JPH08247897A - Automatically analyzing method for optical line characteristics - Google Patents

Abstract

PURPOSE: To automatically calculate connection loss even for short connection intervals by using a back propagation type neural network corresponding to the connection intervals even for the near section of connector connecting point intervals. CONSTITUTION: Measurement conditions are optimized in a control arithmetical part 3 in an optical pulse testing part 2 to measure and analyze the measurement conditions in the testing part 2 and in a data processing part 4, respectively. A measured data sequence is fed into the processing part 4. A differential coefft. sequence at each point is prepared from the data sequence. A connector connecting position and a difference number sequence at a point of distance intervals, a half of pulse width, are obtained based on the width of Fresnel reflection from the differential coefft. sequence. A fusion connecting point position is obtained based on splice from the differential coefficient to obtain a connection point section by using each connecting point position as a reference. A normalized data sequence is inputted to the input layer of a neutral network. A value form an output layer is computed to obtain connector connecting loss. Finally, analyzed results, the connection loss of an optical line, cable loss, reflection amount, their generated positions, and a line loss failure position, are displayed on a display part 6 to terminate analysis.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to an optical pulse tester (OT).
DR) is used to measure the optical characteristics of the optical line and the computer is used to automatically analyze the measurement data to determine the connection loss of the optical line, the cable loss, the amount of reflection, the location where these occur and the location where the failure occurs. It is about how to display.

[0002]

2. Description of the Related Art Conventionally, as a method of automatically analyzing the connection loss of an optical line, a method of obtaining it from a step of an approximate straight line in both sections of a connection point is known (for example, Nakamura et al. Analytical method ", Proceedings of the 1990 IEICE Spring National Convention, B-899).

In the conventional automatic analysis method, the first processing step of appropriately setting and measuring the measurement conditions of the optical pulse tester according to the target optical line, and the waveform data obtained in the first processing step From the above, a second processing step for estimating the light receiving level at the start point of the optical line with an approximate straight line in the data section missing in the mask at the near end portion, and normalizing all the data by setting the light receiving level at the start point to 0 dB; A differential waveform is obtained by calculating the difference of the converted data with a half of the pulse width, and the point at which the sign of this differential waveform changes from positive to negative is regarded as the peak of Fresnel reflection, and the third processing is performed as the connector connection point. Process and third
The differential waveform obtained in the processing step is further calculated to obtain the differential waveform, and the point at which the sign of this differential waveform changes from positive to negative or from negative to positive is regarded as the peak due to the splice, and is defined as the fusion splicing point. From the fourth processing step to perform, the fifth processing step using the last Fresnel reflection as the cable end, and the respective connection point positions obtained in the third and fourth processing steps, in the case of fusion splicing, splice In the case of connector connection, only the width, the sixth processing step in which the section including the width and the tail of the Fresnel reflection is the connection point section, and the cable section is between the connection point sections, and the approximate value of each cable section is at least 2 Multiplied,
The seventh processing step of obtaining each cable loss from the difference between the values on each approximation straight line at the position of the connection point and each connection loss due to the fusion splicing point and the connector connection point, and approximating the cable sections before and after each connection point. Eighth processing step obtained from a straight line step, ninth processing step of correlating connection point distance information of a database composed of connection point type and distance information and detection result with position and connection loss, and each connector connection The tenth processing step in which the difference between the peak value of the data in the point section and the approximate straight line is the height of the Fresnel reflection, and the total line loss is calculated from the difference between the two points from the cable start point to the cable end.
And the processing steps of.

The characteristic of the above-mentioned conventional analysis method is that noises generated inside the optical pulse tester and rotation of the polarization plane of the optical fiber are obtained by obtaining the approximate straight line of the front and rear cable sections by the least square method for calculating the characteristic loss. The connection loss could be calculated without being affected by the accompanying noise. In addition, the interval between connector connection points that can determine the connection loss is reported to be 200 m when the pulse width is 500 ns (Enomoto et al., "Study on automatic analysis accuracy of OTDR waveform including Fresnel reflection", 1994 Electronic Information and Communication Proceedings of the National Spring Conference of the Society, B-982).

[0005]

However, in the conventional analysis method, when the distance between the connector connecting points becomes closer, the tail of the Fresnel reflection affects the rear connector connecting points, so that the cable section is secured. The connector connection loss could not be calculated from the approximate straight line.

