A multi-axis high-speed machining system and method for transformer winding insulation parts

Abstract

The present invention relates to the technical field of transformer winding insulation processing, and specifically to a multi-axis high-speed processing system and method for transformer winding insulation. The system uses incoming material multi-axis adjustment and processing multi-axis adjustment equipment to perform multi-degree-of-freedom adjustment on transformer winding insulation, which can meet the processing requirements of different transformer winding insulation parts and improve the processing flexibility and applicability. Intelligent control is achieved through model training, and different types of transformer winding insulation parts can be automatically identified and adjusted to achieve intelligent measurement and control of transformer winding insulation parts. The multi-axis motion parameters of the processing equipment are adjusted to accurately adapt to transformer winding insulation parts of different shapes and characteristics, thereby improving the adaptability and intelligence level of the system. The fault model learns the correspondence strategy between the vibration of the processing equipment, the temperature of the transformer winding insulation parts and the processing equipment and the shape, size and position of the transformer winding insulation parts under different fault modes, thereby achieving fault diagnosis and classification of the transformer winding insulation parts.

Description

Technical Field

[0001] The invention relates to the technical field of transformer winding insulation part processing, and in particular to a transformer winding insulation part multi-axis high-speed processing system and method. Background Art

[0002] As the capacity of the power grid increases, the incidence of transformer short-circuit accidents has become dominant, accounting for more than 50% of all accidents, posing a serious threat to the power system. Insulation is the most important component in the transformer to prevent short circuits, and its performance and quality directly affect the reliability and service life of the transformer. In order to improve the ability to resist short circuits, the transformer winding group is a continuous structure, and insulation parts are required as support and oil and gas channels. Therefore, winding insulation parts have become a key rigid product in the transformer manufacturing industry, and the demand for each transformer ranges from thousands to tens of thousands of pieces, and the demand is huge.

[0003] At present, the manufacturing equipment for winding insulation parts is semi-automatic equipment. The domestic use of a mixed working mode of manual operation and machine operation has many disadvantages, which seriously restricts production efficiency, product quality and production safety. The equipment has low processing efficiency and it is difficult to flexibly adjust the raw materials, resulting in low product processing accuracy and material waste; manual operation is used, the automation level is low, the production efficiency is low, and the equipment takes a long time to adjust different varieties. Symmetry adjustment needs to be achieved by personnel observation, and intelligent identification and adjustment cannot be achieved; the system has no intelligent fault handling capability, resulting in safety hazards in the equipment, causing equipment damage or posing a potential threat to personnel.

[0004] The existing transformer winding insulation processing equipment has low processing efficiency, low degree of automation and lack of fault handling capabilities. Therefore, a multi-axis high-speed processing system and method for transformer winding insulation is proposed, which is of great significance for achieving high-speed and high-quality processing of transformer winding insulation parts and efficient, intelligent and effective processing of transformer insulation parts. Summary of the invention

[0005] The invention provides a multi-axis high-speed machining system and method for transformer winding insulation parts.

[0006] The technical solution of the present invention is as follows:

[0007] A multi-axis high-speed machining system for transformer winding insulation parts, comprising:

[0008] Raw data acquisition module: collects images of transformer winding insulation parts in normal operation and different fault modes, and extracts shape, size and position data of transformer winding insulation parts from the images; collects vibration data of posture adjustment module and punch adjustment module in different directions in normal operation and different fault modes; collects temperature data of transformer winding insulation parts, posture adjustment module and punch adjustment module in normal operation and different fault modes, and forms raw data;

[0009] Feature extraction module: used to extract features from the raw data collected in normal operation and different fault modes using wavelet transform after preprocessing with Gaussian filtering, and generate image feature data, vibration feature data, and temperature feature data for model training;

[0010] Model training module: used to establish associations between the extracted image feature data under normal operating mode and the data labels of normal operating mode, and train the machine learning model. The machine learning model learns the corresponding strategy between the image data features under normal operating mode and the shape, size, and position of the transformer winding insulation parts;

[0011] Posture adjustment module: used to adjust the posture of the transformer winding insulation to be processed, including an up-down position adjustment module, a front-to-back position adjustment module, a left-to-right position adjustment module, a rotation adjustment module and a material table, wherein the up-down position adjustment module is used to adjust the up-down position of the transformer winding insulation, the front-to-back position adjustment module is used to adjust the front-to-back position of the transformer winding insulation, the left-to-right position adjustment module is used to adjust the left-to-right position of the transformer winding insulation, the rotation adjustment module is used to adjust the rotation angle of the transformer winding insulation, and the material table is used to place the transformer winding insulation;

[0012] Punch adjustment module: used to adjust the punch structure, arranged above the position adjustment module, including a punch left-right adjustment module, a punch front-back adjustment module, and a punch base. The punch left-right adjustment module is used to adjust the left-right position of the punch, the punch front-back adjustment module is used to adjust the front-back position of the punch, and the punch base is used to connect the punch;

[0013] Fault model training module: used to establish associations between the extracted data features under different fault modes and the data labels of different faults, and train the fault model. The fault model learns the corresponding strategies of the vibration and temperature of the posture adjustment module and the punch adjustment module under different fault modes, the temperature of the transformer winding insulation parts and the fault of the transformer winding insulation parts;

[0014] Fault diagnosis module: The new data collected in real time is input into the trained fault model, and the fault model outputs the corresponding classification label to perform fault classification diagnosis.

[0015] Specifically, the model training module also includes: extracting features from newly acquired images during the actual processing process to obtain image feature data, inputting the image feature data into a trained machine learning model, and the machine learning model outputs the predicted shape, size and position of the transformer winding insulation part according to the learned corresponding strategy, and compares the predicted shape, size and position of the transformer winding insulation part with the actually acquired shape, size and position to obtain the comparison difference.

[0016] If the difference is less than or equal to the set threshold, the processing is normal.

[0017] If the difference is greater than the set threshold, the model training module sends a signal to the posture adjustment module, and the posture adjustment module adjusts the posture of the transformer winding insulation.

[0018] After adjustment, the image of the transformer winding insulation is collected again, and the image feature data is obtained after feature extraction. The image feature data is input into the trained machine learning model, and the machine learning model outputs the predicted shape, size and position of the transformer winding insulation again, which is compared with the shape, size and position of the transformer winding insulation collected again. If the difference is less than or equal to the set threshold, it is processed normally.

[0019] Otherwise, it is determined that a fault has occurred and a fault signal is sent to the fault model training module. The fault model training module performs classification diagnosis of the fault and outputs a fault type label.

[0020] Furthermore, it also includes a posture adjustment module that transmits the real-time motion parameters of the up and down position adjustment module, the front and back position adjustment module, the left and right position adjustment module, the rotation adjustment module and the material table to the punch adjustment module. The punch adjustment module adjusts the motion parameters of the punch left and right adjustment module, the punch front and back adjustment module and the punch base in real time according to the received parameters. The punch adjustment module and the posture adjustment module coordinate and cooperate to achieve real-time posture adjustment of the transformer winding insulation.

[0021] Furthermore, the real-time vibration and temperature data of the posture adjustment module and the punch adjustment module, and the real-time temperature data of the transformer winding insulation parts collected after adjustment are input into the fault model training module. The fault model training module outputs the fault type label based on the correspondence strategy between the vibration and temperature of the posture adjustment module and the punch adjustment module under different fault modes, the temperature of the transformer winding insulation parts and different faults of the transformer winding insulation parts.

[0022] Preferably, the posture adjustment module includes an up-down position adjustment module, a front-back position adjustment module, a left-right position adjustment module, a rotation adjustment module and a material platform arranged from bottom to top;

[0023] The up-down position adjustment module is composed of several up-down adjustment shafts, the front-back position adjustment module is universally connected to the up-down adjustment shafts, the left-right position adjustment module is slidably connected to the front-back position adjustment module, the rotation adjustment module is slidably connected to the left-right position adjustment module, and the sliding direction of the rotation adjustment module is perpendicular to the sliding direction of the left-right position adjustment module; the material platform is rotationally connected to the rotation adjustment module;

[0024] The vibration sensor is arranged on the up and down position adjustment module, the front and back position adjustment module, the left and right position adjustment module, the rotation adjustment module and the material table, and the temperature sensor is arranged on the up and down position adjustment module, the front and back position adjustment module, the left and right position adjustment module, the rotation adjustment module and the material table.

[0025] Preferably, the punch adjustment module includes a bracket and a punch left-right adjustment module, a punch front-back adjustment module, and a punch base which are arranged on the bracket and are sequentially connected from top to bottom. The punch front-back adjustment module is slidably arranged on the punch left-right adjustment module, and the punch base is slidably arranged on the punch front-back adjustment module, and the sliding direction of the punch base is perpendicular to the punch front-back adjustment module.

[0026] The vibration sensor and the temperature sensor are both arranged on the punch left-right adjustment module, the punch front-back adjustment module and the punch base.

[0027] Specifically, wavelet transform is used in the feature extraction module to perform feature extraction, including extracting shape, size and position data features of the transformer winding insulation image, extracting vibration data features of each axis of the posture adjustment module and the punch adjustment module in different directions, and extracting temperature data features of the transformer winding insulation, posture adjustment module and punch adjustment module.

[0028] Specifically, the vibration data in the data acquisition module includes: amplitude, frequency, and vibration mode data of the posture adjustment module and the punch adjustment module.

[0029] Furthermore, the data labels of different faults in the fault model training module include mechanical fault labels, electrical fault labels, and thermal fault labels.

[0030] The beneficial effects of the present invention are:

[0031] 1. The present invention uses a posture adjustment module and a punch adjustment module to perform multi-degree-of-freedom adjustment on transformer winding insulation parts, which can meet the processing requirements of different transformer winding insulation parts and has high processing flexibility. The posture adjustment module and the punch adjustment module cooperate with each other and can accurately adapt to transformer winding insulation parts of different shapes, sizes and positions, thereby improving the intelligence level of the system.

[0032] 2. The model training module establishes a corresponding strategy between the acquired image and the shape, size and position of the transformer winding insulation parts based on the intelligent algorithm. At the same time, it cooperates with the posture adjustment module and the punch adjustment module to realize the intelligent measurement and control of the transformer winding insulation parts.

[0033] 3. Intelligent fault diagnosis function can be realized. The fault model training module learns the vibration and temperature of the posture adjustment module and the punch adjustment module under different fault modes, the temperature of the transformer winding insulation parts and the corresponding strategy of the transformer winding insulation parts fault, so as to realize the fault diagnosis and classification of the transformer winding insulation parts. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In the attached picture:

[0035] Figure 1 It is a structural schematic diagram of a posture adjustment module for multi-axis high-speed machining of transformer winding insulation parts in an embodiment;

[0036] Figure 2 It is a structural schematic diagram of a punch adjustment module for multi-axis high-speed machining of transformer winding insulation parts in an embodiment;

[0037] The components represented by the reference numerals in the figure are:

[0038] 1. Up and down adjustment shaft; 2. Conveying shaft; 3. Conveying shaft linear guide; 4. Conveying shaft slider; 5. Left and right adjustment shaft; 6. Left and right adjustment shaft guide; 7. Left and right adjustment slider; 8. Rolling adjustment shaft; 9. Rotating bracket; 10. Material table; 11. Processing left and right adjustment shaft; 12. Left and right adjustment guide; 13. Processing left and right adjustment slider; 14. Processing front and back adjustment shaft; 15. Front and back adjustment guide; 16. Front and back adjustment slider; 17. Punch base; 18. Punch. DETAILED DESCRIPTION

[0039] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings.

[0040] Example

[0041] This embodiment provides a multi-axis high-speed machining system for transformer winding insulation parts. Figure 1 , including the following steps:

[0042] Raw data acquisition module: collects images of transformer winding insulation parts under normal operation and different fault modes, and extracts shape, size and position data of transformer winding insulation parts from the images; collects vibration data of posture adjustment module and punch adjustment module in different directions under normal operation and different fault modes; collects temperature data of transformer winding insulation parts, posture adjustment module and punch adjustment module under normal operation and different fault modes to form raw data.

[0043] Specifically, in the present invention, a high-resolution CCD camera is used for image acquisition. The high-resolution CCD camera is installed above the material table 10, and can capture images of transformer winding insulation parts under normal operation and different fault modes, and extract the shape, size, and position data of the transformer winding insulation parts from the images; a vibration sensor is used to respectively capture vibration data of processing equipment in different directions under normal operation and different fault modes, and the vibration sensor is arranged on the posture adjustment module and the punch adjustment module; a temperature sensor is used to respectively capture temperature data of transformer winding insulation parts and processing equipment under normal operation and different fault modes, and the temperature sensor is arranged on the posture adjustment module and the punch adjustment module.

[0044] Feature extraction module: It is used to extract features from the raw data collected under normal operation and different fault modes after preprocessing with Gaussian filtering, and then use wavelet transform to generate image feature data, vibration feature data, and temperature feature data for model training.

[0045] Among them, wavelet transform is used for feature extraction, including extracting the shape, size and position data features of the transformer winding insulation image, extracting the vibration data features of each axis of the posture adjustment module and the punch adjustment module in different directions, and extracting the temperature data features of the transformer winding insulation, posture adjustment module and punch adjustment module. The vibration data features include amplitude, frequency, vibration mode data, and the temperature data features are the temperature change rate.

[0046] The shape, size and position data of the transformer winding insulation collected by the original data acquisition module, the vibration data of the posture adjustment module and the punch adjustment module in different directions, and the temperature data of the transformer winding insulation, the posture adjustment module and the punch adjustment module are Gaussian filtered by a Gaussian filter to eliminate data noise and smooth the signal to obtain preprocessed data; appropriate wavelet basis functions and scales are selected to perform wavelet transform on the preprocessed data, and the data signal is decomposed into wavelet coefficients at different scales and frequencies to obtain a wavelet representation in the time-frequency domain; feature selection and extraction are performed according to the statistical characteristics of the wavelet coefficients to generate data features for model training.

[0047] Model training module: used to establish an association between the extracted image feature data under normal operating mode and the data label of normal operating mode, and train the machine learning model. The machine learning model learns the correspondence strategy between the image data features under normal operating mode and the shape, size and position of the transformer winding insulation parts.

[0048] Specifically, the model establishes the association between the image feature data and the shape, size and position of the transformer winding insulation under the normal operation mode by learning the relationship between the input data features and the corresponding output normal operation labels. The present invention selects an intelligent control algorithm based on deep dual Q learning (D-DQN) to train the neural network model, initializes the neural network structure and parameters of the intelligent control algorithm based on D-DQN, inputs the data features extracted from the feature extraction module into the neural network model, and uses the reinforcement learning framework to train the neural network model. The model learns the corresponding strategy between the image data features and the shape, size and position of the transformer winding insulation under the normal operation mode, and obtains a trained machine learning model.

[0049] Posture adjustment module: used to adjust the posture of the transformer winding insulation parts to be processed, including an up and down position adjustment module, a front and back position adjustment module, a left and right position adjustment module, a rotation adjustment module and a material table, wherein the up and down position adjustment module is used to adjust the up and down position of the transformer winding insulation parts, the front and back position adjustment module is used to adjust the front and back position of the transformer winding insulation parts, the left and right position adjustment module is used to adjust the left and right position of the transformer winding insulation parts, the rotation adjustment module is used to adjust the rotation angle of the transformer winding insulation parts, and the material table is used to place the transformer winding insulation parts.

[0050] The posture adjustment module includes an up-down position adjustment module, a front-back position adjustment module, a left-right position adjustment module, a rotation adjustment module and a material table arranged from bottom to top;

[0051] The up-down position adjustment module is composed of several up-down adjustment shafts, the front-back position adjustment module is universally connected to the up-down adjustment shafts, the left-right position adjustment module is slidably connected to the front-back position adjustment module, the rotation adjustment module is slidably connected to the left-right position adjustment module, and the sliding direction of the rotation adjustment module is perpendicular to the sliding direction of the left-right position adjustment module; the material platform is rotationally connected to the rotation adjustment module;

[0052] The vibration sensor is arranged on the up and down position adjustment module, the front and back position adjustment module, the left and right position adjustment module, the rotation adjustment module and the material table, and the temperature sensor is arranged on the up and down position adjustment module, the front and back position adjustment module, the left and right position adjustment module, the rotation adjustment module and the material table.

[0053] In this embodiment, four upper and lower adjustment shafts 1 are symmetrically distributed on the left and right sides. The front and rear position adjustment module is two conveying shafts 2 fixedly connected to the upper and lower adjustment shafts 1 on the left and right sides respectively, and each conveying shaft 2 is sequentially provided with a conveying shaft linear guide 3 and a conveying shaft slider 4 that can slide along the conveying shaft linear guide 3; the left and right position adjustment module is a left and right adjustment shaft 5 fixedly connected above the left and right conveying shaft sliders 4, and the left and right adjustment shaft 5 is sequentially provided with left and right adjustment shaft guides 6 and left and right adjustment sliders 7 that can slide along the left and right adjustment guides 6; a rotating bracket 9 is fixedly provided on the left and right adjustment sliders 7, and the rotation adjustment module is a rolling adjustment shaft 8 connected to the rotating bracket 9 bearing, and a material table 10 is connected above the rolling adjustment shaft 8.

[0054] The upper and lower adjustment shafts 1 specifically realize the upper and lower position adjustment of the transformer winding insulation parts through the operation of the internal hydraulic cylinder. By adjusting the up and down movement of the four upper and lower adjustment shafts 1, the upper and lower positions and flipping postures of the transformer winding insulation parts can be adjusted; by adjusting the sliding of the left and right position adjustment modules on the front and rear position adjustment modules, the front and rear positions of the transformer winding insulation parts can be adjusted; by adjusting the sliding of the rotation adjustment modules on the left and right position adjustment modules, the left and right positions of the transformer winding insulation parts can be adjusted; by rotating the rolling adjustment shaft 8, the rotation angle of the transformer winding insulation parts can be adjusted.

[0055] In the posture adjustment module, a real-time adjustment strategy is determined based on the shape, size and position of the transformer winding insulation parts to achieve the adjustment of the position and posture of the transformer winding insulation parts. The position and posture of the transformer winding insulation parts are adjusted through the coordinated movement of the up and down adjustment axis 1, the left and right adjustment axis 5, the rolling adjustment axis 8 and the conveying axis 2.

[0056] Punch adjustment module: used to adjust the punch structure, arranged above the posture adjustment module, including a punch left and right adjustment module, a punch front and back adjustment module, and a punch base. The punch left and right adjustment module is used to adjust the left and right position of the punch, the punch front and back adjustment module is used to adjust the front and back position of the punch, and the punch base is used to connect the punch 18.

[0057] The punch adjustment module includes a bracket and a punch left-right adjustment module, a punch front-back adjustment module, and a punch base which are arranged on the bracket and connected in sequence from top to bottom. The punch front-back adjustment module is slidably arranged on the punch left-right adjustment module, and the punch base is slidably arranged on the punch front-back adjustment module. The sliding direction of the punch base is perpendicular to the sliding direction of the punch front-back adjustment module.

[0058] The vibration sensor and the temperature sensor are both arranged on the punch left-right adjustment module, the punch front-back adjustment module and the punch base.

[0059] In this embodiment, the punch left-right adjustment module is a processing left-right adjustment shaft 11, and the left-right adjustment guide rails 12 and the left-right adjustment slider 13 that can slide along the left-right adjustment guide rails 12 are arranged in sequence below the processing left-right adjustment shaft 11; the punch front-back adjustment module is a processing front-back adjustment shaft 14 arranged below the processing left-right adjustment slider 13, and the front-back adjustment guide rails 15 and the front-back adjustment slider 16 that can slide along the front-back adjustment guide rails 15 are arranged in sequence below the processing front-back adjustment shaft 14, a punch base 17 is connected below the front-back adjustment slider 16, and a punch is connected below the punch base 17.

[0060] The punch adjustment module determines its own adjustment parameters based on the real-time adjustment strategy of the posture adjustment module to achieve high-speed and stable punching of transformer winding insulation parts. The speed, acceleration and movement direction parameters of each axis in the punch left and right adjustment module and the punch front and back adjustment module are adjusted to ensure that the punch accurately covers the entire transformer winding insulation parts during punching.

[0061] The punch adjustment module and the posture adjustment module cooperate with each other to realize the intelligent control and adjustment of the transformer winding insulation. The posture adjustment module transmits the real-time motion parameters of the up-down position adjustment module, the front-back position adjustment module, the left-right position adjustment module, the rotation adjustment module and the material table to the punch adjustment module. The punch adjustment module adjusts the motion parameters of the left-right adjustment module, the front-back adjustment module and the punch base in real time according to the received parameters. The punch adjustment module and the posture adjustment module coordinate to realize the real-time posture adjustment of the transformer winding insulation.

[0062] Fault model training module: used to associate the extracted data features under different fault modes with the data labels of different faults, train the fault model, and learn the corresponding strategies between the vibration and temperature of the posture adjustment module and the punch adjustment module under different fault modes, the temperature of the transformer winding insulation and the fault of the transformer winding insulation. Among them, the data labels of different faults include mechanical fault labels, electrical fault labels, and thermal fault labels.

[0063] The present invention selects convolutional neural network (CNN) for fault model training, based on the data features extracted from different fault modes in the feature extraction module, in the fault mode, different fault features are included, such as sudden disappearance of vibration data or sudden change of amplitude, sudden change of vibration frequency, sharp rise or fall of temperature data, etc. The data input into the convolutional neural network is sequentially processed by convolution to capture local features, reduced by pooling to reduce data dimension, and mapped to different fault categories by full connection. At the same time, the loss function and optimizer are selected to adjust the model parameters to minimize the loss function, and a trained fault model is obtained.

[0064] Fault diagnosis module: The new data collected in real time is input into the trained fault model, and the fault model outputs the corresponding classification label to perform fault classification diagnosis.

[0065] The data of the multi-axis high-speed machining system collected in real time is input into the trained CNN model. The trained CNN model extracts features from the data, including vibration data features and temperature data features. The vibration data features include vibration frequency, vibration amplitude and mode, and the temperature data features are the temperature change rate. The model outputs classification labels corresponding to the data features to perform classification diagnosis of faults. According to the results of the fault classification diagnosis, the system automatically triggers the corresponding processing method. If it is in a normal state, it will continue to monitor the various indicators and sensor data of the machining system to ensure that the machining equipment operates stably within the normal range. If it is in a faulty operation state, it will trigger an alarm, notify relevant maintenance personnel, and generate a fault report.

[0066] In the present invention, the model training module also includes: extracting features from newly acquired images in the actual processing process to obtain image feature data, inputting the image feature data into the trained machine learning model, and the machine learning model outputs the shape, size and position of the transformer winding insulation part predicted in this operating state according to the learned corresponding strategy, and compares the shape, size and position of the transformer winding insulation part actually acquired to obtain the difference value of the comparison,

[0067] If the difference is less than or equal to the set threshold, the processing is normal.

[0068] If the difference is greater than the set threshold, the model training module sends a signal to the posture adjustment module, and the posture adjustment module adjusts the posture of the transformer winding insulation.

[0069] After adjustment, the image of the transformer winding insulation is collected again, and the image feature data is obtained after feature extraction. The image feature data is input into the trained machine learning model, and the machine learning model outputs the predicted shape, size and position of the transformer winding insulation again, which is compared with the shape, size and position of the transformer winding insulation collected again. If the difference is less than or equal to the set threshold, it is processed normally.

[0070] Otherwise, it is determined that a fault has occurred and a fault signal is sent to the fault model training module. The fault model training module performs classification diagnosis of the fault and outputs a fault type label.

[0071] The real-time vibration and temperature data of the posture adjustment module and the punch adjustment module and the real-time temperature data of the transformer winding insulation parts collected after adjustment are input into the fault model training module. The fault model training module outputs the fault type label based on the correspondence strategy between the vibration and temperature of the posture adjustment module and the punch adjustment module under different fault modes, the temperature of the transformer winding insulation parts and different faults of the transformer winding insulation parts.

Claims

1. A multi-axis high-speed machining system for transformer winding insulation parts, characterized in that: include: Raw data acquisition module: collects images of transformer winding insulation parts in normal operation and different fault modes, and extracts shape, size and position data of transformer winding insulation parts from the images; The vibration data of the posture adjustment module and the punch adjustment module in different directions under normal operation and different fault modes are collected respectively; The temperature data of the transformer winding insulation, the posture adjustment module and the punch adjustment module in normal operation and different fault modes are collected to form raw data; Feature extraction module: used to extract features from the raw data collected in normal operation and different fault modes using wavelet transform after preprocessing with Gaussian filtering, and generate image feature data, vibration feature data, and temperature feature data for model training; Model training module: used to establish associations between the extracted image feature data under normal operating mode and the data labels of normal operating mode, and train the machine learning model. The machine learning model learns the corresponding strategy between the image data features under normal operating mode and the shape, size, and position of the transformer winding insulation parts; Posture adjustment module: used to adjust the posture of the transformer winding insulation to be processed, including an up-down position adjustment module, a front-to-back position adjustment module, a left-to-right position adjustment module, a rotation adjustment module and a material table, wherein the up-down position adjustment module is used to adjust the up-down position of the transformer winding insulation, the front-to-back position adjustment module is used to adjust the front-to-back position of the transformer winding insulation, the left-to-right position adjustment module is used to adjust the left-to-right position of the transformer winding insulation, the rotation adjustment module is used to adjust the rotation angle of the transformer winding insulation, and the material table is used to place the transformer winding insulation; Punch adjustment module: used to adjust the punch structure, arranged above the position adjustment module, including a punch left-right adjustment module, a punch front-back adjustment module, and a punch base. The punch left-right adjustment module is used to adjust the left-right position of the punch, the punch front-back adjustment module is used to adjust the front-back position of the punch, and the punch base is used to connect the punch; Fault model training module: used to establish associations between the extracted data features under different fault modes and the data labels of different faults, and train the fault model. The fault model learns the corresponding strategies of the vibration and temperature of the posture adjustment module and the punch adjustment module under different fault modes, the temperature of the transformer winding insulation parts and the fault of the transformer winding insulation parts; The machine learning model outputs the predicted shape, size and position of the transformer winding insulation parts according to the learned corresponding strategy, and compares them with the actual collected shape, size and position of the transformer winding insulation parts to obtain the difference. If the difference is less than or equal to the set threshold, the processing is normal. If the difference is greater than the set threshold, the posture adjustment is performed. Fault diagnosis module: The new data collected in real time is input into the trained fault model, and the fault model outputs the corresponding classification label to perform fault classification diagnosis.

2. A multi-axis high-speed machining system for transformer winding insulation parts according to claim 1, characterized in that: The model training module also includes: extracting features from newly acquired images during the actual processing process to obtain image feature data, inputting the image feature data into a trained machine learning model, and the machine learning model outputs the predicted shape, size and position of the transformer winding insulation part according to the learned corresponding strategy, and compares the predicted shape, size and position of the transformer winding insulation part with the actually acquired shape, size and position to obtain the comparison difference. If the difference is less than or equal to the set threshold, the processing is normal. If the difference is greater than the set threshold, the model training module sends a signal to the posture adjustment module, and the posture adjustment module adjusts the posture of the transformer winding insulation. After adjustment, the image of the transformer winding insulation is collected again, and the image feature data is obtained after feature extraction. The image feature data is input into the trained machine learning model, and the machine learning model outputs the predicted shape, size and position of the transformer winding insulation again, which is compared with the shape, size and position of the transformer winding insulation collected again. If the difference is less than or equal to the set threshold, it is processed normally. Otherwise, it is determined that a fault has occurred and a fault signal is sent to the fault model training module. The fault model training module performs classification diagnosis of the fault and outputs a fault type label.

