Method and system for identifying wear degree of contact for switch

Abstract

The invention discloses a method and a system for identifying the wear degree of a contact for a switch. The burning area can cause the abrasion of the contact for the switch, and the abrasion of the contact can cause the burning of the contact for the switch. RGB values on the image are detected, and a burning area on the surface of the contact for the switch is found through clustering. And the contact position of the switch contact also affects the burning area, so that a superposed image of the distance matrix and the burning image of the contact is constructed, and the abrasion degree is judged by jointly using the burning condition. And the distance matrix can also analyze surface scratches and wear traces so as to judge the wear condition. And the abrasion of the switch contact can influence the resistance, so that the abrasion degree can be judged according to the change of the resistance.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for identifying the degree of wear of a switch contact. Background Art

[0002] With the widespread use of electrical switching equipment, the contact surface of the switch contacts will gradually wear out during long-term switching operations due to the influence of factors such as current, temperature, and contact pressure. Contact resistance is one of the important factors affecting switch contact burning. The increase in contact resistance is often related to contact wear, so the change in contact resistance can be monitored to determine the degree of wear. When the switch is in the closed state, the contact resistance between the contacts should be as low as possible to reduce heat. However, if the contact surface is rough or oxidized, or the contact surface is uneven due to aging, wear, etc., the contact resistance will increase. The increased contact resistance causes the current passing through the contact to generate higher heat at the contact point, which can easily cause burning.

[0003] The degree of contact wear directly affects the reliability and safety of the switch. Traditional contact wear detection methods usually rely on manual inspection or use sensors for detection, but these methods have the problems of low detection efficiency and difficulty in achieving real-time monitoring. Therefore, there is an urgent need for a method and system that can accurately, real-time and automatically identify the degree of switch contact wear. Summary of the invention

[0004] The object of the present invention is to provide a method and system for identifying the degree of wear of switch contacts, so as to solve the above-mentioned problems existing in the prior art.

[0005] In a first aspect, an embodiment of the present invention provides a method for identifying the degree of wear of a switch contact, comprising:

[0006] Acquire a contact burning image and contact resistance at multiple time points; the contact burning image is a plane image of the switch contact;

[0007] Clustering the contact burn image with RGB values ​​to obtain a burn clustering region and a corresponding burn clustering center; the burn clustering center indicates a position with the highest degree of burn in the burn clustering region;

[0008] Based on the contact burn image, the burn cluster area and the corresponding burn cluster center, the location and degree of wear are detected to obtain the burn wear characteristics;

[0009] Inputting the resistances at the multiple time points into a time convolution network to obtain resistance wear features; the resistance wear features represent features of resistance changes caused by wear;

[0010] The burning wear characteristics and the resistance wear characteristics are input into the discrimination network to jointly judge the influence of the burning degree and the resistance change on the wear degree to obtain a wear discrimination value; the wear discrimination value indicates the degree of wear of the switch contact.

[0011] Optionally, the detecting the location and degree of wear based on the contact burn image, the burn cluster area and the corresponding burn cluster center to obtain the burn wear feature includes:

[0012] Acquire a contact plane; the contact plane represents a plane formed by contact surfaces of two switch contacts;

[0013] According to the contact plane, the two-dimensional matrix is ​​associated with the positional relationship of the switch contacts to obtain a distance matrix;

[0014] Based on the distance matrix and the burn clustering regions, a burn adjacency matrix is ​​obtained; the burn adjacency matrix represents a relationship between two connections of multiple positions in the distance matrix;

[0015] Based on the burn adjacency matrix, a plurality of shape position sets are obtained;

[0016] Putting a plurality of shape position sets into a switch structure; the switch structure is a structure that outputs stored data when turned on and does not output stored data when turned off;

[0017] Through the switch structure, the burn wear feature is obtained based on the contact burn image, multiple shape position sets, burn clustering areas and corresponding burn clustering centers.

