Transformer Automatic Wiring Device

Abstract

The present invention belongs to the technical field of transformer manufacturing, and specifically discloses an automatic winding device for a transformer, which includes a column and a driving box. A sliding kit is sleeved on the column. A longitudinal movement component is arranged on one side of the sliding kit, and a wire reel driving component is arranged on the other side of the sliding kit. The position of the wire reel is adapted to the position of the transformer mold; it also includes a control system, which regulates the rotational speed parameter of the sensing layer through a BP neural network model based on the winding working condition so that the actual adjacent wire spacing obtained by the application layer is equal to the adjacent wire spacing parameter. The present invention adopts a vertical winding layout, sets the transformer mold on the top of the driving box, and arranges a longitudinal movement component and a wire reel driving component on the column. The equipment occupies a small area, and the control system judges the winding working condition to select appropriate rotational speeds for the transformer mold and the wire reel, ensuring the adjacent wire spacing while guaranteeing the wire tension, and meeting the requirements of different winding working conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of transformer manufacturing, and particularly relates to an automatic winding device for transformers. Background Art

[0002] With the continuous advancement of power infrastructure construction, the demand for transformers has been continuously increasing, thus driving the development of the transformer winding market. Traditional manual winding methods have disadvantages such as inaccurate number of turns and uneven winding density, and are gradually being replaced by automated winding equipment. Currently, large production workshops all adopt the method of manual assistance with horizontal winding machines for transformer winding.

[0003] For example, Chinese Patent with publication number CN109103015A discloses an automatic winding machine for power transformer coils with bidirectional winding and its winding method. Through PLC programming, automatic winding of the conductor and insulating paper tape is achieved, and a wire table rotating mechanism is used to rotate it by 180 degrees to wind the conductor in the reverse direction on the wire roller, improving production efficiency. However, in this patent, due to the use of traditional horizontal winding, the entire equipment occupies a large area. When winding the transformer, depending on the application scenario, in addition to tightly winding the conductor, loose winding is also required. For example, in a radio frequency transformer, loose winding can reduce the influence of distributed capacitance on signal transmission and improve heat dissipation performance. In this solution, it is impossible to select an appropriate winding speed according to different winding conditions, and it is difficult to ensure the spacing between adjacent conductors when sparse winding is required. Summary of the Invention

[0004] The purpose of the present invention is to overcome the defects existing in the prior art and provide an automatic winding device for transformers.

[0005] The present invention provides an automatic winding device for transformers, including:

[0006] A column and a drive box arranged on a bottom plate. A rack is longitudinally arranged on the outer side of the column, and a sliding sleeve is sleeved on the column. A longitudinal movement component is arranged on one side of the sliding sleeve for driving the sliding sleeve to reciprocate longitudinally along the column. A wire reel drive component is arranged on the other side of the sliding sleeve for driving the wire reel to rotate. The drive box is used for driving the transformer mold to rotate. The position of the wire reel is adapted to the position of the transformer mold, so that when the wire reel and the transformer mold rotate simultaneously, winding of the transformer mold is achieved;

[0007] A control system, including an input layer, a perception layer, and an application layer. The input layer is used for inputting the specification parameters of the transformer mold and the adjacent conductor spacing parameters, and judging the winding conditions of the transformer mold based on the specification parameters and the adjacent conductor spacing parameters;

[0008] The sensing layer is used to obtain the rotational speed parameters of the longitudinal movement component, the wire reel driving component, and the driving box;

[0009] The application layer is used to obtain the true adjacent wire spacing of the transformer mold;

[0010] Based on the winding working conditions, the rotational speed parameters of the sensing layer are regulated through a BP neural network model so that the true adjacent wire spacing obtained by the application layer is equal to the adjacent wire spacing parameter.

