JPH04100456A - Diagnostic management system for communication network - Google Patents

Abstract

PURPOSE:To specify the faulty point of a communication network without relying on the knowledge and experience of a person (experienced) by digitizing eye pattern signals and calculating a self-learning algorithm, and then, discriminating eye patterns after classifying the eye patterns by shapes according to a classification deciding rule. CONSTITUTION:An eye pattern X digitized by an eye pattern signal digitizing section 2 is transmitted to the diagnostic information receiving section 8 of a computer 9 by means of a diagnostic information transmitting section 6. The section 8 gives the pattern X to a self-learning algorithm calculating section 3 and the section 3 calculates the loss function F. when the pattern X is classified to one class OMEGAk of N (a positive integer) pieces of classes OMEGA1, OMEGA2,...#, Wn and the average risk R of wrong classifications on all eye patterns collected so far by using the loss function Fk. After calculation, the section 8 minimizes the average risk R. An eye pattern classification discriminating section 4 specifies the faulty point of a communication network by classifying the eye patterns by shapes according to a classification deciding rule and discriminating the classified result.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a diagnostic management system for diagnosing and managing a communication network including a modulation/demodulation device, and in particular, the present invention relates to a diagnostic management system for diagnosing and managing a communication network including a modulation/demodulation device. This invention relates to a communication network diagnostic management system that uses a self-learning function to classify and judge the eye patterns of devices and identify failure locations.

[Prior Art] Conventionally, in a diagnostic management system for diagnosing and managing a communication network including this type of modem, the eye pattern of the modem is measured and projected on a screen, and the line quality and line quality of the communication network are determined by classifying the shape of the eye pattern. It was common practice to determine the characteristics of the modem and identify the fault location in the communication network (for example,
[Data transmission technology and equipment] 1 Masao Hibino et al., rNEC
"Technical Report" Vol. 39 No. 7, p. 92-9 1 Published by Nippon Electric Culture Center, July 1986; "Data Communication Using Modems and Telephone Networks", edited by Takao Hayashi, p. 54-55. (See CQ Publishing, January 1988, etc.).

However, the shape classification9 of eye patterns and the identification of faulty locations have been performed based on the knowledge and experience of experts.For example, FIG. 26B method 2400 bps)
Fig. 3(b) is an example of an eye pattern pattern (in the case of V, 26B system 2400 bps) when the 4-level fluctuation is bad. The method relied on human (expert) visual judgment of shape classification 1 and identification of failure locations.

[Problem to be solved by the invention]

In the conventional communication network diagnosis management system described above, a human (expert person) classifies and determines the shape of the eye pattern projected on the screen based on knowledge and experience to identify the failure location. The shape classification 0 judgment and the identification of the fault location can only be done by a skilled person and cannot be automated, which is a disadvantage. An object of the present invention is to provide a diagnosis management system for a communication network, which identifies fault locations by self-learning shape classification and judgment of eye patterns, which are measured values of characteristics of analog signals.

[Means to solve the problem]

As shown in FIG. 1, the communication network diagnostic management system of the present invention includes an eye pattern signal measuring section 1 that measures an eye pattern signal, and digitizing the eye pattern signal measured by the eye pattern signal measuring section 1. An eye pattern signal digitizing section 2 that performs the following steps, and a loss function when an eye pattern is classified into an arbitrary class by a self-learning algorithm using the eye pattern signal digitized by the eye pattern signal digitizing section 2. Self-learning algorithm calculation unit 3 that performs calculations so that the average risk is minimized
and an eye pattern classification/judgment unit 4 that classifies and determines the shape of the eye pattern based on the calculation results of the self-learning algorithm calculation unit 3 and identifies a fault location in the communication network.

C Effect] In the communication network diagnostic management system of the present invention, the eye pattern signal measuring section 1 measures the eye pattern signal, and the eye pattern signal digitizing section 2 converts the eye pattern signal measured by the eye pattern signal measuring section l. The self-learning algorithm calculation unit 3 uses the eye pattern signal digitized by the eye pattern signal digitization unit 2 to calculate an average using a loss function when the eye pattern is classified into an arbitrary class by the self-learning algorithm. Calculations are performed to minimize the risk, and the eye pattern classification/judgment unit 4 classifies and determines the shape of the eye pattern based on the calculation results by the self-learning algorithm calculation unit 3 to identify faulty locations in the communication network.

〔Example〕

Next, the present invention will be explained in detail with reference to the drawings.

FIG. 2 is a block diagram showing the configuration of a communication network diagnostic management system according to an embodiment of the present invention.