[0018] Optionally, the burn wear feature is obtained by using a switch structure based on the contact burn image, a plurality of shape position sets, a burn clustering area and a corresponding burn clustering center, including:

[0019] Obtain a random sequence; the values ​​in the random sequence are positive integers from 0 to n; if all the values ​​in the random sequence are 0, it means that no data is output;

[0020] Extracting a corresponding shape position set in the switch structure according to the random sequence as a second shape position set;

[0021] According to the second shape position set, corresponding positions are connected in sequence, and an image is drawn to obtain a second shape image;

[0022] Based on the second shape image, the contact burn image, the burn clustering area and the corresponding burn clustering center, the burn wear feature is obtained through the burn detection network.

[0023] Optionally, the obtaining of the burn wear feature through a burn detection network based on the second shape image, the contact burn image, the burn cluster area and the corresponding burn cluster center includes:

[0024] In the contact burn image, the value of the burn cluster area is retained, and the values ​​of other positions are set to 0, so as to obtain a first burn cluster image;

[0025] superimposing the second shape image and the first cauterization cluster image to obtain a first superimposed image;

[0026] The burn detection network includes a first burn detection network and a second burn detection network;

[0027] obtaining, by the first burn detection network, a first burn position feature based on the first superimposed image;

[0028] Obtaining a second burn wear feature based on the first superimposed image and the corresponding burn cluster center through the second burn detection network;

[0029] The first burning position feature and the second burning wear feature are input into a fusion network to obtain a burning wear feature.

[0030] Optionally, obtaining a first burn position feature based on the first superimposed image through the first burn detection network includes:

[0031] The first burn detection network includes a first convolutional network and a first deconvolutional network;

[0032] The first convolution network includes a plurality of three-dimensional convolution kernels with a size of 2*2*2; the first deconvolution network includes a plurality of two-dimensional convolution kernels with a size of 2*2;

[0033] The three-dimensional convolution kernel of the first convolution network is convolved on the first superimposed image with a step size of 1 to obtain a first change feature; the first change feature represents a position in the detection burn distance clustering matrix that undergoes an equivalent change due to a change in the RGB value of the burn;

[0034] With a step size of 1, the two-dimensional convolution kernel of the first convolutional network is deconvolved on the first change feature to obtain a first burn position feature.

[0035] Optionally, obtaining the second burn wear feature based on the first superimposed image and the corresponding burn cluster center through the second burn detection network includes:

[0036] The second burn detection network includes a plurality of three-dimensional convolution kernels of size 2*2*2;

[0037] A wide straight line passing through the center of the burn cluster and parallel to the first overlay graph is taken as the first straight line;

[0038] A long straight line passing through the center of the burn cluster and parallel to the first overlay graph is used as the second straight line;

[0039] The three-dimensional convolution kernel corresponding to the second burn detection network is convolved on the first straight line of the first superimposed image with a step size of 1 to obtain a first convolution feature; the first convolution feature represents the relationship between the burn state and the position of the midpoint of the burn cluster in the width direction;

[0040] The three-dimensional convolution kernel corresponding to the second burn detection network is convolved on the second straight line of the first superimposed image with a step size of 1 to obtain a second convolution feature; the second convolution feature represents the relationship between the burn state and the position of the midpoint of the burn cluster in the long direction;

[0041] The first convolution feature is superimposed with the second convolution feature to obtain the second burn and wear feature.

[0042] Optionally, the step of obtaining a plurality of shape position sets based on the burn adjacency matrix includes:

[0043] Obtaining a first row sequence; the first row sequence is the sequence number of a random row of the burn adjacency matrix;

[0044] Find the sequence number of the column whose first row sequence is 1 in the burn adjacency matrix to obtain the first column sequence;

[0045] The position corresponding to the first row sequence and the first column sequence is taken as the first position;

[0046] Put the first position into the shape position collection;

[0047] Find the number of a row with the same first column sequence as the second row sequence;

[0048] Find the sequence number of the column whose second row sequence is 1 in the burning adjacency matrix to obtain the second column sequence;

[0049] The position corresponding to the second row sequence and the second column sequence is taken as the second position;

[0050] Put the second position into the shape position collection;

[0051] The serial numbers of connected rows and columns in the burn adjacency matrix are repeatedly detected until the corresponding position is equal to the first position;

[0052] All rows and columns in the burn adjacency matrix are traversed to obtain multiple shape position sets.

