Method and apparatus for analyzing alarms coming from a communications network

• US 7,103,801 B2
• Filed: 11/01/2002
• Issued: 09/05/2006
• Est. Priority Date: 11/08/2001
• Status: Active Grant

• Current Assignee: WSOU Investments, LLC (WSOU Holdings, LLC)
• Original Assignee: Alcatel SA (Nokia Corporation)
• Inventors: Toure, Famory, Marilly, Emmanuel
• Primary Examiner(s): Iqbal, Nadeem
• Application Number:
  US10/285,437
• Publication Number:
  US 20030088669A1
• Time in Patent Office: 1,404 Days
• Field of Search: 714/26, 714/32, 714/33, 714/39, 714/37, 709/224, 709/207, 340/521, 340/522
• US Class Current: 714/26
• CPC Class Codes: H04L 41/16 using machine learning or artificial intelligence

Abstract

The invention enables alarms (MA) coming from a network (4), such as a communications network, to be analyzed by a step of transforming the detected alarms (MA) into a signal (V(t)) which expresses variation in time of a numerical value representative of all of the detected alarms. This transformation can be implemented using a learning system, in particular a neural network (6). The signal can then be analyzed using various techniques, in particular time/frequency analysis techniques, in order to make a diagnosis.

Claims

1. A method of analyzing alarms coming from a network, comprising:

transforming the detected alarms into a signal V(t) which expresses the variation in time of a digital value representative of all of the detected alarms; and

processing the signal V(t), on the basis of which usable information is extracted.

2. A method according to claim 1, wherein the transformation is performed by means of a learning system.

3. The method according to claim 1, wherein the transformation is performed by means of a neural network.

4. The method according to claim 2, wherein each alarm is made up of data fields, the inputs of the learning system are subdivided into groups, with each group being specifically dedicated to a respective field that is taken into account.

5. The method according to claim 4, wherein a particular input of a group of inputs is selected as a function of the content of the field to which said group is dedicated, with each input of a group corresponding to a specific content of a field.

6. The method according to claim 5, wherein at least one group of inputs is provided that is dedicated to a respective one of the following alarm fields:

a first level of alarm localization indicating a type of network equipment;

a second level of alarm localization indicating a piece of equipment;

a reason for an alarm; and

an indication of priority.

7. The method according to claim 4, wherein a detected alarm is applied to the input of the learning system by selectively setting one input in each group of inputs into a determined logic state.

8. The method according to claim 3, wherein each input corresponds to an input neurone of a neural network, the input neurones being subdivided into groups of inputs.

9. The method according to claim 2, wherein a prior stage of training the learning system is performed by correlating the inputs and outputs of the learning system on the basis of examples of alarm messages.

10. The method according to claim 2, wherein said signal V(t) is obtained in the form of the value of a binary word expressed by output neurones of the learning system.

11. The method according to claim 1, wherein said processing comprises time/frequency analysis of said signal V(t) in order to make a diagnosis.

12. The method according to claim 11, wherein said time/frequency analysis is performed by a technique of the Cohen class.

13. The method according to claim 12, wherein the Wigner Ville distribution is used to produce a bilinear distribution of the energy of said signal V(t).

14. The method according to claim 11, wherein peaks in the distribution obtained by the time/frequency analysis are localized, and respective vectors are extracted there from which correspond to identified alarm signatures.

15. The method according to claim 14, wherein said distribution is interpreted in terms of times and frequencies on a linear scale.

16. The method according to claim 1, wherein said processing comprises time/scale analysis of said treatments V(t) in order to make a diagnosis.

17. The method according to claim 16, wherein wavelet transforms are produced to change said signal V(t) into coefficients for use in making the diagnosis by means of a scalogram.

18. The method according to claim 1, wherein said processing comprises Fourier analysis of the frequency only of said signal V(t), in order to make a diagnosis on the basis of the resulting spectrum.

19. The method according to claim 18, wherein peaks in the spectrum are localized, and respective vectors are extracted therefrom corresponding to identified alarm signatures.

20. An apparatus, for analyzing alarms coming from a network, comprising:

transformation means for transforming detected alarms into a signal V(t) which expresses variation in time of a numerical value representative of all of the detected alarms;

processing means for processing the signal V(t), with usable information being extracted therefrom.