[0011] A further solution is that the longitudinal movement component includes two fixed plates arranged in parallel. A gear is arranged between the two fixed plates. One side of the fixed plate is fixedly connected to the sliding sleeve. The gear is rotatably connected to the fixed plate and meshes with the rack; A fixed frame is arranged on the top of the fixed plate. A first motor is arranged on the fixed frame. The output end of the first motor is connected to a driving shaft. A spiral tooth is arranged on the driving shaft. The spiral tooth meshes with the gear;

[0012] The wire reel driving component includes a mounting frame. The mounting frame is fixedly connected to the other side of the sliding sleeve. A second motor and a speed reducer are arranged on the mounting frame. The second motor is connected to the speed reducer. A first rotating shaft for installing the wire reel is arranged at the output end of the speed reducer;

[0013] Inside the driving box, a third motor and a support column are arranged. The top of the support column is rotatably connected to a first pulley. The output end of the third motor is provided with a second pulley. The first pulley and the second pulley are connected by a belt; A second rotating shaft for installing the transformer mold is arranged on the top of the first pulley.

[0014] A further solution is that the input layer includes:

[0015] An input unit for inputting the specification parameters and adjacent wire spacing parameters of the transformer mold;

[0016] A judgment unit, connected to the input unit, for judging and outputting the winding working conditions of the transformer mold;

[0017] The judgment process of the judgment unit is as follows:

[0018] Based on the specification parameters of the transformer mold, the winding width w, the number of winding turns s, and the wire diameter k are extracted. Then the adjacent wire spacing parameter is expressed as , if The result of is in the first interval, and the winding working condition output by the judgment unit is tight winding; if The result of is in the second interval, and the winding working condition output by the judgment unit is sparse winding; The maximum value of the first interval is less than or equal to the minimum value of the second interval.

[0019] A further solution is that the sensing layer includes: a first rotational speed sensor for real-time monitoring of the rotational speed of the first motor; a second rotational speed sensor for real-time monitoring of the rotational speed of the first rotating shaft; and a third rotational speed sensor for real-time monitoring of the rotational speed of the second rotating shaft.

[0020] A further solution is that the application layer is an image recognition module for real-time monitoring of the wire spacing between adjacent wires of the transformer mold;

[0021] The image recognition module includes an image acquisition unit, an image processing unit, an edge detection unit, and a spacing calculation unit;

[0022] The image acquisition unit is used to acquire the image information of the transformer mold at fixed time intervals;

[0023] The image processing unit is used to perform grayscale processing and filtering processing on the image information;

[0024] The edge detection unit extracts the edges of two adjacent windings based on the Canny edge detection operator, and performs curve fitting on the detected edge points of the two adjacent windings, constructs a two-dimensional coordinate system based on the graphic information, and obtains the coordinates x i ,y i， The curve obtained by fitting is:

[0025] y = a0 + a1x + a2x 2 +…+a n x n ;

[0026] where a0, a1, a2…a n are coefficients determined by minimizing the sum of squared errors ; in the formula, is the ordinate of the corresponding point x i on the fitting curve, and m is the number of points;

[0027] The spacing calculation unit is used to calculate the spacing between two adjacent windings ; the calculation process is: perform equidistant sampling on the two obtained adjacent fitting curves to obtain the corresponding points and ;

[0028] Based on the corresponding points and , ;

[0029] where (x1, y1) and (x2, y2) are the coordinates of the points and respectively.

[0030] Start the first motor, the second motor, and the third motor simultaneously, record the rotational speed parameters of the first motor, the first rotating shaft, and the second rotating shaft at fixed time intervals, and at the same time record the adjacent wire spacing of the transformer mold to generate a data set;

[0031] The input layer of the BP neural network model is an output neuron for inputting the adjacent wire spacing; the output layer of the BP neural network model includes three output neurons, which are respectively used to output the output values of the first rotational speed sensor, the second rotational speed sensor, and the third rotational speed sensor; the hidden layer of the BP neural network model includes two neurons, the activation function is the Sigmoid function, and the two neurons are respectively used to normalize the rotational speed parameters and the adjacent wire spacing parameters of the data set;

[0032] Divide the processed data set into a training set and a test set. In the training set, form an output sample matrix X with the output values of the normalized first rotational speed sensor, second rotational speed sensor, and third rotational speed sensor, with a shape of (k, 3), and form an input sample matrix Y with the normalized adjacent wire spacing, with a shape of (k, 1), where k is the number of training set samples;

[0033] Input the input sample matrix Y and the output sample matrix X into the BP neural network model for iterative training, and respectively output the predicted values of the first rotational speed sensor, the second rotational speed sensor, and the third rotational speed sensor.