[0053] Optionally, the step of associating the two-dimensional matrix with the positional relationship of the switch contacts according to the contact plane to obtain a distance matrix includes:

[0054] Constructing a two-dimensional matrix; the initial values ​​of the two-dimensional matrix are all 0;

[0055] Associating the position of the switch contact projected vertically in the contact plane with the position in the two-dimensional matrix to obtain a distance matrix;

[0056] The vertical distance between the switch contact and the contact plane is taken as the value of the corresponding position in the distance matrix.

[0057] Optionally, obtaining a cauterization adjacency matrix based on the distance matrix and the cauterization clustering regions includes:

[0058] In the distance matrix, the values ​​of the burn clustering area are retained, and the values ​​of other positions are set to 0, thereby obtaining a burn distance clustering matrix;

[0059] The positions corresponding to the same distance values ​​in the burn distance clustering matrix are connected, and an image is drawn to obtain a burn adjacency matrix.

[0060] In a second aspect, an embodiment of the present invention provides a system for identifying the degree of wear of switch contacts, including:

[0061] An acquisition module, used to acquire a contact burning image and contact resistance at multiple time points; the contact burning image is a plane image of the switch contact;

[0062] A clustering module, used for clustering the contact burn image by RGB value to obtain a burn cluster area and a corresponding burn cluster center; the burn cluster center indicates a position with the highest degree of burn in the burn cluster area;

[0063] A burn wear detection module, used to detect the location and degree of wear based on the contact burn image, the burn cluster area and the corresponding burn cluster center, and obtain the burn wear characteristics;

[0064] A resistance wear detection module, used for inputting the resistances at the plurality of time points into a time convolution network to obtain resistance wear features; the resistance wear features represent features of resistance changes caused by wear;

[0065] The wear discrimination module is used to input the burning wear characteristics and the resistance wear characteristics into the discrimination network, jointly judge the influence of the burning degree and the resistance change on the wear degree, and obtain the wear discrimination value; the wear discrimination value indicates the degree of wear of the switch contact.

[0066] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0067] The embodiment of the present invention also provides a method and system for identifying the degree of wear of switch contacts.

[0068] In the present invention, because the burnt area will cause the wear of the switch contacts, and the wear of the contacts will cause the burnt contacts of the switch. The RGB values ​​on the image are detected, and the burnt area on the surface of the switch contacts is found through clustering. And because the contact position of the switch contacts will also affect the burnt area, a superimposed image of the distance matrix and the contact burnt image is constructed, and the burnt situation is used together to determine the degree of wear. And the distance matrix can also analyze surface scratches and wear marks to determine the wear situation. And because the wear of the switch contacts will affect the resistance, the degree of wear can be determined by the change in resistance. A more accurate technical effect of determining the degree of wear of the switch contacts by detecting the change in resistance and the burnt situation is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 The present invention provides a flow chart of a method for identifying the degree of wear of switch contacts.

[0070] Figure 2 It is a schematic diagram of obtaining a distance matrix in a method for identifying the degree of wear of a switch contact provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0072] Example 1

[0073] like Figure 1 As shown, an embodiment of the present invention provides a method for identifying the degree of wear of a switch contact, the method comprising:

[0074] S101: Acquire a contact burning image and contact resistance at multiple time points; the contact burning image is a planar image of a switch contact.

[0075] When the switch contact is in a switching operation, the surface of the switch contact will be burned. The contact burning image is an image of the surface of the switch contact detected by the optical sensor.

[0076] In this embodiment, the switch contact is a copper-chromium alloy self-supporting contact.

