Machine learning method and system for dry-type transformer manufacturing process optimization

Abstract

The invention relates to the field of electrical equipment manufacturing, in particular to a machine learning method and system for dry-type transformer manufacturing process optimization. According to the method, production process parameters and quality index data of the dry-type transformer are collected, correlation coefficient screening features are calculated, and ranking is eliminated through recursive features; based on the process parameters, constructing a quality regulation and control model by using support vector regression, and integrating a Bagging method; performing machine vision quality detection on the produced transformer, and judging defects; for a transformer traceability process with a quality problem, adjusting correlation coefficients and model parameters; the intelligent level of dry-type transformer manufacturing is improved, online quality prediction and regulation are realized, the detection efficiency and accuracy are improved, a basis is provided for quality problem positioning and improvement, and the production efficiency and the quality level are improved.

Description

Technical Field

[0001] The present invention relates to the field of power equipment manufacturing, and specifically to a machine learning method and system for optimizing the manufacturing process of dry-type transformers. Background Art

[0002] Dry-type transformers are crucial electrical equipment in power systems, and their manufacturing quality directly impacts the safe and stable operation of the power grid. Traditional dry-type transformer manufacturing relies heavily on manual quality control, resulting in low production efficiency and inconsistent quality.

[0003] The manufacturing process for dry-type transformers is complex, involving multiple steps, including winding, impregnation, and curing. The process parameters of each step have a significant impact on product quality. Traditional manufacturing methods rely primarily on operator experience to set and adjust process parameters, lacking scientific guidance and optimization methods, resulting in unstable product quality.

[0004] Dry-type transformer quality inspections primarily rely on manual spot checks, which have limited sampling rates and low efficiency, making it easy to miss defective products. Furthermore, manual inspections are susceptible to subjective factors and inconsistent testing standards, leading to inaccurate quality assessments.

[0005] When quality issues arise, traditional production models lack effective process traceability and parameter adjustment mechanisms. Manual investigation of the cause and process adjustments based on experience are often required, leading to long lead times, low efficiency, and difficulty ensuring the accuracy and effectiveness of adjustments.

[0006] As market demand for dry-type transformers continues to increase, traditional manufacturing models are no longer able to adapt to the requirements of large-scale, high-quality production. There is an urgent need to introduce intelligent manufacturing and management methods to improve production efficiency and product quality consistency.

[0007] In order to solve the above problems, a machine learning method and system for optimizing the dry-type transformer manufacturing process is urgently needed to improve the production efficiency and quality level of dry-type transformers and meet the power system's production needs for high-quality dry-type transformers. Summary of the Invention

[0008] To achieve the above objectives, the present invention provides a machine learning method and system for optimizing the dry-type transformer manufacturing process. The specific technical solutions are as follows:

[0009] Collect process parameters of dry-type transformer production process and obtain quality indicators of dry-type transformer;

[0010] Calculate the correlation coefficient between process parameters and quality indicators, select process parameters that are strongly correlated with quality indicators as features, and use recursive feature elimination to sort the features;

[0011] Based on the collected process parameters, the support vector regression (SVR) algorithm is used to build a dry-type transformer production quality prediction model, and the bagging learning method is integrated into the production quality control model.

[0012] Conduct quality inspection on dry-type transformers after production. Collect image data of the finished dry-type transformers and use machine vision models to detect the image data to determine whether the dry-type transformers have quality defects.

[0013] The process of dry-type transformers with quality problems is traced back, the correlation coefficients between process parameters and quality indicators are adjusted, and the parameters in the dry-type transformer production quality prediction model are adjusted.

[0014] Preferably, the process parameters of the dry-type transformer production process are collected, and the process parameters include: winding tension F, winding speed v, immersion time t imm , immersion temperature T imm , curing temperature T cure , curing time t cure ;

[0015] Collect the quality indicators of dry-type transformers, including: insulation resistance R ins , inductance L, transformation ratio N, core loss P iron , copper loss P copper and partial discharge q;

[0016] Clean the collected process parameter and quality indicator data, remove outliers, and use interpolation to fill missing values; construct a data set based on the preprocessed process parameter and quality indicator data;

[0017] Assume that the entire data set D is represented as: In the dataset, x i =x1,x2,…,x d ] is the characteristic vector of the process parameters of the i-th sample. The characteristic vector of the process parameters contains d process parameters, x d =[F,v,t imm ,T imm ,T cure ,t cure ];y i is the quality index value corresponding to the i-th sample, y i =[y1,y2,…,y d ], y d =[R ins ,L,N,P iron ,P copper ,q], n is the total number of samples.

[0018] Preferably, the Pearson correlation coefficient of each process parameter and the quality index is calculated, and the correlation coefficient is greater than the preset threshold THp r The process parameters are used as candidate features;

[0019] Use the recursive feature elimination (RFE) algorithm to rank the importance of candidate features. The specific steps are as follows:

[0020] Input candidate feature set X = [x1, x2, ..., x d ], where x d represents the dth eigenvector, the target feature number k, k<6, represents the number of features finally selected, and sets the base model M for evaluating feature importance;

[0021] Initialize the feature ranking list to store the feature ranking results; record the current feature set S = [1, 2, ..., d], which represents the index value of the features that are not excluded;

[0022] Repeat the following steps until |S| = k:

[0023] In the current feature subset X S Train the base model M and obtain the model parameters σ;

[0024] Calculate the importance score of each feature: c j =|σ j |,σ j Represents the model feature weight coefficient;

[0025] Find the feature index j with the lowest importance min =argmin j∈S c j ; j min Remove from S;

[0026] Output the feature ranking list R. According to the preset target number of features k, select the first k features in the ranking list R as the final feature subset:

[0027] X k =[X[d-k+1],X[d-k+2],…,X[d]], where X k represents the k selected feature vectors.

[0028] Preferably, the support vector regression SVR algorithm is used to build a dry-type transformer production quality prediction model; the process parameter characteristic value X is selected based on the recursive feature elimination RFE algorithm. k And quality indicator data y i , as the training set of the dry-type transformer production quality prediction model; the objective function of the SVR algorithm is:

[0029]

[0030] Among them, w and b are the parameters of the SVR model, C is the penalty coefficient, ξ i and is the slack variable, ε is the error tolerance, φ(X k ) is the feature mapping function, i is the parameter, and n is the number of samples;

[0031] The Gaussian kernel function RBF is selected as the feature mapping function of SVR, and the hyperparameters A and γ of SVR are optimized through grid search;

[0032] The expression of Gaussian kernel function RBF is: K(X g ,X l )=exp(-γ||X g -X l || 2 ); where γ is the hyperparameter of the kernel function, X g and X l is the feature vector of the two samples; K(X g ,X l ) represents the Gaussian kernel function, which is used to calculate the similarity between two samples in the high-dimensional feature space;

[0033] The steps of grid search are as follows:

[0034] Set the search range and step size for A and γ to generate a grid. For each (A, γ) combination in the grid, perform the following steps: train the model using (A, γ) as the hyperparameters of SVR; evaluate the model performance through cross-validation.

[0035] The (A,γ) with the smallest cross-validation evaluation value is selected as the optimal hyperparameter of SVR.

[0036] Preferably, randomly resampling to generate m training subsets includes: setting D as the original training set, and resampling the training subset D obtained in the fth round f for: in, is a sample randomly drawn from D, n1 is the number of subset samples, and the number of subset samples is equal to the number of original training set samples; the resampling process is repeated m times to obtain m training subsets; For samples The corresponding label value; use the optimized SVR model parameters on each training subset to train the SVR sub-model, and finally obtain m sub-models; for each SVR sub-model, use the following steps to train:

[0037] Use Gaussian kernel function RBF as the feature mapping function of SVR;

[0038] Optimize the hyperparameters A and γ of each sub-model through grid search;

[0039] Use the training subset D i Train the SVR sub-model and obtain the SVR sub-model parameter ω i and b i ;

[0040] Evaluate the performance of m SVR sub-models on the test set, using mean square error (MSE) and mean absolute error (MAE) as evaluation metrics; calculate the MSE and MAE of each sub-model, and determine the sub-model weight coefficient based on their reciprocals;

[0041] The prediction results of m SVR sub-models are weighted averaged to obtain the final dry-type transformer quality index prediction value

[0042] in, is the predicted value of the u-th sub-model for the g-th test sample, α u is the weight coefficient of the u-th sub-model.

[0043] Preferably, image data of a dry-type transformer that has been produced is collected, wherein the image data of the dry-type transformer includes an appearance image of the dry-type transformer and an internal structure image of the dry-type transformer;

[0044] The dry-type transformer appearance image includes a high-definition image of the transformer appearance, which is used to detect surface defects, including cracks and dents; the dry-type transformer internal structure image includes an X-ray image of the transformer internal structure, which is used to detect internal defects;

[0045] Use the YOLOv5 target detection algorithm to locate the defect area in the image; the steps of the YOLOv5 algorithm are as follows: divide the input image into E×E grids; predict B bounding boxes for each grid, each bounding box contains 5 predicted values: x, y, w b ,h b ,c; where x, y are the coordinates of the center of the bounding box, w b ,h b is the width and height of the bounding box, and c is the confidence level; P category probabilities are predicted for each bounding box, indicating the probability that the bounding box belongs to each defect category;

[0046] Based on the confidence level and category probability, some bounding boxes are filtered out to obtain the final defect detection results;

[0047] Input the detected defect area image block into the pre-trained ResNet-50 classification model to determine the defect type; the ResNet-50 classification model includes an input layer, a convolutional layer, a residual block, a fully connected layer, and an output layer;

[0048] The input layer is used to receive defect image blocks. The convolution layer, including the superposition of multiple convolution layers and pooling layers, is used to extract hierarchical features of the image. The residual block, including the series connection of multiple residual blocks, each residual block contains two convolution layers and a short-circuit connection, which is used to train deep networks. The fully connected layer is used to flatten the features extracted by the convolution layer and map them to the probability distribution of defect categories through the fully connected layer. The output layer uses the Softmax activation function to output the probability of each defect category.

[0049] According to the defect type and severity, a quality assessment result is given, which includes qualified and unqualified results. Assuming the defect type is q and the severity is s, the quality assessment function is:

[0050]

[0051] Among them, C pass is the set of defect types allowed to pass, S threshold is the severity threshold, Q(q,s)=1 indicates that the quality assessment is qualified, and Q(q,s)=0 indicates that the quality assessment is unqualified.

[0052] Preferably, for dry-type transformers that fail quality inspection, the process parameter data x of their production process is extracted. fail , and obtain the corresponding quality index y fail ;

[0053] The process parameters x of the unqualified samples fail and quality index y fail Add to the original data set D, and get the updated data set D'=D∪(x fail ,y fail );

[0054] Based on the updated data set D', re-select features, recalculate the correlation coefficients between process parameters and quality indicators, and adjust the feature ranking;

[0055] The SVR model and the Bagging ensemble model are retrained using the updated dataset D' to obtain an updated quality prediction model. The prediction performance of the models before and after the update on the test set is compared using the mean square error (MSE) and mean absolute error (MAE). If the mean square error (MSE) and mean absolute error (MAE) values ​​of the updated model are both smaller than those of the unupdated model, the new model is used to guide the production of dry-type transformers.