An object of the present invention is to provide an optical line automatic analysis method capable of obtaining and displaying the connection loss, cable loss, and reflection amount of optical lines whose connector connection points are close to each other, and the position where these are generated and the position where a failure occurs. Especially.

[0007]

In order to achieve the above object, according to claim 1 of the present invention, an optical pulse tester is used to measure optical characteristics including distance and light intensity of an optical line, and the obtained measurement is performed. In the automatic analysis method of optical line characteristics that automatically displays the connection loss, cable loss, and reflection amount of the optical line from the data, and the position where these are generated and the position where a failure occurs, the measurement conditions are adjusted according to the target optical line. A first processing step in which optical characteristics are measured using an optimally set optical pulse tester and measurement data is captured, and a second processing step in which connector connection points and fusion splicing points are detected from the measurement data, and connection If there is a section where the connector connection point intervals are close to each other by referring to a database composed of point types and distance information, a value obtained by normalizing the light reception levels at a plurality of consecutive positions including the connector connection point. A third processing step of calculating the connection loss at the connector connection point by using a back propagation type neural network corresponding to the connector connection point interval, which uses the value obtained by normalizing the connection loss at the connector connection point as an input And a fourth process of obtaining each splice loss of the fusion splicing point and the connector splice point other than the splice point for which the splice loss was obtained in the third treatment step from the step of the approximate straight line of the cable section before and after each splice point Step, a fifth processing step of correlating the connection point distance information of the database with the detection result, a sixth processing step of obtaining the cable loss, the reflection amount at the connector connection position, and the total line loss, and the cable loss, Displaying the connection loss due to fusion splicing and connector connection, the amount of reflection at the connector connection position, the position where these occur, the total line loss, the position where a failure has occurred, etc. We propose an automatic analysis method for an optical line characteristics including the process steps.

Further, in claim 2 of the present invention, instead of the third processing step, if there is a section in which the connector connection point intervals are close to each other by referring to the database, the cable section before and after the connector connection point section is detected. An eighth processing step for calculating the loss in the connector connection point section including the connector connection loss for two points, the light receiving levels at a plurality of consecutive positions including the connector connection point, and the connection loss in the connector connection point section. Using a back propagation type neural network corresponding to the connector connection point interval, which uses the normalized value as the input and the connection loss at each connector connection point as the output, the connection loss at each connector connection point And a ninth processing step for calculating.

Further, in claim 3 of the present invention, instead of the ninth processing step, the received light levels at a plurality of consecutive positions including the connector connection point and the connection loss in the connector connection point section are normalized. The input, the output of the ratio of the front connector connection point of the connection loss of the connector connection point section, using a back propagation type neural network corresponding to the connector connection point interval,
A method for automatically analyzing optical line characteristics according to claim 2, which includes a tenth processing step of calculating a connection loss at each connector connection point.

[0010]

According to the first aspect of the present invention, when there is a section in which the intervals of the connector connection points are close to each other, it has been impossible to analyze by using the back propagation type neural network corresponding to the interval of the connector connection points. Even with a short connector connection point interval, the connection loss can be automatically calculated with high accuracy, and can be displayed together with the cable loss of the optical line, the reflection amount, the position where they occur, and the position where the failure occurs.

According to a second aspect of the present invention, a value obtained by normalizing the light receiving levels at a plurality of consecutive positions including the connector connection point and the connection loss in the connector connection point section is used as the input of the neural network. Thus, the connection loss can be calculated automatically with high accuracy.

According to the third aspect of the present invention, the ratio of the front connector connection point to the connection loss in the connector connection point section is set as the output of the neural network, so that the connection loss can be automatically increased with higher accuracy. It can be calculated and the configuration of the neural network can be simplified.

[0013]

[Embodiment 1] FIG. 1 shows the configuration of a system for carrying out the method for automatically analyzing the characteristics of an optical line according to the present invention. In the figure, 1 is an optical line, 2 is an optical pulse test unit (tester), and 3 is A control / arithmetic unit, 4 is a data processing unit, 5 is a database, 6 is a display unit, and 7 is a data bus.

