Comparison of the use of neural networks versus statistical models in fault detection for cable television networks

In this work, we present a model-based method for reliably detecting reverse pilot faults within cable amplifier networks. This method has the advantage over traditional fixed bound fault detection techniques in that it is able to track changes in the environmental conditions and accurately detect changes in signal behaviour. The resulting method offers increased fault detection sensitivity and reduced false alarms rate. We have implemented a general approach based on the use of a modeling engine which is capable of capturing the behaviour of the reverse pilot of cable television amplifiers. Two modeling engines were developed for this purpose. The first one is based on the use of feedforward neural networks, and the second one is based on the use of statistical analysis techniques. The resulting fault detection system, employing either modeling engine, was able to provide good temporal localization of the start of fault conditions and a clear indication of the presence of the fault through its occurrence.

[1]  Nikitas J. Dimopoulos,et al.  Training asymptotically stable recurrent neural networks , 1995, 1995 IEEE International Conference on Systems, Man and Cybernetics. Intelligent Systems for the 21st Century.

[2]  Nikitas J. Dimopoulos,et al.  Techniques for confident and reliable fault detection in large scale engineering plants , 1995, 1995 IEEE International Conference on Systems, Man and Cybernetics. Intelligent Systems for the 21st Century.

[3]  Nikitas J. Dimopoulos,et al.  Neural Network Fault Detection for the Reverse Pilot of Cable Television Networks , 1998 .

[4]  Nikitas J. Dimopoulos,et al.  Early fault detection in cable television networks (the case of the reverse pilot) , 1997, 1997 IEEE Pacific Rim Conference on Communications, Computers and Signal Processing, PACRIM. 10 Years Networking the Pacific Rim, 1987-1997.