KR101946725B1 - Method and apparatus for diagnosis power cable using neural network - Google Patents

Abstract

The present invention relates to a device and method for power cable diagnosis using a neural network. To this end, the method comprises the steps of: setting, as an input layer, electrical and physical characteristic factors obtained through electrical and material tests on a power cable and the material characteristic factors in a compound; setting a plurality of output layers having at least one cumulative failure probability value in a cumulative failure probability distribution obtained through a failure test of a double interface sheet for the power cable and a failure test using a mini cable model; applying a random weight to each of the characteristic factors set as the input layer, generating a data group composed of the weighted characteristic factors, combining characteristic factors values in the data group and determining whether the combined result value converges on at least one output layer of the plurality of output layers so as to provide the data group to any one of the converged output layers, wherein the plurality of data groups are generated by changing the weight applied to each of the characteristic factors so as to converge on each of the plurality of output layers; and generating a Weibull distribution for each output layer by performing a normalization process with values of characteristic factors in the data group input to the output layer and a cable reference value of a specific reference cable. According to the present invention, the effect of each characteristic factor on the power cable can be determined.

Description

[0001] METHOD AND APPARATUS FOR DIAGNOSIS POWER CABLE USING NEURAL NETWORK [0002]

The present invention relates to an apparatus and a method for diagnosing a power cable using a neural network.

In general, the optimum design and manufacturing technology of cable compound extrusion process is developed, and the reliability evaluation technology is developed by measuring the deterioration diagnosis of cable through various experiments, and new products in the power system are developed.

In addition, by applying various types of applied voltage to cables or by measuring the dielectric strength of cables and power facilities in various ways, abnormalities such as water tree generation and voids inside the insulator are diagnosed.

In addition to the development of prototype cable manufacturing technology through various tests such as chemical / mechanical / electrical testing of semiconducting compounds related to the development of new products for cables and extrusion of interlayer / insulation layer interface sheet and tape, Proceed with evaluation.

In addition, for the development of new products for cables, it is necessary to limit the evaluation of the influence of the impulse evaluation of the mini model cable, the design factor of the power cable, and the improvement test of AC and BDS.

However, in order to proceed with the evaluation as described above, there is a problem that it takes a lot of time and cost.

Korean Patent No. 10-1223883 (registered on Jan. 11, 2013)

The present invention assigns a random weight value to converge to the cumulative failure probability value set as an output layer to each electrical, physical, chemical and material characteristic factors affecting the deterioration diagnosis of the power cable obtained through the experiment, The present invention provides a power cable diagnostic method and apparatus using a neural network capable of determining the influence of each characteristic factor on a power cable by performing normalization using an applied characteristic parameter and a specific reference cable value.

In addition, the present invention can statistically analyze the correlation of electrical, physical, chemical and material characteristic factors affecting the deterioration diagnosis of the power cable obtained through the experiment, A method and an apparatus for diagnosing a power cable using a neural network are provided.

It is to be understood, however, that the technical scope of the present invention is not limited to the above-described technical problems, and other technical problems may exist.

As a technical means for achieving the above technical object, a method of diagnosing a power cable using a neural network according to an embodiment of the present invention is a method of diagnosing a power cable by using electrical and physical characteristics obtained through electrical and material tests on a power cable, The method comprising the steps of: setting a material characteristic parameter as an input layer; determining at least one cumulative failure probability value in a cumulative failure probability distribution obtained through a failure test of a double interface sheet for the power cable and a failure test using a mini cable model The method comprising: setting a plurality of output layers; applying a random weight to each of the characteristic factors set in the input layer, and generating a data group composed of weighted characteristic factors; After combining and combining the resultant values, A plurality of output layers, and a plurality of output layers, the plurality of output layers including a plurality of output layers, a plurality of output layers, and a plurality of output layers, Generating a data group, performing a normalization process using the characteristic parameter values in the data group input to the output layer and the cable reference value of the specific reference cable, and generating the Weibull distribution for each output layer .

According to an embodiment of the present invention, the electrical and physical characteristics are at least two of room temperature tensile strength, elongation, heating, cable volume resistivity, heat loss, density and degree of crosslinking, A base resin, carbon, and an additive.

According to an embodiment of the present invention, the cumulative failure probability distribution can be obtained through AC and impulse breakdown tests using the double interface sheet of the power cable, and AC and impulse breakdown tests using the mini cable model of the power cable.

According to an exemplary embodiment of the present invention, the cumulative failure probability value is calculated by using the AC and impulse breakdown test using the double interface sheet of the power cable, the cumulative failure probability obtained through the AC and impulse breakdown tests using the mini cable model of the power cable And may be a value of 63.2%, 10% and 1% in the distribution.

As a technical means for achieving the above technical object, a power cable diagnostic apparatus using a neural network according to an embodiment of the present invention is characterized in that the electrical and physical characteristic factors obtained by electrical and material tests on a power cable, A cumulative failure probability value obtained through a failure test of a double interface sheet for the power cable and a failure test using a mini cable model, And generating a data group composed of weighted characteristic factors by applying a random weight to the characteristic factors set in each of the input layers, The combined resultant values are combined into a plurality of output layers And outputs the generated data group to any one of output layers converged by determining which one of the plurality of output layers is converged and changing a weight applied to each of the plurality of output layers so as to converge on each of the plurality of output layers, Generating a Weibull distribution by generating a Weibull distribution for each output layer by performing a normalization process using a value of a characteristic parameter in a data group input to the output layer and a cable reference value of a specific reference cable, Section.

According to an embodiment of the present invention, the electrical and physical characteristics are at least two of room temperature tensile strength, elongation, heating, cable volume resistivity, heat loss, density and degree of crosslinking, A base resin, carbon, and an additive.

According to an embodiment of the present invention, the output unit may be configured to perform the AC and impulse breakdown tests using the double interface sheet of the power cable, and the cumulative failure probability distributions obtained through the AC and impulse breakdown tests using the mini cable model of the power cable The cumulative failure probability value can be extracted.

According to an exemplary embodiment of the present invention, the output unit calculates the cumulative failure probability distribution obtained by the AC and impulse destructive test using the double interface sheet of the power cable and the AC and impulse destructive test using the mini cable model of the power cable, %, 10%, and 1% of the output layer, and set the extracted values in the plurality of output layers.

The above-described task solution is merely exemplary and should not be construed as limiting the present invention. In addition to the exemplary embodiments described above, there may be additional embodiments described in the drawings and the detailed description of the invention.

According to any one of the above-mentioned objects of the present invention, the cumulative failure probability value set as the output layer is applied to each of the electrical, physical, chemical and material characteristic factors that influence the deterioration diagnosis of the power cable obtained through the experiment A power cable diagnosis method and apparatus using a neural network that assigns a random weight to converge and performs normalization using weighted characteristic factors and specific reference cable values to determine the influence of each characteristic factor on the power cable , It is possible to set important characteristic parameter values of the prototype power cable in various ways as well as to reduce the economic cost of manpower and test equipment because data can be predicted without testing.

In addition, according to any one of the above-mentioned objects of the present invention, the correlation between electrical, physical, chemical, and material characteristic factors affecting the deterioration diagnosis of the power cable obtained through the experiment can be statistically It is possible to evaluate the influence of each of the characteristic factors, thereby improving the reliability of the electric power facility by lowering the failure and defect of the electric power cable.