First, the measurement principle of the optical pulse test section 2 will be described. When light is incident on the optical fiber, the light returning from the middle of the optical fiber to the incident end is the Fresnel reflected light from the connector connection point and the backscattered light that returns part of the Rayleigh scattered light that occurs in the optical fiber. There is.

Here, when an optical pulse obtained by the pulse generator and the semiconductor laser is incident on the optical fiber, the backscattered light and Fresnel reflected light generated in the optical fiber are
It returns to the entrance end after a time proportional to the distance from the entrance end to each generation position. A waveform can be obtained by converting the returned light into an electric signal by a light receiving element. In the embodiment, the light receiving level, which is logarithmically converted from the light receiving power received by the optical pulse test unit 2 and is displayed in decibel (dB), is shown on the vertical axis and the distance is shown on the horizontal axis.

2 and 3 are flow charts showing a first embodiment of the automatic optical path characteristic analyzing method of the present invention. The automatic analysis method of the optical line characteristic of this embodiment will be described with reference to FIGS. 1, 2, 3, 4, and 5.

First, in the control / arithmetic unit 3, the optical pulse test unit 2 is instructed to optimize the measurement conditions via the data bus 7 (step SA1). The point that the measurement conditions are optimized is the same as the conventional one. Next, an optical pulse test is carried out (step SA2).

Next, the data measured by the optical pulse test section 2 is analyzed by the data processing section 4 as described below (steps SA3 to SA17).

First, the measurement data string (Y, X) is loaded into the data processing unit 4 (step SA3). Here, Y is a light receiving level (dB), and X is a point with respect to distance.
The actual distance is X multiplied by the distance ΔL of 1 point. Next, the differential coefficient dY / dX at each point is obtained from the measurement data string (Y, X), and the differential coefficient string (dY /
dX, X) is created (step SA4).

Next, whether or not each detected singular point is a peak of Fresnel reflection generated at the connector connection point is defined as a singular point at which the sign of the differential coefficient dY / dX changes from plus or zero to minus. , The shape is used for the judgment. As a result, when it is determined that there is Fresnel reflection, the position of the singular point is set to the position Xf _l0 of the connector connection point (step S
A5). However, l0 = 1, 2, ..., And 10 indicates the number of the connector connection point as seen from the start position of the optical line.

Next, a difference sequence (ΔY, X) of points having a distance interval corresponding to half the pulse width is created (step SA6), and a differential coefficient sequence (dΔ) of the difference sequence (ΔY, X).
Y / dX, X) is created (step SA7).

Next, each of the extracted singular points is generated at the fusion splicing point, with a point where the sign of the differential coefficient sequence dΔY / dX changes from plus or zero to minus, or minus or zero to plus. Whether or not it is the peak of the spliced is determined from its shape. As a result, when the splice is determined, the position of the singular point is set as the position Xk _l1 of the fusion splicing point (step SA8). However, l1 =
1, 2, ..., And 11 indicates the number of fusion splicing point as seen from the start position of the optical line.

Next, based on the positions Xf _l0 and Xk _l1 of the respective connection points, the splice width in the case of fusion splicing, the width of the Fresnel reflection in the case of connector connection, and the section including the skirt are the connection point sections. And In addition, a cable section is defined between each connection point section (step SA9).

Next, the connection point distance information and the detection results Xf _l0 , X of the database 5 created from the design document at the time of construction of the track.
Corresponding to _k11 , it is determined whether there is a connector connection point interval of T (m) or less (step SA10). In addition, T
Is experimentally obtained for each pulse width.

When the above-mentioned determination result is Yes, Xf _{s at the} s-th and s + 1-th connector connection points when the distance between the connector connection points is T (m) or less as viewed from the start position of the optical line.
Received light level Yf _s −m from −m0 to Xf _s + m1
From 0 to Yf _s + m1, normalization processing is performed with the maximum light receiving level Ymax _{s set} to 1 and the minimum light receiving level Ymin _{s set} to 0, and a normalized data string Ii is obtained. However, i = 1, 2, ...
(M0 + m1 + 1). FIG. 4 shows an example of the waveform data in the optical pulse test section, and FIG. 5 shows the normalized waveform data.