[0077] The contact resistance is the resistance of the switch contact when the switch is turned on.

[0078] S102: clustering the contact burn image according to RGB values ​​to obtain a burn cluster region and a corresponding burn cluster center; the burn cluster center indicates a position with the highest degree of burn in the burn cluster region.

[0079] The burn clustering region indicates finding one or more burn regions according to the burn locations.

[0080] The R value, the G value and the B value in the RGB value are clustered respectively, and the intersection of the three cluster sets is obtained to obtain the burn cluster area and the burn cluster center corresponding to the burn cluster area.

[0081] The contact burning image uses RGB values ​​to represent the degree of burning.

[0082] In this embodiment, k-means is used for clustering.

[0083] The burn clustering region is represented by a plurality of points on the contact burn image.

[0084] S103: Based on the contact burn image, the burn cluster area and the corresponding burn cluster center, the position and degree of wear are detected to obtain the burn wear characteristics.

[0085] The burning wear characteristics represent the characteristics of the wear caused by burning and the common wear reflected in the distance from the contact plane.

[0086] S104: Inputting the resistances at the multiple time points into a time convolution network to obtain resistance wear features; the resistance wear features represent features of resistance changes caused by wear.

[0087] Wherein, the temporal convolutional network (TCN).

[0088] S105: Inputting the burning wear characteristics and the resistance wear characteristics into the discrimination network, jointly judging the influence of the burning degree and the resistance change on the wear degree, and obtaining the wear discrimination value.

[0089] In this embodiment, the discrimination network is a fully connected neural network (Fully Connected Neural Network, FCNN), and the labeled historical wear degree of switch contacts is used to train the discrimination network.

[0090] Optionally, the detecting the location and degree of wear based on the contact burn image, the burn cluster area and the corresponding burn cluster center to obtain the burn wear feature includes:

[0091] A contact plane is obtained; the contact plane represents a plane formed by contact surfaces of two switch contacts.

[0092] Wherein, the contact plane is as follows Figure 2 shown.

[0093] According to the contact plane, the two-dimensional matrix is ​​associated with the positional relationship of the switch contacts to obtain a distance matrix.

[0094] The positions in the distance matrix are used to represent the positions on the contact plane.

[0095] The position of the switch contact perpendicular to the contact plane represents the position of the switch contact in the distance matrix. The distance of the switch contact perpendicular to the contact plane is the value of the corresponding position in the distance matrix.

[0096] Based on the distance matrix and the burn clustering regions, a burn adjacency matrix is ​​obtained; the burn adjacency matrix represents a relationship between two connections of multiple positions in the distance matrix;

[0097] Based on the burn adjacency matrix, a plurality of shape position sets are obtained;

[0098] A plurality of shape position sets are placed into a switch structure; the switch structure is a structure that outputs stored data when turned on and does not output stored data when turned off.

[0099] Through the switch structure, the burn wear feature is obtained based on the contact burn image, multiple shape position sets, burn clustering areas and corresponding burn clustering centers.

[0100] Optionally, the first cauterization matrix is ​​obtained based on the contact cauterization image and a plurality of shape position sets through the switch structure, including:

[0101] A random sequence is obtained; the values ​​in the random sequence are positive integers from 0 to n; and all the values ​​in the random sequence are 0, indicating that no data is output.

[0102] Wherein, n represents the total number of elements in the plurality of shape position sets.

[0103] A corresponding shape position set in the switch structure is extracted according to the random sequence as a second shape position set.

[0104] According to the second shape position set, corresponding positions are connected in sequence, and an image is drawn to obtain a second shape image.

[0105] Among them, the subscripts of the corresponding positions in the second shape position set are connected in sequence.

[0106] The second shape image is a binary image, and the connection position is indicated by black.

[0107] Among them, a second shape position set corresponds to a second shape image.

[0108] Based on the second shape image, the contact burn image, the burn clustering area and the corresponding burn clustering center, the burn wear feature is obtained through the burn detection network.