1 is a block diagram illustrating a power cable diagnostic apparatus using a neural network according to an embodiment of the present invention.
2 is a flowchart illustrating a process of generating a plurality of data groups using a characteristic factor to generate a Weibull distribution according to an embodiment of the present invention.
3 is a graph for explaining influence evaluation using inverse normalization data at a volume resistivity of 20 캜 among characteristic parameters.
4 is a graph for explaining the influence evaluation using the denormalized data at a volume resistivity of 90 deg. C among characteristic parameters.
FIG. 5 is a graph for explaining the evaluation of the influence of the characteristic factors on the deterioration diagnosis of the power cable at room temperature tension.
FIG. 6 is a graph for explaining an influence evaluation on deterioration diagnosis of a power cable at a room temperature of 150 or more in a characteristic factor.
FIG. 7 is a graph for explaining the influence evaluation on the deterioration diagnosis of the power cable at a room temperature of 100 or more in the characteristic factor.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, which will be readily apparent to those skilled in the art. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

Throughout the specification of the present invention, when a part is referred to as "including " an element, it is understood that it may include other elements as well, without excluding other elements unless specifically stated otherwise.

Like reference numerals are used for similar elements in describing each drawing.

The terms first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another.

For example, without departing from the scope of the present invention, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component. The term "and / or" includes any combination of a plurality of related listed items or any of a plurality of related listed items.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art and should not be construed as ideal or overly formal meaning unless explicitly defined herein do.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an explosive detection system according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram illustrating a power cable diagnostic apparatus 100 using a neural network according to an embodiment of the present invention.

As shown in FIG. 1, the power cable diagnostic apparatus 100 using a neural network may include a factor learning unit 110 and a weir distribution generating unit 150. Here, the factor learning unit 110 may include a memory 102, an input unit 104, an operation unit 106, and an output unit 108.

The factor learning unit 110 transforms a characteristic factor for a specific power cable stored in a memory 102 into various types through learning through a neural network, and then generates a data group composed of modified characteristic factors Can be generated. Specifically, the factor learning unit 110 changes characteristic factors by applying different random weights to characteristic factors according to various situations, and determines to which cumulative failure probability value a data group composed of changed characteristic factors converge . Through this process, the factor learning unit 110 can generate a plurality of data groups converged to a plurality of accumulated failure probability values and accumulate them in the memory 102. [

The accumulated data group is provided to the weave distribution generating unit 150, and the weighed distribution generating unit 150 can generate the accumulated data group by the accumulated failure probability value and the weave distribution for the power cable.

In the memory 102, factors affecting the deterioration diagnosis of each power cable, that is, values for electrical characteristics, physical characteristics, and material characteristic factors are stored. In this case, each factor can be obtained using a very low-frequency tan delta measurement method. In the memory 102, values corresponding to respective factors are stored in corresponding indices corresponding to the respective power cables.

In the examples of the present invention, electrical characteristics and physical characteristics are related to room temperature tensile strength, elongation, heating, cable volume resistivity, heat loss, density and crosslinking degree as shown in Table 1 below, The factors are related to the base resin, carbon, additives, and the like.

[Table 1]

In addition, the memory 102 may store a data group composed of values for the changed characteristic factors output through the output unit 108. [ Specifically, the memory 102 may store a data group in which a specific cumulative failure probability value is matched with a value for the characteristic factors changed by the operation unit 106 so as to converge to a specific cumulative failure probability value.

The input unit 104 may extract the values of the characteristic factors matched to the selected power cable from the memory 102 by selecting one of the power cables and then set the value of the characteristic factors in the input layer of the input unit 104. Specifically, the input unit 104 may extract values for eighteen characteristic factors matched to the selected power cable from the memory 102, and then set the values for each input layer of the input unit 104. For this purpose, the input unit 104 may be composed of 18 input layers.

All the characteristic factor values set in the respective input layers of the input unit 104 can be input to the hidden layer of the operation unit 106 by applying a random weight value.

The arithmetic operation unit 106 includes a plurality of hidden layers for determining which of the output units 108 the output result of the arithmetic operation after giving a random weight to the values of the eighteen characteristic factors converge to one of the output layers, It contains 15 hidden layers. Here, each hidden layer receives values of 18 characteristic factors assigned different random weights, and combines the values of the input values and outputs the combined values to a certain output of the output unit 108 using a sigmoid function It is possible to judge whether or not it converges on the layer.