Then, the normalized data sequence Ii is input to the input layer of the neural network obtained by learning in advance, and the value obtained by multiplying the output value O1 from the output layer of the neural network by (Ymax _s -Ymin _s ) is obtained. The sth connector connection loss Lf _s (dB) and a value obtained by multiplying the output value O2 by (Ymax _s −Ymin _s ) are multiplied by the s + 1th connector connection loss Lf _{s + 1} (d).
B) (step SA11). It should be noted that m0 and m1 are experimentally obtained in advance for each pulse width, reading resolution, and connector connection point interval.

On the other hand, when the determination result is No, the approximate straight line of each cable section is obtained by the least squares method, and each cable loss is obtained from the difference between the values on the approximate straight line at the position of the connection point (step SA12).

Next, the connector connection loss and the fusion splicing loss other than those obtained by the neural network are obtained from the steps of the approximate straight line of the cable section before and after each connection point (step SA13).

Next, the connection point distance information of the database 5 and the detection result are made to correspond by the position and the connection loss (step SA
14) _{Find the} Fresnel reflection height Rf ₁₀ from the difference between the peak value of the data at each connector connection point section and the approximate straight line (step SA15), and calculate the total line loss from the difference between the cable start point and the cable end. Is calculated (step SA16).

Finally, the analysis result is displayed on the display unit 6 as the analysis result, the connection loss of the optical line, the cable loss, the reflection amount, the position where these are generated, the total line loss and the position where the failure occurs (step SA17). .

Next, the structure of the neural network used in step SA11 will be described. In the present invention, a backpropagation neural network having a three-layer structure is used. The number of units in the input layer is (m0 + m1 +
1). The output layer is composed of two units. The calculation method of the neural network will be described below.

The input / output relationship of the units in the input layer is linear, and the units in the intermediate layer and the output layer are sigmoid functions, that is, input / output defined by f (p) = 1 / (1 + exp (-2p / u0)). Have a relationship However, p is an input of a sigmoid function, u0 is a parameter which determines the shape of a sigmoid function.

At this time, the output Hj of the intermediate layer unit j of the neural network is It is expressed as However, Wji indicates the coupling coefficient of the input layer and the intermediate layer, and θj indicates the offset of the intermediate layer.

The output Ok of the output layer unit k is It is expressed as However, Vj represents the coupling coefficient of the intermediate layer and the output layer, and γk represents the offset of the output layer.

In learning the neural network, the coupling coefficients Wji and Vkj are calculated from the output value Ok and the teacher signal Otk.
And the offsets θj and γk are changed. The changing method will be described below.

First, an error .delta.k is calculated from the output value Ok and the teacher signal Otk as .delta.k = (Ok-Otk) .Ok. (1-Ok).

Next, from the error δk, the coupling coefficient Vkj of the intermediate layer and the output layer, and the output Hj of the intermediate layer, the coupling coefficient connected to the intermediate layer unit j and the error σj with respect to the offset of the intermediate layer unit are calculated. And ask.

Next, the product of the error δk of the output layer unit k, the output Hj of the intermediate layer unit j and the constant α is added to obtain the coupling coefficient Vkj from the intermediate layer unit j to the output layer unit k, and The offset γ k of the output layer unit k is corrected by adding the product of the error δ k and the constant β. Vkj = Vkj + αδkHj γk = γk + βδk Further, the product of the error σj in the intermediate layer unit j and the output Ii of the input layer unit i and the constant α is added to calculate from the input layer unit i to the intermediate layer unit. Coupling coefficient Wji connected to j
And the offset θj of the intermediate layer unit j is corrected by adding the product of the error σj and the constant β. Wji = Wji + α · σj · Ii θj = θj + β · σj The coupling coefficient Wji, Vkj, offset θj,
Change γ k. The coupling coefficients Wji, Vkj and the offsets .theta.j, .gamma.k must be learned and calculated for each pulse width, reading resolution, and connector connection point interval.

Next, an example in which a neural network is used to calculate the connection loss of adjacent connector connection points will be shown using actual data.

FIG. 6 shows an experimental system for obtaining the waveform data of the optical pulse test section where Fresnel reflection is overlapped. Waveform data of the optical pulse test section with overlapping Fresnel reflection was obtained by measuring three types of simulated lines with connector connection point intervals of 30 m, 50 m, and 100 m. At the time of measurement, the connector connection loss before and after was given a loss from 0.5 to 3.0 dB at intervals of 0.5 dB, and a total of 36 types of waveforms were obtained. This 3
Of the 6 types, 18 types were used for learning the neural network, and the remaining 18 types were used as evaluation data for obtaining the accuracy of the neural network after learning. The measurement conditions of the optical pulse test unit are a distance range of 10 km and a pulse width of 20.
ns, 500 ns, wavelength 1.55 μm, number of data is 50
00.

FIG. 7 shows an example of the waveform data of the optical pulse test section when the connector connection point interval is 100 m, and FIG. 8 shows an example of the waveform data of the optical pulse test section when the connector connection point interval is 50 m. 9 shows examples of waveform data of the optical pulse test unit when the connector connection point interval is 30 m.

FIG. 10 shows the configuration of the neural network used in this embodiment, and FIG. 11 shows the number of units. In this embodiment, the connector connection point interval is 3 from the waveform data of the optical pulse test section when the pulse width is 500 ns.
A neural network for calculating the connection loss at 0 m, 50 m, and 100 m was created.

The section for performing the normalization process is m0 = 10, m1 = 109, 50 when the connector connection point interval is 100 m.
When m, m0 = 10, m1 = 89, and when 30m, m0 =
10, m1 = 69. The teacher signal Otk used for learning is the connector connection loss before and after obtained from the waveform data of the optical pulse test section at 20 ns (Ymax _s −Ymin _s )
The value divided by was used.

The analysis accuracy of the neural network after learning is the connector connection loss obtained by the neural network from the waveform data of the optical pulse test section at 500 ns and the connection loss obtained from the waveform data of the optical pulse test section at 20 ns. The error of loss is shown by cumulative relative frequency.

FIG. 12 shows the analysis accuracy of the neural network at each connector connection point interval. 100m, 5
It can be seen that the analysis accuracy is sufficient for the evaluation of the optical line, with 0 m being within 0.3 dB. 0.5 at 30m
Within dB, it is a sufficient value for evaluating whether or not the connection point is out of order.

In the evaluation of the optical line, when the analyzable connector connection point interval is a pulse width of 500 ns, it was 50 m in the present invention, which was 200 m in the conventional case,
It can be seen that the resolution is improved.

[0047]

[Embodiment 2] FIG. 13 is a flowchart showing a second embodiment of the present invention. Here, step SA11 in FIG. 3 showing the first embodiment is replaced with other steps SB1 and SB2.
Is replaced with. That is, step SA1 in FIG.
If Yes in 0, the process proceeds to steps SB1 and SB2, and if No, the process proceeds to step SA12. Hereinafter, steps SB1 and SB2 will be described. Note that step SA1
Since the steps from SA to SA10 are the same as those in the first embodiment, the description thereof will be omitted.

In step SB1, the s-th and S-th distances from the start position of the optical line when the connector connection point interval is T (m) or less.
At the + 1st connector connection point, an approximate straight line of the cable section before and after the connector connection point section is obtained by the least squares method, and the connector connection including the connector connection loss for two points is obtained from the step difference of the connection point position on each approximate straight line. Loss of point section Lmix _s
Ask for. FIG. 14 shows an example of waveform data in the optical pulse test section.

In step SB2, Xf _s -m0 to Xf _s
Light receiving level Yf _s −m0 to Yf _s + m in the section up to + m1
In 1, the normalization processing is performed by setting the maximum light receiving level Ymax _s to 1 and the minimum light receiving level Ymin _s to 0, and the normalized data string Ii is obtained. However, i = 1, 2, ... (m0 + m
1 + 2), and I _{m0 + m1 + 2} is a value obtained by dividing the normalized loss Lmix _s of the connector connection point section by (Ymax _s −Ymin _s ).

Then, the normalized data string Ii is input to the input layer of the neural network, which has been learned in advance, and the value obtained by multiplying the output value O1 from the output layer by (Ymax _s -Ymin _s ) is connected to the connector. Loss Lf _s , output value O2 (Ymax _s
-Ymin _s ) is taken as the connector connection loss Lf _{s + 1} .

After the end of step SB2, the process proceeds to step SA12 of FIG. The processes after step SA12 are the same as in the case of the first embodiment, so the explanation is omitted.