[0034] A further solution is that the data set includes a first data set and a second data set. The first data set includes several groups of output values of the first rotational speed sensor, the second rotational speed sensor, and the third rotational speed sensor corresponding to different adjacent wire spacings under the tight winding condition;

[0035] The second data set includes several groups of output values of the first rotational speed sensor, the second rotational speed sensor, and the third rotational speed sensor corresponding to different adjacent wire spacings under the sparse winding condition.

[0036] A further solution is that the normalization process of the rotational speed parameters includes mapping the output values of the first rotational speed sensor, the second rotational speed sensor, and the third rotational speed sensor to the same interval [0, 1]. The rotational speed parameter normalization formula is:

[0037] normalized_s=(s - min_s) / (max_s - min_s);

[0038] where normalized_s is the normalized rotational speed, s is the output rotational speed of the first rotational speed sensor, the second rotational speed sensor, and the third rotational speed sensor, max_s is the maximum rotational speed, and min_s is the minimum rotational speed;

[0039] The normalization formula for the adjacent wire spacing parameter is as follows:

[0040] normalized_gap=(gap - min_gap) / (max_gap - min_gap);

[0041] Among them, normalized_gap is the normalized adjacent wire spacing, gap is the adjacent wire spacing output by the image recognition module, max_gap is the maximum adjacent wire spacing, and min_gap is the minimum adjacent wire spacing.

[0042] A further solution is that the control system further includes a motor drive module, which is respectively connected to the first motor, the second motor, and the third motor and is used to drive the first motor, the second motor, and the third motor. The BP neural network model is connected to the motor drive module through a signal conversion module. The signal conversion module is used to convert the output signal of the BP neural network model into a pulse signal, and the motor drive module controls the rotation speed, forward and reverse rotation of the first motor, the second motor, and the third motor based on the pulse signal.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] The present invention adopts a vertical winding layout, sets the transformer mold on the top of the drive box, and sets a longitudinal movement component and a wire reel drive component on the column. The equipment occupies a small area. By judging the winding conditions through the control system, the appropriate rotation speed of the transformer mold and the wire reel speed are selected, ensuring the adjacent wire spacing while guaranteeing the wire tension, improving the winding accuracy, and meeting the requirements of different winding conditions.

[0045] The longitudinal movement component of the present invention can automatically adjust the height of the wire reel in real time, effectively avoiding the overlap of wires during winding. The rotation speed of the transformer can be automatically adjusted through the drive box, facilitating the accurate control of the number of winding turns.

[0046] The present invention accurately obtains the rotation speeds of the first motor, the first rotating shaft, and the second rotating shaft through the input layer, accurately calculates the adjacent wire spacing through the image recognition module, and forms a data set to perform iterative training on the BP neural network model, obtaining a network model that outputs the rotation speeds of the first motor, the first rotating shaft, and the second rotating shaft based on the preset adjacent wire spacing. During wire winding, the rotation speeds of the first motor, the first rotating shaft, and the second rotating shaft are regulated through the BP neural network model based on the winding conditions, so that the actual adjacent wire spacing obtained by the image recognition module is equal to the preset adjacent wire spacing, improving the product accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The following drawings are only for illustrative description and explanation of the present invention and are not used to limit the scope of the present invention, where:

[0048] Figure 1 : Schematic structural diagram of the present invention;

[0049] Figure 2 : Block diagram of the control system connection of the present invention;

[0050] Figure 3 : Schematic winding diagram of the transformer mold;

[0051] In the figure: 1, bottom plate; 2, column; 3, rack; 4, sliding kit; 5, fixing plate; 6, gear; 7, drive shaft; 8, first motor; 9, mounting bracket; 10, second motor; 11, reducer; 12, first rotating shaft; 13, wire reel; 14, drive box; 15, support column; 16, third motor; 17, first pulley; 18, second pulley; 19, second rotating shaft; 20, lower fixing plate; 21, upper fixing plate; 22, transformer mold; 23, input unit; 24, judgment unit; 25, first rotational speed sensor; 26, second rotational speed sensor; 27, third rotational speed sensor; 28, image recognition module; 29, image acquisition unit; 30, image processing unit; 31, edge detection unit; 32, spacing calculation unit; 33, data set; 34, first data set; 35, second data set; 36, BP neural network model; 37, signal conversion module; 38, motor drive module. Detailed implementation manners

