I would like to start a discussion here on the best ways of realizing fast to immediate data evaluation in sensor networks with complex monitoring tasks. A further complication is that the state of the monitored system would change (gradually, abruptly or both), and the monitoring system would have to be able to cope with this e.g. in terms of its predictive capabilities.
An example could be an aerospace structural health monitoring system: If an abrupt change occurred, say structural damage caused by an impact event, the system would have to evaluate sensor data fast enough to suggest changes to operational conditions that would limit loads on the damaged structure. At the same time, despite the structural change within the monitored system, the monitoring approach should remain able to, say, correctly interpret the consequences of a secondary impact, or the behaviour of the system (damage propagation and the like) over an extended operational period.
Now how are such capabilities best achieved? Would the required flexibility demand a model-free approach? Are such model-free approaches fast enough for the speed expected? Is model adaptation a possibility? What about inverse FEM approaches? Can they be ruled out for lack of flexibility, or are they to be preferred for reasons of speed? To what degree are they applicable if things get non-linear?
I hope for a broad discussion, and most probably I have not mentioned all the issues involved yet. Please let me know what you think about this issue.
Dear Dirk,
my personal ideas concerning variable conditions (for instance, presence/absence of damage, temperature variations, different functional objectives) is that Structural Health Monitoring (SHM) systems should be designed to adapt themselves to face different issues, especially in highly-demanding environments (space, seismic areas, ecc.) and for long-term monitoring plans. In this respect, the ability of re-organization or even self-organization (for instance, the possibility to by-pass damaged sensors by defining and eventually modifying the network hierarchy, in real time and by means of remote controllers) of the Wireless Sensor Networks (WSN) is probably a key-feature which lacks in traditional sensor networks. I hope these few words can contribute to the discussion concerning your interesting questions.
In respect to more specific issues, you might be interested in http://dx.doi.org/10.1155/2012/594842, in which some ideas are outlined to develop a technical and economic analysis, with focus on a simple case study which concerns the design of a WSN able to alternately work for OMA (Operational Modal Analysis) or SHM, depending on the designer needs. Best regards, Marco
- Badges
- Science method