I am trying to do fault classification/identification and I need to know whether it is possible to do the classification using system output?
In the first step, system outputs are used for parameter estimation of system critical parameters. system a nonlinear reaction wheel model by Bialke. In the second step, I am trying to use clustering algorithms to classify different faults based on their region of operation. However, my supervisor is insisting that classifying faults based on system parameters (estimated parameters from system outputs) is redundant and has no value. He says that using system outputs only, one should be able to classify faults and rank their importance to system failure.
I was wondering if someone has worked in this area and could help me find some sources in this regard.
Redundency is a desirable property for a damage detection algorithm since one increases the dimension of the subspace in which the fault lies. The identification step is necessary in order to capture the excited dynamics which is usually of lower dimension. Including all the signal will falsify the results due to unmodelled dynamics and measurement noise.
Hi Afshin,
You may want to take a look at PCA methods. An excellent reference on this topic is:
Chiang, L.H., Russell, E.L., Braatz, R.D."Fault "Detection and Diagnosis in Industrial Systems. Springer, 2001.
Best regards,
E.Arroyo
Maybe you will find something interesting in these publications: the abstracts are given in my research gate profile:
[1] D. Trawczynski, J. Sosnowski, P. Gawkowski, Analyzing Fault Susceptibility of ABS Microcontroller, Volume 5219 of the series Lecture Notes in Computer Science, Computer Safety, Reliability, and Security pp 360-372
[2] Trawczyński Dawid, Sosnowski Janusz, Gawkowski Piotr: Testing Distributed ABS System with Fault Injection, w: Innovations in Computing Sciences and Software Engineering / Sobh Tarek, Elleithy Khaled ( red. ), 2010, Springer, ISBN 978-90-481-9111-6, ss. 201-206
HIi
Here is an example of combining a model with data for in-operation SHM. Data-driven subspace is on the way. Cheers, Maher
This is a big subject. For a starter, you may read the attached article (that appeared in an Encyclopedia by Springer Verlag). The Braatz paper recommended by someone else is also a good resource.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems