Capacitor production monitoring method, device, equipment and medium

Abstract

The invention provides a capacitor production monitoring method, device and equipment and a medium method, and relates to the technical field of capacitor production monitoring. The method comprises steps S1 to S8. The method comprises the following steps: S1, acquiring a capacitor production parameter set; S2, monitoring whether characteristic parameters are normal or not according to a preset rule; and S3, regularly analyzing the characteristic parameters by using a boxplot and IQR in a fixed batch. And S4, if the characteristic parameters are abnormal, marking an abnormal production parameter set. S5, comparing the abnormal and normal production parameter sets, and finding out a first abnormal process parameter set through analysis such as a causal diagram, S6, determining a second abnormal process parameter set by means of a fault cause analysis model, and S7, extracting repeated process parameters of the abnormal and normal production parameter sets as a final abnormal process parameter set. And S8, inputting the abnormal production parameter set and the final abnormal process parameter set into the fault solving model to obtain an abnormal production parameter suggested value, so that accurate positioning and problem solving of the abnormal process parameters in capacitor production are realized, and the production quality is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of capacitor production monitoring, in particular to a capacitor production monitoring method, device, equipment and medium. BACKGROUND

[0002] In the capacitor production process, there are many parameters that affect product quality, including material selection, production process, equipment state, etc. The existing technology has obvious shortcomings in capacitor production quality control. When the product quality problem occurs, it is usually necessary to stop the machine for problem analysis, which not only consumes time, but also causes the production line to stop and causes significant economic losses. The inefficiency of stoppage analysis not only affects the execution of the production plan, but also may lead to a decrease in customer satisfaction. In addition, equipment downtime can also reduce the reliability and life of the equipment, further increasing maintenance costs.

[0003] Therefore, the existing technology needs to be improved in the aspect of capacitor production quality control to reduce downtime and improve production efficiency. SUMMARY

[0004] The present application provides a capacitor production monitoring method, device, equipment and medium to improve at least one of the above technical problems.

[0005] In a first aspect, the present application provides a capacitor production monitoring method, which comprises steps S1 to S8.

[0006] S1, obtaining a set of production parameters in the capacitor production process. The set of production parameters includes a set of process parameters that can be controlled in the production process and a set of characteristic parameters of the product itself.

[0007] S2, according to the characteristic parameters, based on the pre-set parameter monitoring rules, monitoring whether the characteristic parameters of the product are normal.

[0008] S3, according to the characteristic parameters, periodically and batch by batch through the box plot and IQR analysis, analyzing whether the characteristic parameters of the product are normal.

[0009] S4, when the characteristic parameters are not normal, obtaining the production parameter set of the product with abnormal characteristic parameters, and marking it as an abnormal production parameter set.

[0010] S5, according to the abnormal production parameter set and the normal production parameter set in the database, through the cause and effect diagram and / or multilayer mean and / or correlation analysis, analyzing the process parameters that cause the characteristic parameters to be abnormal, defined as a first abnormal process parameter set.

[0011] S6, inputting the abnormal production parameter set of the abnormal capacitor into the fault cause analysis model, obtaining the process parameters that cause the characteristic parameters to be abnormal, defined as a second abnormal process parameter set.

[0012] S7, obtaining repeated process parameters in the first abnormal process parameter set and the second abnormal process parameter set, and marking as a final abnormal process parameter set.

[0013] S8, inputting the abnormal production parameter set and the final abnormal process parameter set into a fault solving model to obtain a recommended value of the abnormal production parameter.

[0014] In an optional embodiment, the process parameters of the process parameter set include: material number, production equipment, tank wet detection, wet detection voltage, wet detection capacity, tester number, energizing voltage, energizing process voltage, process current, energizing temperature, pure water conductivity, tank liquid room temperature conductivity, voltage, capacity, production team information, and production personnel information.

[0015] In an optional embodiment, the characteristic parameters of the characteristic parameter set include: leakage current.

[0016] In an optional embodiment, the parameter monitoring rules include monitoring reference value, rule one, rule two, rule three, rule four, and rule five.

[0017] The monitoring reference value includes parameter benchmark, parameter specification upper limit, parameter specification lower limit, parameter control upper limit, and parameter control lower limit. The parameter control upper limit is the parameter benchmark plus 1.5 times the deviation threshold. The parameter control lower limit is the parameter benchmark minus 1.5 times the deviation threshold. The deviation threshold is calculated according to historical data and the parameter benchmark. The deviation threshold is calculated according to historical data and the parameter benchmark. The calculation model of the deviation threshold is: , wherein, is the number of characteristic parameter measurement values in the historical data, represents the characteristic parameter measurement value, is the parameter benchmark.

[0018] Rule one: if the characteristic parameter measurement value is greater than the parameter specification upper limit or less than the parameter specification lower limit, it is considered abnormal.

[0019] Rule two: if five or more consecutive characteristic parameter measurement values are greater than or less than the parameter benchmark, it is determined that the characteristic parameter is abnormal from the fifth point.

[0020] Rule three: if five or more consecutive characteristic parameter measurement values are continuously increasing or continuously decreasing, it is determined that the group of characteristic parameter measurement values that are continuously increasing or continuously decreasing is abnormal.

[0021] Rule four, if the six consecutive characteristic parameter measurement values are alternately greater than and less than the parameter reference value, and the value greater than the parameter reference value exceeds the upper limit of the parameter control, and the value less than the parameter reference value is less than the lower limit of the parameter control, then the set of characteristic parameter measurement values is determined to be abnormal.

[0022] Rule five, the plurality of consecutive characteristic parameter measurement values appear periodically.

[0023] In an optional embodiment, the fault cause analysis model is an N-tag independent classification model. The network structure of the fault cause analysis model comprises an input layer, a first hidden layer, a first random inactivation layer, a second hidden layer, a second random inactivation layer, an output layer, and a conditional random field.

[0024] The input layer is provided with N neurons. N is the number of production parameters, and each production parameter corresponds to a neuron.

[0025] The first hidden layer is provided with 4N neurons. The first hidden layer uses a ReLU activation function.

[0026] The inactivation rate of the first random inactivation layer is 0.2.

[0027] The second hidden layer is provided with 4N neurons, and the second hidden layer uses a ReLU activation function.

[0028] The inactivation rate of the second random inactivation layer is 0.2.

[0029] The output layer is provided with N neurons and N linear activation functions, which are respectively used to output the non-normalized predicted values of the N production parameters. Each neuron corresponds to a production parameter.

[0030] The conditional random field is provided with an NxN transition matrix to represent the transition probability between any two tags. The conditional random field uses a Viterbi algorithm for decoding.

[0031] The first random inactivation layer and the second random inactivation layer are deleted after the model is trained. The input layer, the first hidden layer, the second hidden layer, and the output layer adopt a full connection structure, and each neuron in each layer is connected to all neurons of the previous layer.

[0032] In an optional embodiment, the loss function of the fault cause analysis model is :

[0033] .

[0034] wherein, is the set of all parameters to be optimized, is the total number of samples, is a maximum joint probability calculated by a Viterbi algorithm, denotes a probability, is a target label, denotes the th feature parameter measurement.

[0035] In an optional embodiment, the fault cause analysis model comprises steps A1 to A6 during training.

[0036] A1, a set of production parameters in a product production process are collected. These data include product state labels of qualified or unqualified products, and production parameter labels causing product unqualification.

[0037] A2, the collected data are preprocessed to obtain first training data. The preprocessing includes data cleaning to remove error records or outliers, and normalization processing.

[0038] A3, a network model of the fault cause analysis model is constructed according to the network structure of the fault cause analysis model.

[0039] A4, the fault cause analysis model is randomly initialized. The conditional random field needs to initialize the transition matrix.

[0040] A5, an optimization algorithm and a loss function are set. The optimization algorithm uses adaptive matrix estimation.

[0041] A6, the fault cause analysis model is trained according to the first training data, the optimization algorithm and the loss function , and the model performance is evaluated using k-fold cross-validation technology to obtain the trained fault cause analysis model.

[0042] In an optional embodiment, step S6 specifically comprises steps S61 to S62.

[0043] S61, the maximum and minimum values of each production parameter are obtained, and each abnormal production parameter of the abnormal capacitor is normalized according to the maximum and minimum values to obtain fault input data of the fault cause analysis model.

[0044] S62, the fault input data are input into the fault cause analysis model to obtain process parameters causing abnormal characteristic parameters, which are defined as a second set of abnormal process parameters.

[0045] In an optional embodiment, the network structure of the fault solving model comprises an input layer, a mask input layer, an encoder layer, a decoder layer, a second mask application layer and an output layer.

[0046] The input layer is provided with N neurons corresponding to N production parameters. The input layer uses a linear activation function.

[0047] The mask input layer is provided with N neurons to receive a mask vector corresponding to the input data. Each element is either 0 or 1, where 0 indicates that the corresponding production parameter is marked as abnormal, and 1 indicates normal.

[0048] The encoder layer includes connected in sequence The first mask application layer, the first hidden layer, the second hidden layer, and the bottleneck layer. The first mask application layer is used to multiply the data of the input layer with the mask vector, ensuring that all parameters marked as abnormal (i.e., positions with 0 in the mask) are set to zero, while normal parameters remain unchanged. The first hidden layer is provided with 128 neurons and uses a ReLU activation function. The second hidden layer is provided with 64 neurons and uses a ReLU activation function. The bottleneck layer is provided with 32 neurons and uses a ReLU activation function.

[0049] The decoder layer includes the third hidden layer and the fourth hidden layer. The third hidden layer is provided with 64 neurons and uses a Leaky ReLU activation function. The fourth hidden layer is provided with 128 neurons and uses a Leaky ReLU activation function.

[0050] The second mask application layer is used to apply the mask vector, ensuring that only those parameters marked as abnormal are updated by the model, while other parameters remain unchanged.

[0051] The output layer is provided with N neurons for reconstructing the input data, using a linear activation function to output the recommended values of the production parameters that maintain the original scale.

[0052] In an optional embodiment, the loss function of the fault resolution model is:

[0053]

[0054] where N is the number of production parameters, represents the th feature parameter measurement, is the corresponding mask vector, is the corresponding value of reconstructed or estimated by the fault resolution model.

[0055] In an optional embodiment, the fault resolution model training includes the following steps B1 to B5.

[0056] B1, obtain historical data during normal production to train the fault resolution model to learn the correct parameter pattern.

[0057] B2, pre-processing the collected data to obtain second training data. Wherein, the pre-processing includes standardization or normalization processing to scale all features to the same range.

[0058] B3, creating a mask vector, constructing an N-dimensional mask vector for each input sample. Wherein, the element is 0, indicating that the corresponding production parameter is marked as abnormal, and 1 indicating normal. In the training stage, all mask vectors are set to all 1, because we want the model to learn to reconstruct all input features.

[0059] B4, constructing a fault resolution model according to the network structure of the fault resolution model.

[0060] B5, training the fault resolution model according to the second training data and the loss function , using Adam optimizer to train the fault resolution model, and obtaining the trained fault resolution model.

[0061] Secondly, the application provides a capacitor production monitoring device, which comprises a data recording module, a data monitoring module, a data analysis module, an abnormal data acquisition module, a first correlation analysis module, a second correlation analysis module, a comparison module and a suggestion module.

[0062] The data recording module is used to obtain a production parameter set in the capacitor production process. Wherein, the production parameter set comprises a process parameter set that can be controlled in the production process and a characteristic parameter set of the product itself.

[0063] The data monitoring module is used to monitor whether the characteristic parameters of the product are normal based on the pre-set parameter monitoring rules according to the characteristic parameters.

[0064] The data analysis module is used to analyze whether the characteristic parameters of the product are normal through box plot and IQR analysis on a regular basis.

[0065] The abnormal data acquisition module is used to obtain the production parameter set of the product with abnormal characteristic parameters when the characteristic parameters are abnormal, and mark it as an abnormal production parameter set.

[0066] The first correlation analysis module is used to analyze the process parameters that cause the characteristic parameters to be abnormal through causal diagram and / or multilayer mean and / or correlation analysis according to the abnormal production parameter set and the normal production parameter set in the database, and define it as a first abnormal process parameter set.

[0067] The second correlation analysis module is used to input the abnormal production parameter set of the abnormal capacitor into a fault cause analysis model to obtain the process parameters that cause the characteristic parameters to be abnormal, and define it as a second abnormal process parameter set.

[0068] A comparison module is configured to obtain repeated process parameters in the first set of abnormal process parameters and the second set of abnormal process parameters, and mark the repeated process parameters as a final set of abnormal process parameters.

[0069] A suggestion module is configured to input the set of abnormal production parameters and the final set of abnormal process parameters into a fault resolution model, and obtain suggested values of the abnormal production parameters.

[0070] In a third aspect, the present application provides a capacitor production monitoring device, which comprises a processor, a memory, and a computer program stored in the memory. The computer program can be executed by the processor to implement the capacitor production monitoring method according to any one of the first aspect.

[0071] In a fourth aspect, the present application provides a computer-readable storage medium comprising a stored computer program, wherein the computer-readable storage medium controls the device where the computer-readable storage medium is located to execute the capacitor production monitoring method according to any one of the first aspect when the computer program is running.

[0072] By adopting the above technical solutions, the present application can achieve the following technical effects:

[0073] The capacitor production monitoring method of the present application can accurately determine the process parameters that cause capacitor defects through the fault cause analysis model, significantly improving the accuracy of problem diagnosis. This method can quickly identify problem parameters, thereby greatly reducing downtime analysis time and improving the operation efficiency of the production line. By accurately and quickly finding the source of the problem and providing a solution, the present application effectively reduces economic losses caused by downtime.