[0109] Optionally, the obtaining of the burn wear feature through a burn detection network based on the second shape image, the contact burn image, the burn cluster area and the corresponding burn cluster center includes:

[0110] In the contact burn image, the value of the burn cluster area is retained, and the values ​​of other positions are set to 0, so as to obtain a first burn cluster image;

[0111] The second shape image and the first burn cluster image are superimposed to obtain a first superimposed image.

[0112] The width of the first superimposed image is equal to the width of the first cautery distance image, which is equal to the width of the first cautery cluster image. The length of the first superimposed image is equal to the width of the first cautery distance image, which is equal to the width of the first cautery cluster image. The number of channels of the first superimposed image is equal to the number of channels of the first cautery distance image plus the first cautery cluster image.

[0113] The burn detection network includes a first burn detection network and a second burn detection network;

[0114] obtaining, by the first burn detection network, a first burn position feature based on the first superimposed image;

[0115] Obtaining a second burn wear feature based on the first superimposed image and the corresponding burn cluster center through the second burn detection network;

[0116] The first burning position feature and the second burning wear feature are input into a fusion network to obtain a burning wear feature.

[0117] Among them, in this embodiment, the fusion network is a fully connected neural network (Fully Connected Neural Network, FCNN).

[0118] Optionally, obtaining a first burn position feature based on the first superimposed image through the first burn detection network includes:

[0119] The first burn detection network includes a first convolutional network and a first deconvolutional network.

[0120] The first convolutional network is a convolutional neural network (CNN). The first deconvolutional network is a convolutional neural network (CNN) that performs a deconvolution operation.

[0121] The first convolution network includes a plurality of three-dimensional convolution kernels with a size of 2*2*2; the first deconvolution network includes a plurality of two-dimensional convolution kernels with a size of 2*2;

[0122] The three-dimensional convolution kernel of the first convolution network is convolved on the first superimposed image with a step size of 1 to obtain a first change feature; the first change feature represents a position in the detection burn distance clustering matrix that undergoes an equivalent change due to a change in the RGB value of the burn;

[0123] With a step size of 1, the two-dimensional convolution kernel of the first convolutional network is deconvolved on the first change feature to obtain a first burn position feature.

[0124] By using the above method and adding features by deconvolution, the first burning position feature can contain more feature information.

[0125] Optionally, obtaining the second burn wear feature based on the first superimposed image and the corresponding burn cluster center through the second burn detection network includes:

[0126] The second burn detection network includes a plurality of three-dimensional convolution kernels of size 2*2*2;

[0127] A wide straight line passing through the center of the burn cluster and parallel to the first overlay graph is taken as the first straight line;

[0128] A long straight line passing through the center of the burn cluster and parallel to the first overlay graph is used as the second straight line;

[0129] With a step size of 1, the three-dimensional convolution kernel corresponding to the second burn detection network is convolved on the first straight line of the first superimposed image to obtain a first convolution feature; the first convolution feature represents the relationship between the burn state and the position of the midpoint of the burn cluster in the width direction.

[0130] With a step size of 1, the three-dimensional convolution kernel corresponding to the second burn detection network is convolved on the second straight line of the first superimposed image to obtain a second convolution feature; the second convolution feature represents the relationship between the burn state and the position of the midpoint of the burn cluster in the long direction.

[0131] The first convolution feature is superimposed with the second convolution feature to obtain the second burn and wear feature.

[0132] Among them, the first convolution feature, the second convolution feature and the second burn and wear feature are all one-dimensional vectors, and the number of elements of the vector corresponding to the second burn and wear feature is equal to the number of elements of the vector corresponding to the first convolution feature plus the number of elements of the vector corresponding to the second convolution feature.