[0052] In order to make the purpose, technical solutions, design methods and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below through specific embodiments with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0053] Such as Figure 1 And Figure 2As shown in the figure, the present invention provides an automatic transformer winding device, including a winding structure and a control system for driving the winding structure to work according to predetermined parameters. Among them, the winding structure includes a column 2 and a driving box 14 arranged on a bottom plate 1. A rack 3 is longitudinally arranged on the outer side of the column 2, and a sliding kit 4 is sleeved on the column 2. A longitudinal movement component is arranged on one side of the sliding kit 4 for driving the sliding kit 4 to reciprocate longitudinally along the column 2. A wire reel driving component is arranged on the other side of the sliding kit 4 for driving a wire reel 13 to rotate. The driving box 14 is used for driving a transformer mold 22 to rotate. The position of the wire reel 13 is adapted to the position of the transformer mold 22 so that when the wire reel 13 and the transformer mold 22 rotate simultaneously, winding of the transformer mold 22 is realized. The control system includes an input layer, a perception layer, and an application layer. The input layer is used for inputting the specification parameters of the transformer mold 22 and the adjacent wire spacing parameters, and judging the winding working condition of the transformer mold 22 based on the specification parameters and the adjacent wire spacing parameters. The perception layer is used for obtaining the rotational speed parameters of the longitudinal movement component, the wire reel driving component, and the driving box 14. The application layer is used for obtaining the actual adjacent wire spacing of the transformer mold 22. Based on the winding working condition, the rotational speed parameters of the perception layer are regulated through a BP neural network model 36 so that the actual adjacent wire spacing obtained by the application layer is equal to the adjacent wire spacing parameters.

[0054] Specifically, the longitudinal movement component includes two fixed plates 5 arranged in parallel. A gear 6 is disposed between the two fixed plates 5. One side of the fixed plate 5 is fixedly connected to the sliding sleeve 4. The gear 6 is rotatably connected to the fixed plate 5 and meshes with the rack 3. A fixed frame is provided on the top of the fixed plate 5. A first motor 8 is provided on the fixed frame. The output end of the first motor 8 is connected to a drive shaft 7. A helical tooth is provided on the drive shaft 7 and meshes with the gear 6. When the first motor 8 drives the drive shaft 7 to rotate, the gear 6 is driven to rotate, realizing the longitudinal reciprocating movement of the entire sliding sleeve 4. The wire reel driving component includes a mounting frame 9. The mounting frame 9 is fixedly connected to the other side of the sliding sleeve 4. A second motor 10 and a speed reducer 11 are provided on the mounting frame 9. The second motor 10 is connected to the speed reducer 11. A first rotating shaft 12 is provided at the output end of the speed reducer 11 for mounting the wire reel 13. Since the wire reel driving component is disposed on one side of the sliding sleeve 4, when the sliding sleeve 4 moves longitudinally, the wire reel driving component moves synchronously with the sliding sleeve 4. Inside the drive box 14, a third motor 16 and a support column 15 are provided. The top of the support column 15 is rotatably connected to a first pulley 17. The output end of the third motor 16 is provided with a second pulley 18. The first pulley 17 and the second pulley 18 are connected by a belt. A second rotating shaft 19 is provided on the top of the first pulley 17 for mounting the transformer mold 22. A lower fixing plate 20 is fixedly provided on the second rotating shaft 19 to limit the position of the transformer mold 22. An upper fixing plate 21 is also movably provided on the top of the second rotating shaft 19. After the transformer mold 22 is sleeved on the second rotating shaft 19, the transformer mold 22 is fixed by the upper fixing plate 21, enabling the transformer mold 22 to rotate with the second rotating shaft 19.

