Different trains pass over a bridge.
Now how to know from the vibration/ accelerometer data of a bridge over which the train pass the type of train passing?
I suppose that every trains like goods train or passenger train or other kind of trains can create different kind of vibration characteristics.
What are the best algorithms that I can apply over my accelerometer data to differentiate between the slight differences between each data to know the corresponding train type?
No body will tell you "The best" it is the opposite thinking of research. "the --est will close the door". Use examples of previous known data with different kinds of vibration signals and different type of trains and try to use either machine learning models or statistical modeling for automatic recognition of train types after extracting the best set of feature able to do that.
Before to choose an algorithm for this application, I would suggest to take into account that vibrations of a bridge excited by passing trains depend on many different parameters. It is very difficult to discern train typology from its "vibration". As an example, two trains with different masses, speeds and lengths could produce identical vibration (i.e. an identical force that excites the characteristic frequencies of the bridge). Besides, several trains wheels could be damaged by brakings, in this case you can detect relevant acceleration levels in vibrations.
I would suggest to create hidden markov models, as it is a structured pattern recognition, and it could allow to recognize some features regarding the number of axis, bogies or cars...
I would also suggest a microphone, as there might be sounn sources other than rail-wheel contact that can produce sounds that could contribute to separation of classes.
- Badges
- Science topic