[0133] Optionally, the step of obtaining a plurality of shape position sets based on the burn adjacency matrix includes:

[0134] Obtaining a first row sequence; the first row sequence is the sequence number of a random row of the burn adjacency matrix;

[0135] Find the sequence number of the column whose first row sequence is 1 in the burn adjacency matrix to obtain the first column sequence;

[0136] The position corresponding to the first row sequence and the first column sequence is taken as the first position;

[0137] Put the first position into the shape position collection;

[0138] Find the number of a row with the same first column sequence as the second row sequence;

[0139] Find the sequence number of the column whose second row sequence is 1 in the burning adjacency matrix to obtain the second column sequence;

[0140] The position corresponding to the second row sequence and the second column sequence is taken as the second position;

[0141] Put the second position into the shape position collection;

[0142] The serial numbers of connected rows and columns in the burn adjacency matrix are repeatedly detected until the corresponding position is equal to the first position;

[0143] All rows and columns in the burn adjacency matrix are traversed to obtain multiple shape position sets.

[0144] Optionally, the step of associating the two-dimensional matrix with the positional relationship of the switch contacts according to the contact plane to obtain a distance matrix includes:

[0145] Constructing a two-dimensional matrix; the initial values ​​of the two-dimensional matrix are all 0;

[0146] Associating the position of the switch contact projected vertically in the contact plane with the position in the two-dimensional matrix to obtain a distance matrix;

[0147] The vertical distance between the switch contact and the contact plane is taken as the value of the corresponding position in the distance matrix.

[0148] Optionally, obtaining a cauterization adjacency matrix based on the distance matrix and the cauterization clustering regions includes:

[0149] In the distance matrix, the values ​​of the burn clustering area are retained, and the values ​​of other positions are set to 0, thereby obtaining a burn distance clustering matrix.

[0150] Wherein, a plurality of burn clustering regions correspond to a plurality of burn distance clustering matrices, and one burn clustering region corresponds to one burn distance clustering matrix.

[0151] The positions corresponding to the same distance values ​​in the burn distance clustering matrix are connected, and an image is drawn to obtain a burn adjacency matrix.

[0152] The rows and columns of the burn adjacency matrix represent multiple non-zero positions in the burn distance clustering matrix. The number of rows of the burn adjacency matrix is ​​equal to the number of columns, which is equal to n, where n is a positive integer. The number of rows is the number of non-zero positions in the burn distance clustering matrix. The i-th row and the i-th column represent the same position, where i is a positive integer less than or equal to n.

[0153] Example 2

[0154] Based on the above-mentioned method for identifying the degree of wear of switch contacts, an embodiment of the present invention also provides a system for identifying the degree of wear of switch contacts, the system comprising an acquisition module, a clustering module, a burning wear detection module, a resistance wear detection module and a wear judgment module.

[0155] An acquisition module, used to acquire a contact burning image and contact resistance at multiple time points; the contact burning image is a plane image of the switch contact;

[0156] A clustering module, used for clustering the contact burn image by RGB value to obtain a burn cluster area and a corresponding burn cluster center; the burn cluster center indicates a position with the highest degree of burn in the burn cluster area;

[0157] A burn wear detection module, used to detect the location and degree of wear based on the contact burn image, the burn cluster area and the corresponding burn cluster center, and obtain the burn wear characteristics;

[0158] A resistance wear detection module, used for inputting the resistances at the plurality of time points into a time convolution network to obtain resistance wear features; the resistance wear features represent features of resistance changes caused by wear;

[0159] The wear discrimination module is used to input the burning wear characteristics and the resistance wear characteristics into the discrimination network, jointly judge the influence of the burning degree and the resistance change on the wear degree, and obtain the wear discrimination value; the wear discrimination value indicates the degree of wear of the switch contact.

[0160] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those explicitly recited in each claim. More specifically, as reflected in the claims below, inventive aspects lie in less than all the features of the individual embodiments disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present invention.