[0055] Through the above winding structure, the winding work of the transformer mold 22 is realized. To ensure the winding accuracy and coordinate the rotation speeds of the first motor 8, the second motor 10, and the third motor 16, in this embodiment, the input layer includes an input unit 23 and a judgment unit 24. The input unit 23 is used to input the specification parameters of the transformer mold 22 and the adjacent wire spacing parameters. The judgment unit 24 is connected to the input unit 23 and is used to judge and output the winding working condition of the transformer mold 22. In this embodiment, the judgment process of the judgment unit 24 is as follows: As Figure 3 shown, based on the specification parameters of the transformer mold 22, the winding width w, the number of winding turns s, and the wire diameter k are extracted. Then the adjacent wire spacing parameter is expressed as , if the result is within the first interval, the winding working condition output by the judgment unit is tight winding; if The result is in the second interval, and the winding condition output by the judgment unit is sparse winding; the maximum value of the first interval is less than or equal to the minimum value of the second interval. It should be understood that when performing the tight winding condition, since the winding inductance is proportional to the square of the number of turns of the coil and the magnetic permeability of the magnetic core, tight winding can increase the number of turns in a limited space. In theory, it is required that the distance between adjacent wires approaches 0 infinitely. However, in the actual winding process, it is impossible to ensure that adjacent two wires are completely in contact. Therefore, to ensure the feasibility of the solution, it is limited that as long as the result is within the first interval, it can be considered as the tight winding condition. The selection of the winding condition can be determined according to the allowable error of the inductance of the product in actual construction. For example, the influence value of one turn of wire on the inductance is determined, and then the maximum and minimum values of the number of turns of the transformer mold are determined according to the allowable error of the inductance, and thus the range of the first interval is determined. Among them, the calculation process of the influence value of one turn of wire on the inductance is as follows:

[0056] ;

[0057] where L is the inductance, is the magnetic permeability, s is the number of turns, A is the cross-sectional area of the coil, and l is the length of the coil.

[0058] In the above, the perception layer includes: a first rotational speed sensor 25 for monitoring the rotational speed of the first motor 8 in real time; a second rotational speed sensor 26 for monitoring the rotational speed of the first rotating shaft 12 in real time; a third rotational speed sensor 27 for monitoring the rotational speed of the second rotating shaft 19 in real time. The application layer is an image recognition module 28 for monitoring the distance between adjacent wires of the transformer mold 22 in real time; the image recognition module 28 includes an image acquisition unit 29, an image processing unit 30, an edge detection unit 31, and a distance calculation unit 32; among them, the image acquisition unit 29 is used to acquire the image information of the transformer mold 22 at fixed time intervals; the image processing unit 30 is used to perform grayscale processing and filtering processing on the image information; the edge detection unit 31 extracts the edges of adjacent two windings based on the Canny edge detection operator, and performs curve fitting on the detected edge points of adjacent two windings, constructs a two-dimensional coordinate system based on the graphic information, and obtains the coordinates (x i , y i ) of several edge points, and the fitted curve is:

[0059] y = a0 + a1x + a2x 2 + … + a n x n ;

[0060] where a0, a1, a2 … a n are coefficients, which are determined by minimizing the sum of squared errors ; in the formula, is the corresponding point x on the fitting curve i The ordinate of, and m is the number of points;

[0061] The spacing calculation unit 32 is used to calculate the spacing between adjacent two windings ; For the transformer mold 22, every time a winding is wound, that is, one more turn of wire is added. Therefore, the spacing between adjacent two windings is the spacing between adjacent wires. The calculation process is as follows: perform equidistant sampling on the obtained adjacent two fitting curves to obtain the corresponding points on the two curves and ;

[0062] Based on the corresponding points and , ;

[0063] Among them, (x1, y1) and (x2, y2) are the coordinates of the points and respectively.

[0064] In the above, the construction process of the BP neural network model 36 is as follows:

[0065] Start the first motor 8, the second motor 10, and the third motor 16 simultaneously, record the rotational speed parameters of the first motor 8, the first rotating shaft 12, and the second rotating shaft 19 at fixed time intervals, and at the same time record the spacing between adjacent wires of the transformer mold 22 to generate the data set 33; In order to obtain more comprehensive data, when collecting the rotational speed parameters, it is necessary to change the working states of the first motor 8, the second motor 10, and the third motor 16 and change the model of the transformer mold 22, and collect sufficient data under different working conditions.