The communication network diagnosis management system of this embodiment includes a modulation/demodulation unit 5.
Eye pattern signal measurement section 1. a modulation/demodulation device 7 including an eye pattern signal digitization section 2 and a diagnostic information transmission section 6;
Diagnostic information receiving section 8. It is comprised of a computer 9 including a self-learning algorithm calculation section 3 and an eye pattern classification determination section 4.

The modulator/demodulator 7 has a modulator/demodulator 5 that transmits DTE (Data
Terminal Equipment) interface and line.

The variable U4 device 7 and the computer 9 are connected via a signal line through a diagnostic information transmitting section 6 and a diagnostic information receiving section 8.

Next, the operation of the communication network diagnostic management system of this embodiment configured as described above will be explained.

The eye pattern signal measurement section 1 measures the eye pattern signal from the modulation/demodulation section 5 and supplies it to the eye pattern signal digitization section 2 .

The eye pattern signal digitizing section 2 digitizes the eye pattern signal measured by the eye pattern signal measuring section 1. Specifically, the eye pattern signal digitization unit 2 digitizes one signal constituting the eye pattern as a two-dimensional vector, and converts the eye pattern into a set of n (positive integer) pieces XCt of the two-dimensional vectors X,... ,X4.・
...! 11).

The eye pattern X digitized by the eye pattern signal digitizing section 2 is transmitted by the diagnostic information transmitting section 6 to the diagnostic information receiving section 8 of the combiator 9 .

The diagnostic information receiving section 8 of the computer 9 passes the transmitted eye pattern X to the self-learning algorithm calculating section 3.

The self-learning algorithm calculation unit 3 calculates that the eye pattern
(positive integer) classes (a set in which similar eye patterns are classified as one group) Ω, 2 Ω2.・・・
Loss function (size of error when one eye pattern is classified into a certain class) Fm (X, W, , ..., W
N) (However, WI,...W8 are load matrices)
(For example, F, (X.

The average risk R of incorrect classification is calculated for all the eye patterns collected so far using calculation L and loss function F, and this average risk R Minimize, i.e.
Consider leading to R=zo. Applying the stochastic approximation method to this process, we obtain the following self-learning algorithm.

t −, W+t) “t7wmF*(X, WI, ”
', WJ1)) - 〇Here, m"l, ..., N Ek (X, WI, "', WN)" (1
ys(t)>o (squash) Generally speaking, the eye pattern classification determination unit 4 uses the class For an eye pattern at the interface between Ω and class Ω, f k−(X, WI , . . .
・, WI1) = F k(X, V/+ −W,) −
F, (X, W, ,−,W,)=0 holds, and the following classification decision rule is obtained.

Fm (X, WI, -, WN) F-(X, WI
..., Ws) < O, then XεΩ Fh (X, WI, "', WN) F-(
If X, W..., Ws) > 0, then XεΩ.

The eye pattern classification/judgment unit 4 performs shape classification and judgment based on the two classification decision rules described above, and identifies faulty locations in the communication network.

〔Effect of the invention〕

As explained above, the present invention digitizes the eye pattern signal, calculates a self-learning algorithm, and classifies and judges the shape of the eye pattern using classification decision rules, thereby eliminating reliance on the knowledge and experience of humans (experts). This has the advantage that it is possible to automatically classify and determine the shape of eye patterns without having to do so, and it is possible to identify failure points in communication networks.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a communication network diagnostic management system according to the present invention, FIG. 2 is a block diagram showing the configuration of a communication network diagnostic management system according to an embodiment of the present invention, and FIG. ) and (b
) are diagrams each showing an example of an eye pattern (in the case of V, 26B system 2400 bps). In the figure, 1...Eye pattern signal measurement unit, 2...Eye pattern signal digitization unit, 3...Self learning algorithm calculation unit, 4...Eye pattern classification determination unit, 5...Modulation/demodulation unit , 6...Diagnostic information transmitter, 7...Modulator/demodulator, 8...Diagnostic information receiver, 9...Computer.

Claims (1)

[Claims] A diagnostic management system for diagnosing and managing a communication network including a modulation/demodulation device, comprising: an eye pattern signal measuring section that measures an eye pattern signal; and digitizing the eye pattern signal measured by the eye pattern signal measuring section. and an eye pattern signal digitization unit that performs the eye pattern signal digitization unit, and a self-learning algorithm using the eye pattern signal digitized by the eye pattern signal digitization unit to calculate the average risk using a loss function when the eye pattern is classified into an arbitrary class. a self-learning algorithm calculation unit that performs calculations so that the error is minimized; and an eye-pattern classification determination unit that classifies and determines the shape of the eye pattern based on the calculation results of the self-learning algorithm calculation unit and identifies failure points in the communication network. A communication network diagnostic management system comprising: