mjb@goanna.cs.rmit.oz.au (Mat Boek) (02/28/91)
I have been working in applying back-prop networks to the problem of fault diagnosis of machinery through analysis of their vibration signals, for my Masters degree thesis. The method I use involves collecting vibration signals from tranducers attached to critical points on the operating machine, filtering and digitally processing the signal (FFT) to produce spectra, which represent the machine's state at a given time. Certain features are extracted from the spectra, and these features are used as inputs to the network. As yet, I haven't found any evidence of other work on applying neural-net (or expert-systems) to this type of problem, apart from using nets to classify sonar signals (Gorman and Sejnowski, 1987). Does anyone know of work that is being done in this, or a similar area? Any help, comments etc., will be very much appreciated, and queries cheerfully answered, if possible! Please reply directly via e-mail, if response warrants I'll post a summary.. Thanks in advance, Mat Boek. -- Mat Boek, Computer Science ACSnet: mjb@goanna.cs.rmit.oz R.M.I.T. ARPA: mjb%goanna.cs.rmit.oz.au@uunet.uu.net GPO BOX 2476 V, CSNET: mjb%goanna.cs.rmit.oz.au@australia Melbourne, 3001, AUSTRALIA UUCP: ...!uunet!goanna.cs.rmit.oz.au!mjb
Norbert.Zacharias@arbi.informatik.uni-oldenburg.de (Norbert Zacharias) (03/01/91)
mjb@goanna.cs.rmit.oz.au (Mat Boek) writes: >I have been working in applying back-prop networks to the problem of fault >diagnosis of machinery through analysis of their vibration signals, for my >Masters degree thesis. The method I use involves collecting vibration signals >from tranducers attached to critical points on the operating machine, filtering >and digitally processing the signal (FFT) to produce spectra, which represent >the machine's state at a given time. Certain features are extracted from the >spectra, and these features are used as inputs to the network. >As yet, I haven't found any evidence of other work on applying neural-net >(or expert-systems) to this type of problem, apart from using nets to classify >sonar signals (Gorman and Sejnowski, 1987). Does anyone know of work that >is being done in this, or a similar area? Any help, comments etc., will be >very much appreciated, and queries cheerfully answered, if possible! I'm interested too Norbert -- ============================================================================= Norbert Zacharias Norbert.Zacharias@arbi.informatik.uni-oldenburg.de FB Physik 148964@DOLUNI1.bitnet Carl-von-Ossietzky-Universitaet Tel. 0049-441-7983527 Was Du nicht willst das man Dir tu, das will auch nicht was willst denn Du? Heinz Erhard =============================================================================