In addition, the arithmetic unit 106 may set a weight to be applied to the characteristic parameter value so as to converge on a specific output layer. Accordingly, each hidden layer of the arithmetic operation unit 106 applies a weight to a characteristic factor set for each of the input layers so as to converge to a specific output layer, that is, a specific cumulative probability of failure, And output it to the set output layer.

The output unit 108 may be composed of a plurality of output layers in which different cumulative failure probability values are set. Specifically, the output unit 108 may include a plurality of output layers in which different cumulative failure probability values are set based on the accumulated failure probability distribution obtained after the failure test in the commercial / nano anti-conduction model cable or the double interface sheet, May include 12 output layers with cumulative probability of failure of 63.2%, 10%, and 1%.

Each of the output layers of the output unit 108 has a cumulative failure probability value set in the output layer as a data group composed of characteristic factor values (hereinafter referred to as "weighting characteristic factor values") to which weights given from the hidden layer are applied And may be stored in the memory 102. Accordingly, the output unit 108 may receive the data group corresponding to each of the plurality of output layers and store the data group in the memory 102. [

Meanwhile, in the embodiment of the present invention, the cumulative failure probability distribution is calculated by the AC failure test using the double interface sheet of each power cable and the impulse breakdown test, the AC failure test using the mini model cable for each power cable, and the impulse failure test . Specifically, the cumulative failure probability distribution is determined by the AC breakdown test and the impulse breakdown test using the double interface sheet of each power cable, the AC failure test using the mini model cable for each power cable, and the ultra low frequency tan delta Signal. ≪ / RTI >

The factor learning unit 110 having the above-described configuration learns eighteen characteristic factors obtained through a characteristic test on a power cable through a neural network to generate a plurality of data groups, Influence can be evaluated. This will be described with reference to FIG.

FIG. 2 is a flowchart illustrating a process of generating a plurality of data groups using a characteristic factor to generate a Weibull distribution according to an embodiment of the present invention.

As shown in FIG. 2, when a specific power cable (hereinafter referred to as "power cable for influence evaluation") is selected (S200), an AC breakdown test and an impulse breakdown test using a double interface sheet for the influence evaluation power cable And AC / impulse cumulative failure probability distributions in the mini cable model obtained through the AC failure test and the impulse failure test through the mini model cable of the power cable for influence evaluation, Cumulative failure probability value for the AC characteristic and the cumulative failure probability value for the impulse characteristic, the cumulative failure probability value for the AC characteristic and the cumulative failure probability for the impulse characteristic in the mini-cable model are selected from the output section 108 (Step 202). Here, the cumulative failure probability value may be a value when the cumulative failure probability distribution is 63.2%, 10%, and 1%.

Then, electrical and physical characteristics such as room temperature tensile strength, elongation, heating condition, volume resistivity of cable, heat loss, density and crosslinking degree, and base resin, Carbon, additive, and the like from the memory 102, and sets them as input layers of the input unit 104. [0052] Specifically, a value corresponding to eighteen characteristic factors of the influence-evaluating power cable is set to each input layer of the input unit 104 (S204).

Then, the characteristic parameter values set to the respective input layers are input to one hidden layer of the operation unit 106 by applying different random weights (S206). Specifically, the random parameter values are input to one hidden layer, and the hidden layer generates a data group composed of the weighted characteristic parameter values. In addition, the characteristic parameter values Lt; / RTI >

Then, the hidden layer of the arithmetic operation unit 106 determines which output layer among the output layers of the output unit 108 has converged using the sigmoid function (S208) And a data group composed of application characteristic parameter values is provided (S210).

Then, the output layer of the output unit 108 matches the data group provided to the cumulative failure probability value set in the output unit 108 and stores the matched data group in the memory 102 (S212).

The above steps may be repeatedly performed to generate a data group for each output layer, that is, for each cumulative failure probability value. Then, the weave distribution generator 150 normalizes the data group for each cumulative failure probability value (S214) and generates a Weibull distribution based on the normalization (S216).

As described above, the characteristic factor values of the power cable for influence evaluation are changed according to the weights to be randomly set by the operation unit 106, and a decision is made as to which output layer to converge through the combination of the changed characteristic parameter values It is possible to determine the influence of each characteristic factor on the power cable for influence evaluation according to various situations. In other words, when the characteristic parameter value is changed according to the change of the weight value, it is determined which output layer converges to determine the influence of the characteristic parameter on the cable. To this end, the power cable diagnostic apparatus 100 according to the embodiment of the present invention accumulates a group of data composed of accumulated failure probability values stored in the memory 102 and weighted characteristic factor values matched thereto, that is, each accumulated failure probability value It is possible to accumulate the data groups provided to the applied output layer and generate a weave distribution for each cumulative failure probability value based on the accumulated data groups. That is, the weighted characteristic factor value when the cumulative failure probability value converged to 63.2%, 10%, and 1% according to the AC failure test and the cumulative failure probability value according to the impulse fracture test were 63.2%, 10%, and 1% The weighted characteristic parameter value when the convergence converges to 63.2%, 10%, and 1% of the cumulative failure probability value according to the AC failure test in the mini cable model, and the impulse in the mini model cable Weave distributions can be generated by accumulating the weighting factor values when the cumulative failure probability values converge to 63.2%, 10%, and 1% according to the failure test.

For generating the weave distribution, the weave distribution generator 150 of the power cable diagnostic apparatus 100 may perform a normalization process to generate a Weibull distribution. Specifically, the weave distribution generator 150 calculates a weighted characteristic factor value in a data group matched with each cumulative probability of failure stored in the memory 102 to a cable reference (hereinafter, referred to as a reference cable) The value can be normalized to a value between 0 and 1.

For this, the weave distribution generator 150 may be divided into three cases and normalized. Specifically, the weave distribution generating unit 150 performs normalization through the following Equation 1 through comparison between the weighting characteristic factor value and the cable reference value, for example, 0.5, and calculates the volume resistivity and the volume resistivity Weights are applied to weight loss factor We applied the weighted characteristic factor value and the cable reference value to denormalize through the following equation (2). To confirm the breakdown voltage of the mini model cable and double interface sheet The normalization can be performed by the following equation (3).

[Equation 1]

(Xi - min) / (X ^* - min) * 0.5 (Xi? X ^* )

^{0.5 + (Xi - X *)} / (max - X *) * 0.5 ............... (Xi> X *)

Equation (1) will be described below. When the weighting factor parameter Xi is smaller than or equal to the cable reference value X ^* , the normalization value is normalized by using the minimum value min of the weighting factor parameter values, And if not, the normalized value can be calculated by normalizing using the maximum value (max) of the weighted characteristic parameter values.

On the other hand, normalization can be performed through Equation (1) only when the characteristic factors are room temperature tensile, elongation, density, base resin, carbon and additives. Specifically, normalization can be performed by applying Equation (1) only at room temperature tensile, elongation, density, base resin, carbon, and weighting factor values corresponding to additives.

&Quot; (2) "

1 - [(Xi - min) / (X * - min) * 0.5] ............... (Xi≤X *)

1 - [0.5+ (Xi - X *) / (max - X *) * 0.5] ............... (Xi> X *)

Equation (2) will be described. Equation (2) is obtained by comparing the weighted characteristic factor value Xi with the cable reference value X ^{* in} the case of the weighted characteristic factor value for the volume resistivity and the heating loss. Normalized values can be calculated through inverse normalization.

&Quot; (3) "

0.5 + 0.4 * (Xi - min) / (max - min)

Equation (3) is an expression for normalizing a weighting factor value converging to a cumulative failure probability value of 63.2%. Since the cable reference value (X ^* ) shows a data aspect of the lowest point, .

The weave distribution generating unit 150 can obtain the data and the wavelet distribution of the AC and impulse characteristics in the double interface sheet as shown in the following Table 2 through the equations (1), (2) and (3) The data and the wavelet distribution can be obtained by the AC and impulse characteristics in the mini model cable as shown in Table 3. Specifically, the weighed distribution generating unit 150 converges to 63.2%, performs normalization using the Equation 3 in the case of the weighting applied characteristic factor value, and calculates the weighted applied characteristic factor value converged at 10% and 1% Normalized by using Equation 1 in the case of weighted characteristic factor values corresponding to room temperature tensile, elongation, density, degree of crosslinking, base resin, carbon and additive, and weighted characteristic factors converging at 10% and 1% The normalized Weibull distribution as shown in Tables 2 and 3 below can be generated by performing normalization using Equation 2 in the case of the weighted characteristic factor values corresponding to the volume resistivity and the heating loss in the values.

[Table 2]

[Table 3]

As shown in FIGS. 3 to 7, when the normalization through Equations 1, 2 and 3 as described above is denormalized to a value lower than 50 of the volume resistivity, the room temperature tensile strength is 63% Of the impulse is suddenly dropped in the double interface sheet and mini cable model of the impulse. In addition, the relationship between the volume resistivity and the breakdown voltage at 90 ° C was found to be higher when the breakdown voltage was lowered. .

It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention as defined by the appended claims and their equivalents. It will be possible. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

100: Power cable diagnostic device
102: Memory
104:
106:
108:
110: factor learning section
120: Weibull distribution generator

Claims (12)

Setting an electrical and physical characteristic factor obtained through electrical and material tests on the power cable and a material characteristic factor in the compound as an input layer;
Setting a plurality of output layers having at least one cumulative failure probability value in a cumulative failure probability distribution obtained through a failure test of a double interface sheet for the power cable and a failure test using a mini cable model,
A random weight is applied to each of the characteristic factors set to the input layer, and a data group is generated, the weight group including characteristic factors to which the weights are applied, A plurality of output layers, and a plurality of output layers, the plurality of output layers including a plurality of output layers, a plurality of output layers, and a plurality of output layers, Creating a data group,
A power cable diagnosis method using a neural network, comprising the step of performing a normalization process using values of characteristic factors in a data group input to the output layer and a cable reference value of a specific reference cable, .
The method according to claim 1,
Wherein the electrical and physical characteristics are at least two of: room temperature tensile strength, elongation, heating, volume resistivity of the cable, heat loss, density and degree of crosslinking,
Wherein the material characteristic factor is a characteristic of at least two of a base resin, carbon, and an additive.
3. The method of claim 2,
The cumulative failure probability distribution may include:
An AC and impulse breakdown test using a double interface sheet of the power cable, and a power cable diagnosis method using a neural network obtained through an AC and impulse breakdown test using a mini cable model of the power cable.
The method of claim 3,
Wherein the cumulative failure probability value
The cumulative failure probability distributions obtained through the AC and impulse breakdown tests using the double interface sheet of the power cable and the AC and impulse breakdown tests using the mini cable model of the power cable were 63.2%, 10% and 1% A Method of Diagnosing Power Cable Using Network.
5. The method of claim 4,
The generating of the Weibull distribution by output layer may include:
(Xi - min) / (max - min), Xi: the value of the characteristic factor to which the weight is applied, max: the maximum value among the characteristic factor values to which the weight is applied , min: minimum value among characteristic parameter values to which weights are applied ".
5. The method of claim 4,
The generating of the Weibull distribution by output layer may include:
In the case of the characteristic factor values, which are converged at 10% and 1%, at room temperature, elongation, density, degree of crosslinking, base resin, carbon, and weights corresponding to additives, A Method of Diagnosing Power Cable Using Network.
(Xi - min) / (X ^* - min) * 0.5 (Xi? X ^* )
^{0.5 + (Xi - X *)} / (max - X *) * 0.5 ............... (Xi> X *)
Here, X ^* is the cable reference value of the reference cable
Xi: Value of the weighted characteristic factor
max: the maximum value of the weighted characteristic parameter values,
min: The minimum value of the weighted characteristic parameter value
5. The method of claim 4,
The generating of the Weibull distribution by output layer may include:
Wherein the normalization is performed using the following equation in the case of the weighted characteristic factor value corresponding to the volume resistivity and the heating loss weight among the characteristic factors converging at 10% and 1%.
1 - [(Xi - min) / (X * - min) * 0.5] ............... (Xi≤X *)
1 - [0.5+ (Xi - X *) / (max - X *) * 0.5] ............... (Xi> X *)
Here, X ^* is the cable reference value of the reference cable
Xi: Value of the weighted characteristic factor
max: the maximum value of the weighted characteristic parameter values,
min: The minimum value of the weighted characteristic parameter value
An input unit for setting electrical and physical characteristic factors obtained through electrical and material tests on the power cable and material characteristic factors in the compound as input layers,
An output unit for setting each cumulative failure probability value in a cumulative failure probability distribution obtained through a failure test of a double interface sheet for the power cable and a failure test using a mini cable model as a plurality of output layers,
A random weight value is applied to each of the input layers, and then a weighted value is applied to a data group, and a combined result value is input to the plurality of outputs And outputs the generated data group to any one of the output layers converged by changing the weight applied to each of the plurality of output layers so as to converge on each of the plurality of output layers An operation unit for generating a plurality of data groups,
And a weir distribution generator for generating a weir distribution for each output layer by performing a normalization process using characteristic value values in a data group input to the output layer and a cable reference value of a specific reference cable, Cable diagnostics.
9. The method of claim 8,
Wherein the electrical and physical characteristics are at least two of: room temperature tensile strength, elongation, heating, volume resistivity of the cable, heat loss, density and degree of crosslinking,
Wherein the material characteristic parameter is at least two characteristics of a base resin, carbon, and an additive.
10. The method of claim 9,
The output unit includes:
The AC and impulse breakdown tests using the double interface sheet of the power cable and the neural network for extracting the cumulative failure probability values from the cumulative failure probability distributions obtained through the AC and impulse breakdown tests using the mini cable model of the power cable Power cable diagnostic device.
11. The method of claim 10,
The output unit includes:
The values of the cumulative failure probability distributions obtained through AC and impulse breakdown tests using the double interface sheet of the power cable and the AC and impulse breakdown tests using the mini cable model of the power cable were 63.2%, 10% and 1% And setting the extracted output layers to the plurality of output layers.
12. The method of claim 11,
The Weibull distribution generating unit may include:
In the case of the characteristic factor value to which the weight converged to 63.2% is applied, normalization is performed through the following Equation (3), and among the characteristic factor values to which the weight converged at 10% and 1% is applied, the room temperature tension, elongation, In the case of weighted characteristic factor values corresponding to degrees of base resin, carbon and additives, normalization is performed by the following Equation 1, and among the weighted characteristic factor values converging at 10% and 1%, volume resistivity and heating And the normalization is performed using Equation (2) in the case of the weighted characteristic parameter value corresponding to the weight loss.
[Equation 1]
(Xi - min) / (X ^* - min) * 0.5 (Xi? X ^* )
^{0.5 + (Xi - X *)} / (max - X *) * 0.5 ............... (Xi> X *)
&Quot; (2) "
1 - [(Xi - min) / (X * - min) * 0.5] ............... (Xi≤X *)
1 - [0.5+ (Xi - X *) / (max - X *) * 0.5] ............... (Xi> X *)
&Quot; (3) "
0.5 + 0.4 * (Xi - min) / (max - min)
In the above equation
X ^* : Cable reference value of reference cable
Xi: Value of the weighted characteristic factor
max: the maximum value of the weighted characteristic parameter values,
min: The minimum value of the weighted characteristic parameter value

Images