[0056] Preferably, a machine learning system for optimizing the manufacturing process of a dry-type transformer is used to implement the machine learning method for optimizing the manufacturing process of a dry-type transformer, comprising: a data acquisition module, a feature calculation module, a production quality control module, a quality inspection module, and a process traceability adjustment module;

[0057] The data acquisition module is used to collect process parameters of the dry-type transformer production process and obtain quality indicators of the dry-type transformer;

[0058] The feature calculation module is used to calculate the correlation coefficient between the process parameters and the quality indicators, select the process parameters that are strongly correlated with the quality indicators as features, and sort the features using recursive feature elimination;

[0059] The production quality control module uses the support vector regression (SVR) algorithm based on the collected process parameters to build a dry-type transformer production quality prediction model and integrates the bagging learning method into the production quality control model;

[0060] The quality inspection module is used to perform quality inspection on the produced dry-type transformers, collect image data of the produced dry-type transformers, detect the image data through a machine vision model, and determine whether the dry-type transformers have quality defects;

[0061] The process traceability adjustment module is used to trace the process of dry-type transformers with quality problems, adjust the correlation coefficients between process parameters and quality indicators, and adjust the parameters in the dry-type transformer production quality prediction model.

[0062] An electronic device includes: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the machine learning method for optimizing the dry-type transformer manufacturing process by calling the computer program stored in the memory.

[0063] A computer-readable storage medium stores instructions. When the instructions are executed on a computer, the computer executes the machine learning method for optimizing the dry-type transformer manufacturing process.

[0064] Beneficial effects of the present invention: The present invention establishes a data association between the production process and product quality by collecting process parameters and quality indicator data of the production process, providing a basis for subsequent data analysis and model construction.

[0065] By calculating the correlation coefficient between process parameters and quality indicators, this method can quantitatively assess the impact of each process parameter on product quality and select key process parameters with the greatest impact as features. Using a recursive feature elimination algorithm to sort features can further optimize feature subsets, remove redundant and noisy features, and improve the model's generalization performance.

[0066] The present invention uses the support vector regression (SVR) algorithm to construct a quality control model, which can establish a nonlinear mapping relationship between process parameters and quality indicators, thereby realizing the prediction and control of production quality. SVR has good generalization performance and noise resistance, and can handle small sample and nonlinear problems. Based on the SVR model, the integration of the bagging learning method can further improve the stability and robustness of the model. By taking a weighted average of the prediction results of multiple SVR sub-models, the differences and fluctuations of a single model can be reduced, and the accuracy of quality prediction can be improved.

[0067] This invention uses machine vision technology to perform quality inspections on finished dry-type transformers, enabling automatic defect identification and location. By capturing images of the transformer's exterior and internal structure and employing target detection and image classification algorithms, surface and internal defects can be quickly and accurately detected, improving the efficiency and accuracy of quality inspections.

[0068] Through process traceability, this method can quickly locate the key process links and parameters that cause quality issues, providing a basis for cause analysis and improvement. Furthermore, by feeding back data on defective products into the original dataset, the correlation coefficients between process parameters and quality indicators are adjusted, and feature selection results are dynamically updated to adapt to changes in the production process and continuously optimize feature subsets. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A flow chart of the machine learning method for optimizing the dry-type transformer manufacturing process provided by the present invention;

[0070] Figure 2 This is a structural diagram of the machine learning system for optimizing the dry-type transformer manufacturing process provided by the present invention. DETAILED DESCRIPTION

[0071] For a better understanding of the present invention, various aspects of the present invention will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are merely descriptions of exemplary embodiments of the present invention and are not intended to limit the scope of the present invention in any way. Throughout this specification, like reference numerals refer to like elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0072] In the accompanying drawings, the size, dimensions, and shapes of the elements have been slightly adjusted for ease of illustration. The accompanying drawings are for illustration purposes only and are not drawn strictly to scale. As used herein, the terms "substantially," "approximately," and similar terms are used to indicate approximate values, not degrees, and are intended to illustrate inherent deviations in measurements or calculations that would be recognized by a person of ordinary skill in the art. In addition, in the present invention, the order in which the steps are described does not necessarily represent the order in which these steps occur in actual operation, unless otherwise specified or inferred from the context.

[0073] It should also be understood that expressions such as "comprises," "including," "having," "includes," and / or "comprising" are open rather than closed expressions in this specification, indicating the presence of the stated features, elements, and / or components, but do not exclude the presence of one or more other features, elements, components, and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present invention, "may" is used to mean "one or more embodiments of the present invention." And, the term "exemplary" is intended to refer to an example or illustration.

[0074] Unless otherwise defined, all terms used herein (including engineering terms and scientific and technological terms) have the same meaning as commonly understood by those skilled in the art to which the present invention pertains. It should also be understood that, unless otherwise expressly stated in the present invention, words defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.

[0075] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0076] Example 1

[0077] Reference Figure 1 , which is the first embodiment of the present invention, provides a machine learning method for optimizing the dry-type transformer manufacturing process.

[0078] S1: Collect process parameters of dry-type transformer production process and obtain quality indicators of dry-type transformer.

[0079] Collect the process parameters of the dry-type transformer production process, including winding tension f, winding speed v, immersion time t imm , immersion temperature T imm , curing temperature T cure, curing time t cure .

[0080] The winding tension F controls the wire tension during the winding process to ensure the winding quality; the winding speed v controls the running speed of the winding machine, affecting the winding efficiency and quality; the dipping time t imm Controlling the time the transformer is immersed in insulating varnish affects the insulation performance; the immersion temperature T imm Controlling the temperature of the insulating varnish affects the viscosity and permeability during the impregnation process; the curing temperature T cure Control the curing temperature of the transformer after impregnation to affect the curing effect of the insulating paint; curing time t cure Controlling the curing time of the transformer after impregnation affects the curing degree of the insulating varnish.

[0081] Collect the quality indicators of dry-type transformers, including: insulation resistance R ins , inductance L, transformation ratio N, core loss P iron , copper loss P copper and partial discharge q.

[0082] Insulation resistance R ins It reflects the main indicators of transformer insulation performance; inductance L reflects the index of transformer inductance; transformation ratio N reflects the ratio of primary and secondary voltages of transformer; core loss P iron Reflects the loss of the transformer under no-load; copper loss P copper It reflects the coil loss of the transformer under load; partial discharge q reflects the local high-frequency discharge caused by transformer insulation defects.

[0083] The collected process parameter and quality indicator data are cleaned, outliers are removed, and missing values ​​are filled using interpolation; a data set is constructed based on the preprocessed process parameter and quality indicator data.

[0084] Assume that the entire data set is represented as: In the dataset, x i =[x1,x2,…,x d is the characteristic vector of the process parameters of the i-th sample. The characteristic vector of the process parameters contains d process parameters. d =[F,v,t imm ,T imm ,T cure ,t cure ];y i is the corresponding quality index value, y i =[y1,y2,…,y d ], y d =[R ins ,T,N,P iron ,P copper,q], n is the total number of samples.

[0085] Normalize the data and scale it to the interval [0, 1]. The normalization formula is: Among them, x is the original data, x min and x max are the minimum and maximum values ​​of the data, respectively, x norm The data are normalized.

[0086] By collecting key process parameters and quality indicators from the dry-type transformer production process and then cleaning and preprocessing the collected data, a high-quality dataset was constructed. This dataset not only includes key factors affecting dry-type transformer quality but also addresses missing and outliers through methods such as interpolation. This provides a solid data foundation for subsequent feature selection and model training, helping to improve the reliability and effectiveness of the optimization method.

[0087] S2: Calculate the correlation coefficient between process parameters and quality indicators, select process parameters that are strongly correlated with quality indicators as features, and use recursive feature elimination to sort the features.

[0088] Calculate the Pearson correlation coefficient p between each process parameter and quality index r , select the correlation coefficient greater than the preset threshold THp r The process parameters are used as candidate features; the Pearson correlation coefficient formula is:

[0089]

[0090] Among them, x z and y z Represent the zth process parameter and quality index value respectively, and Represent the means of the corresponding process parameters and quality indicators respectively.

[0091] The recursive feature elimination (RFE) algorithm is used to sort the importance of candidate features. The basic idea of ​​the RFE algorithm is to repeatedly build the model, removing the least important features from the current feature set each time until the preset number of features is reached.

[0092] The specific steps of the recursive feature elimination RFE algorithm are as follows:

[0093] Input candidate feature set X = [x1, x2, ..., x d ], where x d represents the dth eigenvector, the target feature number k, k<6, represents the number of features finally selected, and sets the base model M to evaluate the importance of features. Usually a linear model such as LinearRegression is selected;

[0094] Initialize the feature ranking list to store the feature ranking results; record the current feature set S = [1, 2, ..., d], which represents the index value of the features that are not excluded;

[0095] Repeat the following steps until |S| = k:

[0096] In the current feature subset X S Train the base model M and obtain the model parameters σ;

[0097] Calculate the importance score c of each feature j =|σ j |,σ j Represents the model feature weight coefficient. The absolute value of the feature weight coefficient is used to represent the importance score of the feature. The larger the absolute value of the weight coefficient, the greater the impact of the feature on the model output and the higher its importance.

[0098] Find the feature index j with the lowest importance min =argmin j∈S c j ; j min Remove from S;

[0099] Output a ranked list R of features, where the features at the end of the list are the least important and the features at the beginning of the list are the most important.

[0100] According to the preset target feature number k, the first k features of the ranking list R are selected as the final feature subset: X k =[X[d-k+1],X[d-k+2],…,X[d]], where X k represents the k selected features.

[0101] By repeatedly training the model and evaluating feature importance, the RFE algorithm can gradually screen out the most relevant and discriminative feature subsets, helping to improve the performance of subsequent quality prediction models.

[0102] By combining the Pearson correlation coefficient with recursive feature elimination, a feature selection method can be used to select key parameters closely related to dry-type transformer quality and of high importance from a wide range of process parameters. This dimensionality reduction and feature selection not only reduces model complexity and improves training efficiency, but also enhances the model's generalization capabilities, enabling more accurate prediction of dry-type transformer quality indicators and providing a reliable basis for quality optimization and control.

[0103] S3: Based on the collected process parameters, the support vector regression (SVR) algorithm is used to construct a dry-type transformer production quality prediction model, and the bagging learning method is integrated into the production quality control model to improve the robustness of the dry-type transformer production quality prediction model.

[0104] The support vector regression (SVR) algorithm is used to build a dry-type transformer production quality prediction model; the process parameter eigenvalues ​​X are selected based on the recursive feature elimination (RFE) algorithm. k And quality indicator data y i , as the training set of the dry-type transformer production quality prediction model; the objective function of the SVR algorithm is:

[0105]

[0106] Among them, w and b are the parameters of the SVR model, C is the penalty coefficient, ξ i and is the slack variable, ε is the error tolerance, φ(X k ) is the feature mapping function, i is the parameter, and n is the number of samples.

[0107] The Gaussian kernel function RBF is selected as the feature mapping function of SVR, and the hyperparameters A and γ of SVR are optimized through grid search.

[0108] The expression of Gaussian kernel function RBF is: K(X g ,X l )=exp(-γ||X g -X l || 2 ); where γ is the hyperparameter of the kernel function, X g and X l is the feature vector of the two samples; K(X g ,X l ) represents the Gaussian kernel function, which is used to calculate the similarity between two samples in the high-dimensional feature space.

[0109] The basic steps of grid search are as follows:

[0110] Set A and γ as the search range and step size of the grid search to generate a grid; for each (A, γ) combination in the grid, perform the following steps: use (A, γ) as the hyperparameters of SVR to train the model; evaluate the model performance through cross-validation.

[0111] The (A,γ) with the smallest cross-validation evaluation value is selected as the optimal hyperparameter of SVR.

[0112] A dry-type transformer quality prediction model, constructed using the Support Vector Regression (SVR) algorithm, can accurately predict various dry-type transformer quality indicators based on key process parameters. The introduction of a Gaussian kernel function and grid search effectively addresses the nonlinear relationship between process parameters and quality indicators, optimizes model hyperparameters, and further improves the accuracy of quality prediction. Quality prediction based on the SVR model can promptly detect abnormal fluctuations in process parameters, providing a basis for adjusting the production process, thereby achieving quality control and optimization of the dry-type transformer manufacturing process.

[0113] Random resampling generates m training subsets, including: setting D as the original training set, and the training subset D obtained by the f-th round of resampling f for: in, is a sample randomly drawn from D, n1 is the number of subset samples, and the number of subset samples is equal to the number of original training set samples; the resampling process is repeated m times to obtain m training subsets; For samples The corresponding label value; use the optimized SVR model parameters on each training subset to train the SVR sub-model, and finally obtain m sub-models; for each SVR sub-model, use the following steps to train:

[0114] Use Gaussian kernel function RBF as the feature mapping function of SVR;

[0115] Optimize the hyperparameters A and γ of each sub-model through grid search;

[0116] Use the training subset D i Train the SVR sub-model and obtain the SVR sub-model parameter ω i and b i ;

[0117] Evaluate the performance of m SVR sub-models on the test set, using mean square error (MSE) and mean absolute error (MAE) as evaluation indicators; calculate the MSE and MAE of each sub-model, and determine the weight coefficient of the sub-model according to their reciprocals:

[0118]

[0119] Among them, MSE u is the mean square error of the u-th sub-model, MAE u is the mean absolute error of the u-th sub-model;

[0120] The prediction results of m SVR sub-models are weighted averaged to obtain the final dry-type transformer quality index prediction value

[0121] in, is the predicted value of the u-th sub-model for the g-th test sample, α u is the weight coefficient of the u-th sub-model.

[0122] The prediction results of the m sub-models are averaged to obtain the final dry-type transformer quality index prediction value: f t (x) is the predicted value of the t-th sub-model for the test sample.

[0123] S4: Perform quality inspection on the dry-type transformers produced, collect image data of the dry-type transformers produced, and inspect the image data through a machine vision model to determine whether the dry-type transformers have quality defects.

[0124] Based on the SVR quality prediction model, the Bagging ensemble learning method is introduced. By randomly resampling the training data, multiple SVR sub-models are constructed and their prediction results are averaged. This effectively reduces the randomness and uncertainty of a single model and improves the stability and robustness of quality prediction results. The Bagging ensemble model can better adapt to process fluctuations and data noise in the dry-type transformer production process, making quality predictions more reliable, thereby guiding the dynamic optimization of production parameters and ensuring consistent product quality.

[0125] Image data of a dry-type transformer that has been produced is collected, wherein the image data of the dry-type transformer includes an appearance image of the dry-type transformer and an internal structure image of the dry-type transformer.

[0126] The dry-type transformer appearance image includes a high-definition image of the transformer's appearance, which is used to detect surface defects, including cracks and dents; the dry-type transformer internal structure image includes an X-ray image of the transformer's internal structure, which is used to detect internal defects, such as insulation failure and short circuit.

[0127] Use the YOLOv5 target detection algorithm to locate the defect area in the image; the basic steps of the YOLOv5 algorithm are as follows: divide the input image into E×E grids; predict B bounding boxes for each grid, each bounding box contains 5 predicted values: x, y, w b ,h b ,c; where x, y are the coordinates of the center of the bounding box, w b ,h b is the width and height of the bounding box, and c is the confidence level. P category probabilities are predicted for each bounding box, indicating the probability that the bounding box belongs to each defect category.

[0128] According to the confidence and category probability, the bounding boxes with low confidence and low category probability are filtered out to obtain the final defect detection results.

[0129] The detected defect area image block is input into a pre-trained ResNet-50 classification model to determine the defect type; the ResNet-50 classification model includes an input layer, a convolutional layer, a residual block, a fully connected layer, and an output layer.

[0130] The input layer is used to receive defect image blocks, and the convolution layer includes a superposition of multiple convolution layers and pooling layers, which is used to extract hierarchical features of the image; the residual block includes a series of multiple residual blocks, each residual block contains two convolution layers and a short-circuit connection, which is used to train deep networks; the fully connected layer is used to flatten the features extracted by the convolution layer and map them to the probability distribution of defect categories through the fully connected layer; the output layer uses a Softmax activation function to output the probability of each defect category.

[0131] Online quality inspection of finished dry-type transformers uses the object detection algorithm YOLOv5 and the image classification model ResNet-50 to quickly and accurately identify external and internal defects in dry-type transformers, enabling automated product quality inspection. Compared to manual inspection, this method is more efficient and accurate, enabling timely detection of quality issues, reducing the outflow of defective products and improving product qualification rates. By comparing test results with pre-set quality assessment standards, product compliance can be automatically determined, providing a basis for quality grading and improvement.