3. A multi-axis high-speed machining system for transformer winding insulation parts according to claim 2, characterized in that: It also includes a posture adjustment module that transmits the real-time motion parameters of the up and down position adjustment module, the front and back position adjustment module, the left and right position adjustment module, the rotation adjustment module and the material table to the punch adjustment module. The punch adjustment module adjusts the motion parameters of the punch left and right adjustment module, the punch front and back adjustment module and the punch base in real time according to the received parameters. The punch adjustment module and the posture adjustment module coordinate and cooperate to achieve real-time posture adjustment of the transformer winding insulation parts.

4. The multi-axis high-speed machining system for transformer winding insulation parts according to claim 2, characterized in that: The real-time vibration and temperature data of the posture adjustment module and the punch adjustment module and the real-time temperature data of the transformer winding insulation parts collected after adjustment are input into the fault model training module. The fault model training module outputs the fault type label based on the correspondence strategy between the vibration and temperature of the posture adjustment module and the punch adjustment module under different fault modes, the temperature of the transformer winding insulation parts and different faults of the transformer winding insulation parts.

5. The multi-axis high-speed machining system for transformer winding insulation parts according to claim 1, characterized in that: The posture adjustment module includes an up-down position adjustment module, a front-back position adjustment module, a left-right position adjustment module, a rotation adjustment module and a material platform arranged from bottom to top; The up-down position adjustment module is composed of several up-down adjustment shafts, the front-back position adjustment module is universally connected to the up-down adjustment shafts, the left-right position adjustment module is slidably connected to the front-back position adjustment module, the rotation adjustment module is slidably connected to the left-right position adjustment module, and the sliding direction of the rotation adjustment module is perpendicular to the sliding direction of the left-right position adjustment module; the material platform is rotationally connected to the rotation adjustment module; The vibration sensor is arranged on the up and down position adjustment module, the front and back position adjustment module, the left and right position adjustment module, the rotation adjustment module and the material table, and the temperature sensor is arranged on the up and down position adjustment module, the front and back position adjustment module, the left and right position adjustment module, the rotation adjustment module and the material table.

6. The multi-axis high-speed machining system for transformer winding insulation parts according to claim 1, characterized in that: The punch adjustment module includes a bracket and a punch left-right adjustment module, a punch front-back adjustment module, and a punch base which are arranged on the bracket and are connected in sequence from top to bottom. The punch front-back adjustment module is slidably arranged on the punch left-right adjustment module, and the punch base is slidably arranged on the punch front-back adjustment module. The sliding direction of the punch base is perpendicular to the sliding direction of the punch front-back adjustment module. The vibration sensor and the temperature sensor are both arranged on the punch left-right adjustment module, the punch front-back adjustment module and the punch base.

7. The multi-axis high-speed machining system for transformer winding insulation parts according to claim 1, characterized in that: The feature extraction module uses wavelet transform to perform feature extraction, including extracting the shape, size and position data features of the transformer winding insulation image, extracting the vibration data features of each axis of the posture adjustment module and the punch adjustment module in different directions, and extracting the temperature data features of the transformer winding insulation, the posture adjustment module and the punch adjustment module.

8. The multi-axis high-speed machining system for transformer winding insulation parts according to claim 1, characterized in that: The vibration data in the data acquisition module includes: amplitude, frequency, and vibration mode data of the posture adjustment module and the punch adjustment module.

9. The multi-axis high-speed machining system for transformer winding insulation parts according to claim 1, characterized in that: The data labels of different faults in the fault model training module include mechanical fault labels, electrical fault labels, and thermal fault labels.

Images