Next, the structure of the neural network used in step SB2 will be described. In the present invention, 3
A back-propagation neural network with a layered structure is used. The number of units in the input layer is (m0 + m1 +
2). The output layer is composed of two units. The calculation method of the neural network is the same as that of the first embodiment, so the explanation is omitted.

Next, an example in which a neural network is used to calculate the connection loss of adjacent connector connection points will be shown using actual data.

FIG. 6 shows an experimental system for obtaining the waveform data of the optical pulse test section in which Fresnel reflection is overlapped. Waveform data of the optical pulse test section with overlapping Fresnel reflection was obtained by measuring three types of simulated lines with connector connection point intervals of 30 m, 50 m, and 100 m. At the time of measurement, the connector connection loss before and after was given a loss from 0.5 to 3.0 dB at intervals of 0.5 dB, and a total of 36 types of waveforms were obtained. This 3
Of the 6 types, 18 types were used for learning the neural network, and the remaining 18 types were used as evaluation data for obtaining the accuracy of the neural network after learning. The measurement conditions of the optical pulse test unit are a distance range of 10 km and a pulse width of 20.
ns, 500 ns, wavelength 1.55 μm, number of data is 50
00. The process up to this point is the same as in the case of the first embodiment.

FIG. 15 shows the configuration of the neural network used in this embodiment, and FIG. 16 shows the number of units. In this embodiment, from the waveform data of the optical pulse test section when the pulse width is 500 ns, the connector connection point interval is 30 m,
A neural network for calculating the connection loss at 50 m and 100 m was created.

The interval for normalization processing is m0 = 10, m1 = 108, 50 when the connector connection point interval is 100 m.
When m, m0 = 10, m1 = 88, and when 30m, m0 =
10, m1 = 68. The teacher signal Otk used for learning is the connector connection loss before and after obtained from the waveform data of the optical pulse test section at 20 ns (Ymax _s −Ymin _s )
The value divided by was used.

The analysis accuracy of the neural network after learning is the connector connection loss obtained by the neural network from the waveform data of the optical pulse test unit at 500 ns and the connection loss obtained from the waveform data of the optical pulse test unit at 20 ns. The error of loss is shown by cumulative relative frequency.

FIG. 17 shows the analysis accuracy of the neural network at each connector connection point interval. 100m, 5
It can be seen that the analysis accuracy is sufficient for the evaluation of the optical line, with 0 m being within 0.3 dB. 0.5 at 30m
Within dB, it is a sufficient value for evaluating whether or not the connection point is out of order.

In the evaluation of the optical line, when the analyzable connector connection point interval is a pulse width of 500 ns, it is 50 m in the present invention, which was 200 m in the prior art.
It can be seen that the resolution is improved.

[0060]

[Third Embodiment] FIG. 18 is a flow chart showing a third embodiment of the present invention. Here, FIG. 13 shows a second embodiment.
The step SB2 in step 2 is replaced with another step SC1. That is, after completion of step SB1 in FIG. 13, the process proceeds to step ASC1. Hereinafter, step SC1 will be described. Note that steps SA1 to SB1 are the first
The description is omitted because it is the same as that of the second embodiment.

In step SC1, from Xf _s -m0 to Xf _s
Light receiving level Yf _s −m0 to Yf _s + in the section up to + m1
In m1, the normalization processing is performed by setting the maximum light receiving level Ymax _s to 1 and the minimum light receiving level Ymin _s to 0, and the normalized data string Ii is obtained. However, i = 1, 2, ... (m0 +
m1 + 2), and I _{m0 + m1 + 2} is the loss Lmix _s of the connector connection point section obtained in step SB1 (Ymax _s −Ymin _s ).
Divide by.

Then, the normalized data string Ii is input to the input layer of the neural network, which has been learned in advance, and the value obtained by multiplying the output value O1 from the output layer by the loss Lmix _s in the connector connection point section is used as the connector. Splice loss Lf _s , (1-O
The value obtained by multiplying 1) by the loss Lmix _s in the connector connection point section is defined as the connector connection loss Lf _{s + 1} .

After step SC1 is completed, the process proceeds to step SA12 in FIG. The processes after step SA12 are the same as in the case of the first embodiment, so the explanation is omitted.

Next, the structure of the neural network used in step SC1 will be described. In the present invention, 3
A back-propagation neural network with a layered structure is used. The number of units in the input layer is (m0 + m1 +
2). The output layer is composed of one unit. The calculation method of the neural network is the same as that of the first embodiment, so the explanation is omitted.