Claims

1. A method for identifying the degree of wear of a switch contact, characterized in that: include: Acquire contact burning images and contact resistance at multiple time points; The contact burning image is a plane image of the switch contact; Clustering the contact burn image according to the RGB value to obtain the burn cluster area and the corresponding burn cluster center; The burn cluster center indicates the location with the highest degree of burn in the burn cluster area; Based on the contact burn image, the burn cluster area and the corresponding burn cluster center, the location and degree of wear are detected to obtain the burn wear characteristics; Inputting the resistances at the multiple time points into a time convolution network to obtain resistance wear characteristics; The resistance wear characteristic represents the characteristic of the change of resistance caused by wear; The burning wear characteristics and the resistance wear characteristics are input into the discrimination network to jointly judge the influence of the burning degree and the resistance change on the wear degree to obtain a wear discrimination value; the wear discrimination value indicates the degree of wear of the switch contact.

2. A method for identifying the degree of wear of switch contacts according to claim 1, characterized in that: The method of detecting the location and degree of wear based on the contact burn image, the burn cluster area and the corresponding burn cluster center to obtain the burn wear characteristics includes: Acquire a contact plane; the contact plane represents a plane formed by contact surfaces of two switch contacts; According to the contact plane, the two-dimensional matrix is ​​associated with the positional relationship of the switch contacts to obtain a distance matrix; Based on the distance matrix and the burn clustering regions, a burn adjacency matrix is ​​obtained; the burn adjacency matrix represents a relationship between two connections of multiple positions in the distance matrix; Based on the burn adjacency matrix, a plurality of shape position sets are obtained; Putting a plurality of shape position sets into a switch structure; the switch structure is a structure that outputs stored data when turned on and does not output stored data when turned off; Through the switch structure, the burn wear feature is obtained based on the contact burn image, multiple shape position sets, burn clustering areas and corresponding burn clustering centers.

3. A method for identifying the degree of wear of switch contacts according to claim 2, characterized in that: The switch structure is used to obtain burn wear characteristics based on the contact burn image, multiple shape position sets, burn cluster areas and corresponding burn cluster centers, including: Obtain a random sequence; the values ​​in the random sequence are positive integers from 0 to n; if all the values ​​in the random sequence are 0, it means that no data is output; Extracting a corresponding shape position set in the switch structure according to the random sequence as a second shape position set; According to the second shape position set, corresponding positions are connected in sequence, and an image is drawn to obtain a second shape image; Based on the second shape image, the contact burn image, the burn clustering area and the corresponding burn clustering center, the burn wear feature is obtained through the burn detection network.

4. A method for identifying the degree of wear of switch contacts according to claim 3, characterized in that: The method of obtaining burn wear characteristics based on the second shape image, the contact burn image, the burn cluster area and the corresponding burn cluster center through a burn detection network includes: In the contact burn image, the value of the burn cluster area is retained, and the values ​​of other positions are set to 0, so as to obtain a first burn cluster image; superimposing the second shape image and the first cauterization cluster image to obtain a first superimposed image; The burn detection network includes a first burn detection network and a second burn detection network; obtaining, by the first burn detection network, a first burn position feature based on the first superimposed image; Obtaining a second burn wear feature based on the first superimposed image and the corresponding burn cluster center through the second burn detection network; The first burning position feature and the second burning wear feature are input into a fusion network to obtain a burning wear feature.

5. A method for identifying the degree of wear of switch contacts according to claim 4, characterized in that: The obtaining of a first burn position feature based on the first superimposed image by the first burn detection network includes: The first burn detection network includes a first convolutional network and a first deconvolutional network; The first convolution network includes a plurality of three-dimensional convolution kernels with a size of 2*2*2; the first deconvolution network includes a plurality of two-dimensional convolution kernels with a size of 2*2; The three-dimensional convolution kernel of the first convolution network is convolved on the first superimposed image with a step size of 1 to obtain a first change feature; the first change feature represents a position in the detection burn distance clustering matrix that undergoes an equivalent change due to a change in the RGB value of the burn; With a step size of 1, the two-dimensional convolution kernel of the first convolutional network is deconvolved on the first change feature to obtain a first burn position feature.