[0132] According to the defect type and severity, a quality assessment result is given, which includes qualified and unqualified results. Assuming the defect type is q and the severity is s, the quality assessment function is:

[0133]

[0134] Among them, C pass is the set of defect types allowed to pass, S threshold is the severity threshold, Q(q,s)=1 indicates that the quality assessment is qualified, and Q(q,s)=0 indicates that the quality assessment is unqualified.

[0135] S5: Conduct process traceability for dry-type transformers with quality problems, adjust the correlation coefficients between process parameters and quality indicators, and adjust the parameters in the dry-type transformer production quality prediction model.

[0136] For dry-type transformers that fail quality inspection, extract the process parameter data x of their production process fail , and obtain the corresponding quality index y fail .

[0137] The process parameters x of the unqualified samples fail and quality index y failAdd to the original data set D, and get the updated data set D'=D∪(x fail ,y fail ).

[0138] Based on the updated data set D', feature selection is performed again, the correlation coefficient between process parameters and quality indicators is calculated again, and the feature ranking is adjusted.

[0139] The SVR model and the Bagging ensemble model are retrained using the updated dataset D' to obtain an updated quality prediction model.

[0140] The prediction performance of the models before and after the update on the test set is compared using the mean square error (MSE) and mean absolute error (MAE). If the mean square error (MSE) and mean absolute error (MAE) values ​​of the updated model are both smaller than those of the unupdated model, the new model is used to guide the production of dry-type transformers.

[0141] For dry-type transformers that fail quality inspections, process parameter traceability and data feedback into the original dataset allow for dynamic updates to the optimization model, enabling closed-loop optimization control of the production process. Newly added unqualified sample data expands the training set, improving the model's generalization and adaptability. Dynamic model updates allow for timely adjustments to the correlation between key process parameters and quality indicators, resulting in more accurate quality predictions and more timely and effective optimization guidance.

[0142] Through the continuous accumulation of production data and continuous self-improvement of the model, the intelligent optimization of the dry-type transformer manufacturing process can be achieved, and product quality and production efficiency can be continuously improved.

[0143] Example 2

[0144] Reference Figure 2 , which is the second embodiment of the present invention, provides a machine learning system for optimizing the dry-type transformer manufacturing process.

[0145] The system includes: a data acquisition module, a feature calculation module, a production quality control module, a quality inspection module and a process traceability adjustment module.

[0146] The data acquisition module is used to collect process parameters of the dry-type transformer production process and obtain quality indicators of the dry-type transformer.

[0147] The feature calculation module is used to calculate the correlation coefficient between the process parameters and the quality indicators, screen out the process parameters that are strongly correlated with the quality indicators as features, and use recursive feature elimination to sort the features.

[0148] The production quality control module uses the support vector regression (SVR) algorithm based on the collected process parameters to build a dry-type transformer production quality prediction model and integrates the bagging learning method into the production quality control model.

[0149] The quality inspection module is used to perform quality inspection on the produced dry-type transformers, collect image data of the produced dry-type transformers, inspect the image data through a machine vision model, and determine whether the dry-type transformers have quality defects.

[0150] The process traceability adjustment module is used to trace the process of dry-type transformers with quality problems, adjust the correlation coefficients between process parameters and quality indicators, and adjust the parameters in the dry-type transformer production quality prediction model.

[0151] Example 3

[0152] The present invention also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories may store computer-readable code that, when executed by the one or more processors, may implement the machine learning method for optimizing the dry-type transformer manufacturing process as described above.

[0153] The method or system according to the embodiment of the present invention may also be implemented with the aid of the architecture of the electronic device of the present invention.

[0154] An electronic device may include a bus, one or more CPUs, read-only memory (ROM), random access memory (RAM), a communication port for connecting to a network, input / output components, a hard disk, and the like.

[0155] A storage device in an electronic device, such as a ROM or a hard disk, can store the machine learning method for optimizing the dry-type transformer manufacturing process provided by the present invention.

[0156] A machine learning method for optimizing the dry-type transformer manufacturing process includes collecting process parameters of the dry-type transformer production process and obtaining quality indicators of the dry-type transformer; calculating the correlation coefficient between the process parameters and the quality indicators, screening the process parameters that are strongly correlated with the quality indicators as features, and sorting the features using recursive feature elimination; based on the collected process parameters, using the support vector regression (SVR) algorithm to construct a dry-type transformer production quality prediction model, and integrating the bagging learning method into the production quality control model; performing quality inspection on the produced dry-type transformers, collecting image data of the completed dry-type transformers, and inspecting the image data through a machine vision model to determine whether the dry-type transformers have quality defects; tracing the process of dry-type transformers with quality problems, adjusting the correlation coefficient between the process parameters and the quality indicators, and adjusting the parameters in the dry-type transformer production quality prediction model.

[0157] Furthermore, the electronic device may further include a user interface. Of course, the architecture of the present invention is only exemplary, and when implementing different devices, one or more components of the electronic device disclosed in the present invention may be omitted according to actual needs.

[0158] Example 4

[0159] The invention also discloses a computer-readable storage medium.

[0160] The computer-readable storage medium has computer-readable instructions stored thereon.

[0161] When the computer-readable instructions are executed by a processor, the machine learning method for optimizing the dry-type transformer manufacturing process according to the embodiments of the present invention described with reference to the above drawings may be executed.

[0162] Storage media include, but are not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), a hard disk, flash memory, etc. In addition, according to embodiments of the present invention, the processes described above with reference to the flowcharts may be implemented as computer software programs.