Next, an example in which a neural network is used to calculate the connection loss of adjacent connector connection points will be shown using actual data.

FIG. 6 shows an experimental system for obtaining the waveform data of the optical pulse test section in which the Fresnel reflection is overlapped. Waveform data of the optical pulse test section with overlapping Fresnel reflection was obtained by measuring three types of simulated lines with connector connection point intervals of 30 m, 50 m, and 100 m. At the time of measurement, the connector connection loss before and after was given a loss from 0.5 to 3.0 dB at intervals of 0.5 dB, and a total of 36 types of waveforms were obtained. This 3
Of the 6 types, 18 types were used for learning the neural network, and the remaining 18 types were used as evaluation data for obtaining the accuracy of the neural network after learning. The measurement conditions of the optical pulse test unit are as follows: distance range 10 km, pulse width 20
ns, 500 ns, wavelength 1.55 μm, number of data is 50
00. The process up to this point is the same as in the case of the first embodiment.

FIG. 19 shows the structure of the neural network used in this embodiment, and FIG. 20 shows the number of units. In this embodiment, from the waveform data of the optical pulse test section when the pulse width is 500 ns, the connector connection point interval is 30 m,
A neural network for calculating the connection loss at 50 m and 100 m was created.

The cut-out width when performing the normalization processing is m0 = 10 and m1 = 1 when the connector connection point interval is 100 m.
At 08 and 50 m, m0 = 10, m1 = 88, and at 30 m, m0 = 10 and m1 = 68. The teacher signal Otk used for learning is the connector connection loss before and after (Ymax _s) obtained from the waveform data of the optical pulse test section at 20 ns.
-Ymin _s ).

The analysis accuracy of the neural network after learning is the connector connection loss obtained from the neural network from the waveform data of the optical pulse test section at 500 ns and the connection loss obtained from the waveform data of the optical pulse test section at 20 ns. The error of loss is shown by cumulative relative frequency.

FIG. 21 shows the analysis accuracy of the neural network at each connector connection point interval. 100m, 5
It can be seen that the analysis accuracy is sufficient for the evaluation of the optical line, with 0 m being within 0.2 dB. 0.4 at 30 meters
Within dB, it is a sufficient value for evaluating whether or not the connection point is out of order.

In the evaluation of the optical line, when the analyzable connector connection point interval is a pulse width of 500 ns, it is 50 m in the present invention, which is 200 m in the conventional case.
It can be seen that the resolution is improved. As described above, the present invention has been specifically described based on the embodiments.
It is needless to say that the present invention is not limited to this, and various changes can be made without departing from the scope of the invention.

[0072]

As described above, according to the first aspect of the present invention, when there is a section in which the connector connection point intervals are close to each other, the back propagation type neural network corresponding to the connector connection point intervals is used. ,
The connection loss can be automatically calculated with high accuracy even at a short connector connection point interval that could not be analyzed in the past, and it is displayed together with the cable loss of the optical line, the reflection amount, the position where they occur, and the position where the failure occurs. be able to.

According to claim 2 of the present invention, the light receiving levels at a plurality of consecutive positions including the connector connection point,
The connection loss can be automatically and highly accurately calculated by inputting a value obtained by normalizing the connection loss in the connector connection point section to the neural network.

According to the third aspect of the present invention, the connection loss is automatically further increased by setting the ratio of the front connector connection point in the connection loss in the connector connection point section as the output of the neural network. It can be calculated with high accuracy and the structure of the neural network can be simplified.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a system for carrying out an automatic optical line characteristic analysis method according to the present invention.

FIG. 2 is a flowchart showing a first embodiment of an automatic optical line characteristic analyzing method according to the present invention.

FIG. 3 is a flowchart showing a first embodiment of an automatic optical line characteristic analyzing method according to the present invention.

FIG. 4 is a diagram showing an example of waveform data in an optical pulse test section of an optical line including adjacent connector connection points.

5 is a diagram in which the waveform data of FIG. 4 is normalized.

FIG. 6 is a diagram showing an experimental system for obtaining waveform data of an optical pulse test section with overlapping Fresnel reflections.

FIG. 7 is a diagram showing an example of waveform data of the optical pulse test unit when the connector connection point interval is 100 m.

