances using a new approach, called embedded systems, to design a smart vibration-reduction system for a 15-foot-long rocket payload faring. Currently, the high noise and vibration levels inside rockets when they are launched significantly increase the cost of manufacturing satellites and other equipment boosted into space. So a system that reduces these levels by even a small amount would cut payload development costs substantially.
In the first phase of the project, Framptons group prepared and ran a detailed computer simulation of the system that showed it should provide a degree of vibration-reduction comparable to that of a centrally controlled smart system.
The most important result of the simulation is that it shows that the embedded system is scalable, says Frampton. That means we should be able to build it as big as we need to and it should continue to function.
In the older approach, all the sensors and actuators are connected to a central computer. It receives information from all the sensors, processes it and then sends instructions to all the actuators on how they should respond. As the size of the structure and the number of sensors and actuators increase, the amount of wiring required increases dramatically. Difference in arrival times of information from the nearest and farthest sensors also increases, as does the time it takes the farthest sensors to receive their orders. The bigger the system, the greater these and other problems become.
In an embedded system, on the other hand, each node contains a PC-strength microprocessor with a relatively simple program and modest amount of memory that allows it to directly control the sensors and actuators wired to its node. The microprocessor also communicates with its nearest neighbors so they can work together. Depending on how the system is set up, the processor also receives data from a certain number of its nearest neighbors so that it can coordinate the actions of its actuators with those of the other nodes. Although each processor has considerably less capability than that of a central computer, it has far less information to handle, and its workload does not increase as the system gets bigger.
Embedded systems are also far more fault tolerant than centrally controlled systems, Frampton points out. If the central processor breaks down, the entire system shuts down. But a decentralized system will continue to work even when several microprocessors fail, although probably with slightly diminished capability.
The second step in Framptons project is to put a 100-node system into an actual rocket faring comparable to the simulated system. Then he will test how well in it performs in the laboratory. This information will allow the engineers to get better estimates not only of the systems performance but also its weight and cost.
Collaborators on the project include Research Assistant Professor Akos Ledeczi, Associate Professor Gabor Karsai and Associate Professor Gautam Biswas from the Vanderbilt Department of Electrical Engineering and Computer Science.
The research is supported by the National Science Foundation, National Aeronautics and Space Administration and Defense Advanced Research Project Agency.
For more news about Vanderbilt research, visit Exploration, Vanderbilts online research magazine, at http://exploration.vanderbilt.edu