6. A method for identifying the degree of wear of switch contacts according to claim 4, characterized in that: The second burn detection network is used to obtain a second burn wear feature based on the first superimposed image and the corresponding burn cluster center, including: The second burn detection network includes a plurality of three-dimensional convolution kernels of size 2*2*2; A wide straight line passing through the center of the burn cluster and parallel to the first overlay graph is taken as the first straight line; A long straight line passing through the center of the burn cluster and parallel to the first overlay graph is used as the second straight line; The three-dimensional convolution kernel corresponding to the second burn detection network is convolved on the first straight line of the first superimposed image with a step size of 1 to obtain a first convolution feature; the first convolution feature represents the relationship between the burn state and the position of the midpoint of the burn cluster in the width direction; The three-dimensional convolution kernel corresponding to the second burn detection network is convolved on the second straight line of the first superimposed image with a step size of 1 to obtain a second convolution feature; the second convolution feature represents the relationship between the burn state and the position of the midpoint of the burn cluster in the long direction; The first convolution feature is superimposed with the second convolution feature to obtain the second burn and wear feature.

7. A method for identifying the degree of wear of switch contacts according to claim 2, characterized in that: Based on the burn adjacency matrix, a plurality of shape position sets are obtained, including: Obtaining a first row sequence; the first row sequence is the sequence number of a random row of the burn adjacency matrix; Find the sequence number of the column whose first row sequence is 1 in the burn adjacency matrix to obtain the first column sequence; The position corresponding to the first row sequence and the first column sequence is taken as the first position; Put the first position into the shape position collection; Find the number of a row with the same first column sequence as the second row sequence; Find the sequence number of the column whose second row sequence is 1 in the burning adjacency matrix to obtain the second column sequence; The position corresponding to the second row sequence and the second column sequence is taken as the second position; Put the second position into the shape position collection; The serial numbers of connected rows and columns in the burn adjacency matrix are repeatedly detected until the corresponding position is equal to the first position; All rows and columns in the burn adjacency matrix are traversed to obtain multiple shape position sets.

8. A method for identifying the degree of wear of switch contacts according to claim 2, characterized in that: The step of associating the two-dimensional matrix with the positional relationship of the switch contacts according to the contact plane to obtain a distance matrix includes: Constructing a two-dimensional matrix; the initial values ​​of the two-dimensional matrix are all 0; Associating the position of the switch contact projected vertically in the contact plane with the position in the two-dimensional matrix to obtain a distance matrix; The vertical distance between the switch contact and the contact plane is taken as the value of the corresponding position in the distance matrix.

9. A method for identifying the degree of wear of switch contacts according to claim 2, characterized in that: The step of obtaining a burn adjacency matrix based on the distance matrix and the burn clustering regions includes: In the distance matrix, the values ​​of the burn clustering area are retained, and the values ​​of other positions are set to 0, thereby obtaining a burn distance clustering matrix; The positions corresponding to the same distance values ​​in the burn distance clustering matrix are connected, and an image is drawn to obtain a burn adjacency matrix.

10. A system for identifying the degree of wear of switch contacts, characterized in that: include: An acquisition module, used for acquiring contact burning images and contact resistances at multiple time points; The contact burning image is a plane image of the switch contact; A clustering module, used for clustering the contact burn image according to RGB values ​​to obtain burn clustering areas and corresponding burn clustering centers; The burn cluster center indicates the location with the highest degree of burn in the burn cluster area; A burn wear detection module, used to detect the location and degree of wear based on the contact burn image, the burn cluster area and the corresponding burn cluster center, and obtain the burn wear characteristics; A resistor wear detection module, used for inputting the resistances at the multiple time points into a time convolution network to obtain resistor wear characteristics; The resistance wear characteristic represents the characteristic of the change of resistance caused by wear; The wear discrimination module is used to input the burning wear characteristics and the resistance wear characteristics into the discrimination network, jointly judge the influence of the burning degree and the resistance change on the wear degree, and obtain the wear discrimination value; the wear discrimination value indicates the degree of wear of the switch contact.

Images