FIG. 8 is a diagram showing an example of waveform data of an optical pulse test unit when the connector connection point interval is 50 m.

FIG. 9 is a diagram showing an example of waveform data of an optical pulse test unit when the connector connection point interval is 30 m.

FIG. 10 is a configuration diagram of a neural network used in the first embodiment.

FIG. 11 is a diagram showing the number of units in the input layer, the intermediate layer, and the output layer of the neural network used in the first embodiment.

FIG. 12 is a diagram showing analysis accuracy after learning in the neural network according to the first embodiment.

FIG. 13 is a flowchart showing a second embodiment of the automatic optical line characteristic analysis method of the present invention.

FIG. 14 is a diagram showing an example of waveform data in an optical pulse test section of an optical line including adjacent connector connection points.

FIG. 15 is a configuration diagram of a neural network used in the second embodiment.

FIG. 16 is a diagram showing the number of units in the input layer, intermediate layer, and output layer of the neural network used in the second embodiment.

FIG. 17 is a diagram showing analysis accuracy after learning in the neural network according to the second embodiment.

FIG. 18 is a flowchart showing a third embodiment of the optical line characteristic automatic analysis method of the present invention.

FIG. 19 is a block diagram of the neural network used in the third embodiment.

FIG. 20 is a diagram showing the number of units in the input layer, the intermediate layer, and the output layer of the neural network used in the third embodiment.

FIG. 21 is a diagram showing analysis accuracy after learning in the neural network according to the third embodiment.

[Explanation of symbols]

1 ... Optical line, 2 ... Optical pulse test unit, 3 ... Control / arithmetic unit,
4 ... Data processing unit, 5 ... Database, 6 ... Display unit, 7
... data bus.

Claims (3)

[Claims] 1. An optical pulse tester is used to measure optical characteristics including the distance and light intensity of an optical line, and the connection loss, cable loss, and reflection amount of the optical line are automatically calculated from the obtained measurement data. In the automatic analysis method of the optical line characteristics that finds and displays the generation position and the failure generation position, the optical characteristics are measured using the optical pulse tester that is optimally set according to the measurement conditions and the measured data. The first processing step of capturing the data, the second processing step of detecting the connector connection points and the fusion splicing points from the measurement data, and the database of connection point types and distance information If there are adjacent sections, the normalized value of the received light level at a plurality of consecutive positions including the connector connection point is input, and the normalized connection loss of the connector connection point is output. A third processing step of calculating the connection loss at the connector connection point by using a back propagation type neural network corresponding to the connector connection point interval, and a connection point other than the connection point for which the connection loss was obtained in the third processing step. A fourth processing step of obtaining each connection loss of the fusion splicing point and the connector connection point from the step of the approximate straight line of the cable section before and after each connection point, and making the connection point distance information of the database correspond to the detection result. 5, the cable loss, the amount of reflection at the connector connection position,
A sixth processing step of obtaining the total line loss, and a step of displaying the cable loss, the connection loss due to the fusion splicing and the connector connection, the reflection amount at the connector connection position, the position where these are generated, the total line loss and the position where the failure occurs 7
A method for automatically analyzing optical line characteristics, comprising: 2. In place of the third processing step, when there is a section in which the connector connection point intervals are close to each other by referring to the database, the connector connection loss of two points is calculated from the cable section before and after the connector connection point section. The eighth processing step of calculating the loss in the connector connection point section including the input, a normalized value of the light reception level at a plurality of consecutive positions including the connector connection point, and the connection loss in the connector connection point section are input. A ninth processing step of calculating the connection loss of each connector connection point using a back propagation type neural network corresponding to the connector connection point interval, which outputs a value obtained by normalizing the connection loss of each connector connection point The method for automatically analyzing optical line characteristics according to claim 1, further comprising: 3. In place of the ninth processing step, a value obtained by normalizing a light reception level at a plurality of consecutive positions including a connector connection point and a connection loss in the connector connection point section is input, and the connector is used. Output is the ratio of the front connector connection point to the connection loss in the connection point section.
The optical line characteristic automatic analysis according to claim 2, further comprising a tenth processing step of calculating a connection loss at each connector connection point by using a back propagation type neural network corresponding to a connector connection point interval. Method.