[0163] For example, the present invention provides a non-temporary machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be run by a processor to execute instructions corresponding to the method steps provided by the present invention, such as: collecting process parameters of the dry-type transformer production process, and obtaining quality indicators of the dry-type transformer; calculating the correlation coefficient between the process parameters and the quality indicators, screening out process parameters that are strongly correlated with the quality indicators as features, and using recursive feature elimination to sort the features; based on the collected process parameters, using the support vector regression SVR algorithm to construct a dry-type transformer production quality prediction model, and integrating the Bagging learning method into the production quality control model; performing quality inspection on the produced dry-type transformers, collecting image data of the completed dry-type transformers, and inspecting the image data through a machine vision model to determine whether the dry-type transformers have quality defects; tracing the process of dry-type transformers with quality problems, adjusting the correlation coefficient between the process parameters and the quality indicators, and adjusting the parameters in the dry-type transformer production quality prediction model.

[0164] When the computer program is executed by a central processing unit (CPU), the functions defined in the method of the present invention are performed. The method, apparatus, and device of the present invention may be implemented in many ways. For example, the method, apparatus, and device of the present invention may be implemented using software, hardware, firmware, or any combination of software, hardware, and firmware.

[0165] The above sequence of steps for the method is for illustration only, and the steps of the method of the present invention are not limited to the sequence specifically described above unless otherwise specifically stated.

[0166] In addition, in some embodiments, the present invention can also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present invention. Therefore, the present invention also covers a recording medium storing a program for executing the method according to the present invention.

[0167] In addition, the parts of the above technical solutions provided in the embodiments of the present invention that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0168] The above-described specific embodiments further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is merely a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A machine learning approach to optimizing the dry-type transformer manufacturing process, characterized by: include: Collect process parameters of dry-type transformer production process and obtain quality indicators of dry-type transformer; Calculate the correlation coefficient between process parameters and quality indicators, select process parameters that are strongly correlated with quality indicators as features, and use recursive feature elimination to sort the features; Based on the collected process parameters, the support vector regression (SVR) algorithm is used to build a dry-type transformer production quality prediction model, and the bagging learning method is integrated into the production quality control model. Conduct quality inspection on dry-type transformers after production. Collect image data of the finished dry-type transformers and use machine vision models to detect the image data to determine whether the dry-type transformers have quality defects. The process of dry-type transformers with quality problems is traced back, the correlation coefficients between process parameters and quality indicators are adjusted, and the parameters in the dry-type transformer production quality prediction model are adjusted.

2. The machine learning method for optimizing the dry-type transformer manufacturing process according to claim 1, characterized in that: Collect the process parameters of the dry-type transformer production process, including winding tension F, winding speed v, immersion time t imm , immersion temperature T imm , curing temperature T cure , curing time t cure ; Collect the quality indicators of dry-type transformers, including: insulation resistance R ins , inductance L, transformation ratio N, core loss P iron , copper loss P copper and partial discharge q; Clean the collected process parameter and quality indicator data, remove outliers, and use interpolation to fill missing values; construct a data set based on the preprocessed process parameter and quality indicator data; Assume that the entire data set D is represented as: In the dataset, x i =x1,x2,…,x d ] is the characteristic vector of the process parameters of the i-th sample. The characteristic vector of the process parameters contains d process parameters, x d =[F,v,t imm ,T imm ,T cure ,t cure ];y i is the quality index value corresponding to the i-th sample, y i =[y1,y2,…,y d ], y d =[R ins ,L,N,P iron ,P copper ,q], n is the total number of samples.

3. The machine learning method for optimizing the dry-type transformer manufacturing process according to claim 2, characterized in that: Calculate the Pearson correlation coefficient between each process parameter and quality index, and select the parameters whose correlation coefficient is greater than the preset threshold THp r The process parameters are used as candidate features; Use the recursive feature elimination (RFE) algorithm to rank the importance of candidate features. The specific steps are as follows: Input candidate feature set X = [x1, x2, ..., x d ], where x d represents the dth eigenvector, the target feature number k, k<6, represents the number of features finally selected, and sets the base model M for evaluating feature importance; Initialize the feature ranking list to store the feature ranking results; record the current feature set S = [1, 2, ..., d], which represents the index value of the features that are not excluded; Repeat the following steps until |S| = k: In the current feature subset X S Train the base model M and obtain the model parameters σ; Calculate the importance score c of each feature j =|σj|, σj represents the model feature weight coefficient; Find the feature index j with the lowest importance min =argmin j∈S c j ; j min Remove from S; Output the feature ranking list R. According to the preset target number of features k, select the first k features in the ranking list R as the final feature subset: X k =[X[d-k+1],X[d-k+2],…,X[d]], where X k represents the k selected features.

4. The machine learning method for optimizing the dry-type transformer manufacturing process according to claim 3, characterized in that: The support vector regression (SVR) algorithm is used to build a dry-type transformer production quality prediction model; the process parameter eigenvalues ​​X are selected based on the recursive feature elimination (RFE) algorithm. k And quality indicator data y i , as the training set of the dry-type transformer production quality prediction model; the objective function of the SVR algorithm is: Among them, w and b are the parameters of the SVR model, C is the penalty coefficient, ξ i and is the slack variable, ε is the error tolerance, φ(X k ) is the feature mapping function, n is the number of samples; The Gaussian kernel function RBF is selected as the feature mapping function of SVR, and the hyperparameters A and γ of SVR are optimized through grid search; The expression of Gaussian kernel function RBF is: K(X g ,X l )=exp(-γ||X g -X l || 2 ); where γ is the hyperparameter of the kernel function, X g and X l is the feature vector of two samples; K(X g ,X l ) represents the Gaussian kernel function, which is used to calculate the similarity between two samples in the high-dimensional feature space; The steps of grid search are as follows: Set the search range and step size for A and γ to generate a grid. For each (A, γ) combination in the grid, perform the following steps: train the model using (A, γ) as the hyperparameters of SVR; evaluate the model performance through cross-validation. The (A,γ) with the smallest cross-validation evaluation value is selected as the optimal hyperparameter of SVR.