[0066] The input layer of the BP neural network model 36 is an output neuron, which is used to input the spacing between adjacent wires; The output layer of the BP neural network model 36 includes three output neurons, which are respectively used to output the output values of the first rotational speed sensor 25, the second rotational speed sensor 26, and the third rotational speed sensor 27; The hidden layer of the BP neural network model 36 includes two neurons, the activation function is the Sigmoid function, and the two neurons are respectively used to normalize the rotational speed parameters and the spacing between adjacent wire parameters of the data set 33;

[0067] Divide the processed data set 33 into a training set and a test set. In the training set, form an output sample matrix X with the output values of the normalized first rotational speed sensor 25, the second rotational speed sensor 26, and the third rotational speed sensor 27, with a shape of (k, 3), and form an input sample matrix Y with the normalized spacing between adjacent wires, with a shape of (k, 1), where k is the number of training set samples;

[0068] The input sample matrix Y and the output sample matrix X are input into the BP neural network model 36 for iterative training, and the predicted values of the first rotational speed sensor 25, the second rotational speed sensor 26, and the third rotational speed sensor 27 are output respectively. During the training process, the error between the output of the BP neural network model and the actual output is calculated, such as using the mean square error (MSE) as the loss function. Then, the weights and biases are adjusted according to the error backpropagation to continuously optimize the parameters of the network until the preset maximum number of iterations is satisfied.

[0069] To adapt to different winding conditions, the data set 33 is divided into a first data set 34 and a second data set 35. The first data set 34 includes several groups of output values of the first rotational speed sensor 25, the second rotational speed sensor 26, and the third rotational speed sensor 27 corresponding to different adjacent wire spacings under the tight winding condition; the second data set 35 includes several groups of output values of the first rotational speed sensor 25, the second rotational speed sensor 26, and the third rotational speed sensor 27 corresponding to different adjacent wire spacings under the sparse winding condition. The first data set 34 and the second data set 35 are respectively input into the BP neural network model 36 for iterative training, so that the BP neural network model 36 can adapt to different winding conditions.

[0070] In the above, the normalization process of the rotational speed parameter includes mapping the output values of the first rotational speed sensor 25, the second rotational speed sensor 26, and the third rotational speed sensor 27 to the same interval [0,1]. The rotational speed parameter normalization formula is:

[0071] normalized_s=(s - min_s) / (max_s - min_s);

[0072] where, normalized_s is the normalized rotational speed, s is the output rotational speed of the first rotational speed sensor 25, the second rotational speed sensor 26, and the third rotational speed sensor 27, max_s is the maximum rotational speed, and min_s is the minimum rotational speed;

[0073] The normalization formula for the adjacent wire spacing parameter is:

[0074] normalized_gap=(gap - min_gap) / (max_gap - min_gap);

[0075] where, normalized_gap is the normalized adjacent wire spacing, gap is the adjacent wire spacing output by the image recognition module 28, max_gap is the maximum adjacent wire spacing, and min_gap is the minimum adjacent wire spacing.

[0076] In order to ensure that the output value of the BP neural network model 36 accurately controls the motor, in this embodiment, the control system further includes a motor drive module 38, which is respectively connected to the first motor 8, the second motor 10, and the third motor 16, and is used to drive the first motor 8, the second motor 10, and the third motor 16. The BP neural network model 36 is connected to the motor drive module 38 through a signal conversion module 37. The signal conversion module 37 is used to convert the output signal of the BP neural network model 36 into a pulse signal. The motor drive module 38 controls the rotation speed, forward and reverse rotation of the first motor 8, the second motor 10, and the third motor 16 based on the pulse signal.

[0077] The embodiments of the present invention have been described above. The above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art in the technical field without departing from the scope and spirit of the described embodiments. The selection of the terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other ordinary skill in the art in the technical field to understand the disclosed embodiments.

Claims

1. Automatic winding device for transformer, characterized in that, Including: A column (2) and a drive box (14) arranged on a bottom plate (1). A rack (3) is longitudinally arranged on the outer side of the column (2), and a sliding sleeve (4) is sleeved on the column (2). A longitudinal movement component is arranged on one side of the sliding sleeve (4) for driving the sliding sleeve (4) to reciprocate longitudinally along the column (2). A wire reel drive component is arranged on the other side of the sliding sleeve (4) for driving a wire reel (13) to rotate. The drive box (14) is used for driving a transformer mold (22) to rotate. The position of the wire reel (13) is adapted to the position of the transformer mold (22) so that when the wire reel (13) and the transformer mold (22) rotate simultaneously, winding of the transformer mold (22) is realized; A control system, including an input layer, a perception layer and an application layer. The input layer is used for inputting the specification parameters of the transformer mold (22) and the adjacent wire spacing parameters, and judging the winding working condition of the transformer mold (22) based on the specification parameters and the adjacent wire spacing parameters; The perception layer is used for obtaining the rotational speed parameters of the longitudinal movement component, the wire reel drive component and the drive box (14); The application layer is used for obtaining the actual adjacent wire spacing of the transformer mold (22); Based on the winding working condition, regulating the rotational speed parameters obtained by the perception layer through a BP neural network model (36) so that the actual adjacent wire spacing obtained by the application layer is equal to the adjacent wire spacing parameters; The longitudinal movement component includes two parallel fixed plates (5). A gear (6) is arranged between the two fixed plates (5). The fixed plates (5) are fixedly connected to one side of the sliding sleeve (4). The gear (6) is rotatably connected to the fixed plates (5) and meshes with the rack (3). A fixed frame is arranged on the top of the fixed plate (5), and a first motor (8) is arranged on the fixed frame. The output end of the first motor (8) is connected with a drive shaft (7). A spiral tooth is arranged on the drive shaft (7), and the spiral tooth meshes with the gear (6); The wire reel drive component includes a mounting frame (9). The mounting frame (9) is fixedly connected to the other side of the sliding sleeve (4). A second motor (10) and a reducer (11) are arranged on the mounting frame (9). The second motor (10) is connected to the reducer (11), and a first rotating shaft (12) for mounting the wire reel (13) is arranged at the output end of the reducer (11); Inside the drive box (14), a third motor (16) and a support column (15) are arranged. The top of the support column (15) is rotatably connected with a first pulley (17). The output end of the third motor (16) is provided with a second pulley (18). The first pulley (17) and the second pulley (18) are connected by a belt. A second rotating shaft (19) for mounting the transformer mold (22) is arranged on the top of the first pulley (17).

2. The automatic transformer winding device according to claim 1, characterized in that The input layer includes: An input unit (23) for inputting the specification parameters of the transformer mold (22) and the adjacent wire spacing parameters; A judgment unit (24), connected to the input unit (23), for judging and outputting the winding working condition of the transformer mold (22); The judgment process of the judgment unit (24) is as follows: Extract the winding width w, the number of winding turns s, and the wire diameter k based on the specification parameters of the transformer mold (22), then the adjacent wire spacing parameter is expressed as , if The result of is in the first interval, and the winding condition output by the judgment unit (24) is tight winding; if The result of is in the second interval, and the winding condition output by the judgment unit (24) is sparse winding; the maximum value of the first interval is less than or equal to the minimum value of the second interval.

3. The automatic transformer winding device according to claim 2, wherein, The sensing layer includes: a first rotational speed sensor (25) for monitoring the rotational speed of the first motor (8) in real time; a second rotational speed sensor (26) for monitoring the rotational speed of the first rotating shaft (12) in real time; a third rotational speed sensor (27) for monitoring the rotational speed of the second rotating shaft (19) in real time.