5. The machine learning method for optimizing the dry-type transformer manufacturing process according to claim 4, characterized in that: Random resampling generates m training subsets, including: setting D as the original training set, and the training subset D obtained by the f-th round of resampling f for: in, is a sample randomly drawn from D, n1 is the number of subset samples, and the number of subset samples is equal to the number of original training set samples; the resampling process is repeated m times to obtain m training subsets; For samples The corresponding label value; Use the optimized SVR model parameters on each training subset to train the SVR sub-model, and finally obtain m sub-models. For each SVR sub-model, use the following steps to train: Use Gaussian kernel function RBF as the feature mapping function of SVR; Optimize the hyperparameters A and γ of each sub-model through grid search; Use the training subset D i Train the SVR sub-model and obtain the SVR sub-model parameter ω i and b i ; Evaluate the performance of m SVR sub-models on the test set, using mean square error (MSE) and mean absolute error (MAE) as evaluation metrics; calculate the MSE and MAE of each sub-model, and determine the sub-model weight coefficient based on their reciprocals; The prediction results of m SVR sub-models are weighted averaged to obtain the final dry-type transformer quality index prediction value in, is the predicted value of the u-th sub-model for the g-th test sample, α u is the weight coefficient of the u-th sub-model.

6. The machine learning method for optimizing the dry-type transformer manufacturing process according to claim 5, characterized in that: Collecting image data of a dry-type transformer after production, wherein the image data of the dry-type transformer includes an appearance image of the dry-type transformer and an internal structure image of the dry-type transformer; The dry-type transformer appearance image includes an image of the transformer appearance, which is used to detect surface defects, including cracks and dents; the dry-type transformer internal structure image includes an X-ray image of the transformer internal structure, which is used to detect internal defects; Use the YOLOv5 target detection algorithm to locate the defect area in the image; the steps of the YOLOv5 algorithm are as follows: divide the input image into E×E grids; predict B bounding boxes for each grid, each bounding box contains 5 predicted values: x, y, w b ,h b ,c; where x, y are the coordinates of the center of the bounding box, w b ,h b is the width and height of the bounding box, and c is the confidence level; P category probabilities are predicted for each bounding box, indicating the probability that the bounding box belongs to each defect category; Based on the confidence level and category probability, some bounding boxes are filtered out to obtain the final defect detection results; Input the detected defect area image block into the pre-trained ResNet-50 classification model to determine the defect type; the ResNet-50 classification model includes an input layer, a convolutional layer, a residual block, a fully connected layer, and an output layer; The input layer is used to receive defect image blocks. The convolution layer, including the superposition of multiple convolution layers and pooling layers, is used to extract hierarchical features of the image. The residual block, including the series connection of multiple residual blocks, each residual block contains two convolution layers and a short-circuit connection, which is used to train deep networks. The fully connected layer is used to flatten the features extracted by the convolution layer and map them to the probability distribution of defect categories through the fully connected layer. The output layer uses the Softmax activation function to output the probability of each defect category. According to the defect type and severity, a quality assessment result is given, which includes qualified and unqualified results. Assuming the defect type is q and the severity is s, the quality assessment function is: Among them, C pass is the set of defect types allowed to pass, S threshold is the severity threshold, Q(q,s)=1 indicates that the quality assessment is qualified, and Q(q,s)=0 indicates that the quality assessment is unqualified.

7. The machine learning method for optimizing the dry-type transformer manufacturing process according to claim 6, characterized in that: For dry-type transformers that fail quality inspection, extract the process parameter data of their production process x fail , and obtain the corresponding quality index y fail ; The process parameters x of the unqualified samples fail and quality index y fail Add to the original data set D, and get the updated data set D'=D∪(x fail ,y fail ); Based on the updated data set D', re-select features, recalculate the correlation coefficients between process parameters and quality indicators, and adjust the feature ranking; The SVR model and the Bagging ensemble model are retrained using the updated dataset D' to obtain an updated quality prediction model. The prediction performance of the models before and after the update on the test set is compared using the mean square error (MSE) and mean absolute error (MAE). If the mean square error (MSE) and mean absolute error (MAE) values ​​of the updated model are both smaller than those of the unupdated model, the new model is used to guide the production of dry-type transformers.

8. A machine learning system for optimizing the manufacturing process of a dry-type transformer, which is used to implement the machine learning method for optimizing the manufacturing process of a dry-type transformer according to any one of claims 1 to 7, characterized in that: include: Data acquisition module, feature calculation module, production quality control module, quality inspection module and process traceability adjustment module; The data acquisition module is used to collect process parameters of the dry-type transformer production process and obtain quality indicators of the dry-type transformer; The feature calculation module is used to calculate the correlation coefficient between the process parameters and the quality indicators, select the process parameters that are strongly correlated with the quality indicators as features, and sort the features using recursive feature elimination; The production quality control module uses the support vector regression (SVR) algorithm based on the collected process parameters to build a dry-type transformer production quality prediction model and integrates the bagging learning method into the production quality control model; The quality inspection module is used to perform quality inspection on the produced dry-type transformers, collect image data of the produced dry-type transformers, detect the image data through a machine vision model, and determine whether the dry-type transformers have quality defects; The process traceability adjustment module is used to trace the process of dry-type transformers with quality problems, adjust the correlation coefficients between process parameters and quality indicators, and adjust the parameters in the dry-type transformer production quality prediction model.

9. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the machine learning method for optimizing the dry-type transformer manufacturing process according to any one of claims 1 to 7 by calling the computer program stored in the memory.

10. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are run on a computer, the computer is caused to execute the machine learning method for optimizing the dry-type transformer manufacturing process according to any one of claims 1 to 7.

Images