4. The automatic transformer winding device according to claim 3, characterized in that The application layer is an image recognition module (28) for monitoring the adjacent wire spacing of the transformer mold in real time; The image recognition module (28) includes an image acquisition unit (29), an image processing unit (30), an edge detection unit (31), and a spacing calculation unit (32); The image acquisition unit (29) is used to acquire the image information of the transformer mold (22) at fixed time intervals; The image processing unit (30) is used to perform grayscale processing and filtering processing on the image information; The edge detection unit (31) extracts the edges of two adjacent windings based on the Canny edge detection operator, performs curve fitting on the edge points of the two adjacent windings detected, constructs a two-dimensional coordinate system based on the graphic information, and obtains the coordinates (x i , y i ) of a number of edge points, and the fitted curve is: y = a0 + a1x + a2x 2 +…+ a n x n ; Among them, i is used to represent different edge point coordinates, a0, a1, a2... a n are coefficients, n is used to determine the number of terms of the fitting curve polynomial, and is determined by minimizing the sum of squared errors ; in the formula, is the ordinate of the corresponding point x i on the fitting curve, and m is the number of points; The spacing calculation unit (32) is configured to calculate the spacing between two adjacent windings ; the calculation process is as follows: equidistant sampling is performed on the obtained two adjacent fitted curves to obtain corresponding points on the two curves and ; Based on corresponding points and , ; Among them, (x1, y1) and (x2, y2) are the coordinates of points and respectively.

5. The automatic transformer winding device according to claim 4, characterized in that, The construction process of the BP neural network model (36) is as follows: Start the first motor (8), the second motor (10), and the third motor (16) at the same time, record the rotational speed parameters of the first motor (8), the first rotating shaft (12), and the second rotating shaft (19) at fixed time intervals, and record the adjacent wire spacing of the transformer mold (22) at the same time to generate a data set (33); The input layer of the BP neural network model (36) is an output neuron for inputting the adjacent wire spacing; The output layer of the BP neural network model (36) includes three output neurons, which are respectively used to output the output values of the first rotational speed sensor (25), the second rotational speed sensor (26), and the third rotational speed sensor (27); The hidden layer of the BP neural network model (36) includes two neurons, the activation function is the Sigmoid function, and the two neurons are respectively used to perform normalization processing on the rotational speed parameters and adjacent wire spacing parameters of the data set (33); Divide the processed data set (33) into a training set and a test set. In the training set, the output values of the first rotational speed sensor (25), the second rotational speed sensor (26), and the third rotational speed sensor (27) after normalization are composed into an output sample matrix X with a shape of (k, 3), and the adjacent wire spacing after normalization is composed into an input sample matrix Y with a shape of (k, 1), where k is the number of training set samples; Input the input sample matrix Y and the output sample matrix X into the BP neural network model (36) for iterative training, and output the predicted values of the first rotational speed sensor (25), the second rotational speed sensor (26), and the third rotational speed sensor (27) respectively.

6. The automatic transformer winding device according to claim 5, characterized in that, The data set (33) includes a first data set (34) and a second data set (35). The first data set (34) includes several groups of output values of the first rotational speed sensor (25), the second rotational speed sensor (26), and the third rotational speed sensor (27) corresponding to different adjacent wire spacings under the tight winding working condition; The second data set (35) includes output values of a number of first speed sensors (25), second speed sensors (26), and third speed sensors (27) corresponding to different adjacent wire spacings under the sparse winding condition.

7. The automatic transformer winding device according to claim 6, characterized in that, The normalization process of the speed parameter includes mapping the output values of the first speed sensor (25), the second speed sensor (26), and the third speed sensor (27) to the same interval [0,1]. The speed parameter normalization formula is: normalized_s=(s-min_s) / (max_s-min_s); where normalized_s is the normalized speed, s is the output speed of the first speed sensor (25), the second speed sensor (26), and the third speed sensor (27), max_s is the maximum speed, and min_s is the minimum speed; The normalization formula for the adjacent wire spacing parameter is: normalized_gap=(gap-min_gap) / (max_gap-min_gap); where normalized_gap is the normalized adjacent wire spacing, gap is the adjacent wire spacing output by the image recognition module (28), max_gap is the maximum adjacent wire spacing, and min_gap is the minimum adjacent wire spacing.

8. The automatic transformer winding device according to claim 7, characterized in that, The control system further includes a motor drive module (38), which is respectively connected to the first motor (8), the second motor (10), and the third motor (16) for driving the first motor (8), the second motor (10), and the third motor (16). The BP neural network model (36) is connected to the motor drive module (38) through a signal conversion module (37). The signal conversion module (37) is used to convert the output signal of the BP neural network model (36) into a pulse signal. The motor drive module (38) controls the speed, forward and reverse rotation of the first motor (8), the second motor (10), and the third motor (16) based on the pulse signal.

Images