[0074] In addition, the fault resolution model can learn and provide optimization suggestions for process parameters based on existing data, which helps to improve product quality and production efficiency. By continuously monitoring and optimizing process parameters, the present application enhances the stability and reliability of production, reduces future possible quality problems. This not only improves customer satisfaction with the product, but also reduces production costs and improves production efficiency through intelligent means. BRIEF DESCRIPTION OF DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the specific embodiments of the present application. It should be understood that the following drawings only show some specific embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0076] Fig. 1 is a flowchart of the capacitor production monitoring method.

[0077] Fig. 2 is a schematic diagram of parameter monitoring.

[0078] Fig. 3 is a schematic diagram of box plot analysis of fault causes

[0079] Fig. 4 is a network structure diagram of a fault cause analysis model.

[0080] Fig. 5 is a network structure diagram of a fault solution model. DETAILED DESCRIPTION

[0081] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0082] Embodiment one, please refer to Figs. 1 to 5 The first embodiment of the present application provides a capacitor production monitoring method, which can be executed by a capacitor production monitoring device (hereinafter referred to as: production monitoring device). In particular, it is executed by one or more processors in the production monitoring device to realize steps S1 to S8.

[0083] S1, a set of production parameters in the capacitor production process is obtained. The set of production parameters includes a set of process parameters that can be controlled in the production process and a set of characteristic parameters of the product itself.

[0084] Preferably, the process parameters of the set of process parameters include: material number, production equipment, block wet detection, wet detection voltage, wet detection capacity, tester number, energizing voltage, energizing process voltage, process current, energizing temperature, pure water conductivity, liquid chamber temperature conductivity, voltage, capacity, production team information, and production personnel information. The characteristic parameters of the set of characteristic parameters include: leakage current.

[0085] It can be understood that the production monitoring device can be a portable notebook computer, a desktop computer, a server, an industrial computer, or other electronic devices with computing performance.

[0086] S2, according to the characteristic parameters, based on the pre-set parameter monitoring rules, whether the characteristic parameters of the product are normal is monitored. Preferably, as shown in Fig. 2 The parameter monitoring rules include monitoring reference values, rule one, rule two, rule three, rule four, and rule five.

[0087] Monitoring reference values ​​include parameter baseline, upper limit of parameter specification, lower limit of parameter specification, upper limit of parameter control, and lower limit of parameter control. The upper limit of parameter control is the parameter baseline plus 1.5 times the deviation threshold. The lower limit of parameter control is the parameter baseline minus 1.5 times the deviation threshold. The deviation thresholds are calculated based on historical data and the parameter baseline. The calculation model is as follows: In the formula, The number of characteristic parameter measurements in historical data, Indicates the first Each characteristic parameter measurement value, Used as the parameter reference.

[0088] Rule 1: If the measured value of a characteristic parameter is greater than the upper limit of the parameter specification or less than the lower limit of the parameter specification, it is considered abnormal. Specifically, if the measured value of a characteristic parameter exceeds the parameter specification range, the product is considered defective; this is a 100% defective rule.

[0089] Rule 2: If five or more consecutive characteristic parameter measurements are greater than or less than the parameter benchmark, then the characteristic parameter is considered abnormal starting from the fifth point.

[0090] Rule 3: If five or more consecutive characteristic parameter measurements increase or decrease continuously, then the set of characteristic parameter measurements that continuously increase or decrease is considered abnormal.

[0091] Rule 4: If six consecutive characteristic parameter measurements alternately exceed and fall below the parameter baseline, and the value exceeding the parameter baseline exceeds the upper limit of parameter control, and the value falling below the parameter baseline falls below the lower limit of parameter control, then the set of characteristic parameter measurements is considered abnormal.

[0092] Rule 5: Multiple consecutive characteristic parameter measurements appear periodically.

[0093] Specifically, Rules 2 through 5 indicate that the monitored product parameters are all within acceptable limits, but there are potential safety hazards that could lead to product defects; these are preventative rules. Rule 2 is usually related to systemic deviations, which may be due to incorrect operating procedures or equipment malfunctions. It helps identify potential systemic problems and suggests the need for process adjustments or calibrations. Rule 3 may be due to tool wear, reduced maintenance levels, or improved operator skills. Its benefit lies in detecting trend changes in the process, indicating the need for inspection and improvement of tools or operating methods. Rules 4 and 5 indicate that there are periodic changes in the system, and these changes may become larger, leading to product defects.

[0094] S3. Based on the characteristic parameters, analyze the product's characteristic parameters regularly and in fixed batches using box plots and IQR to determine if they are normal.

[0095] Specifically, such asFig. 4 As shown, a box plot is a statistical graph used to display the distribution of data by defining the normal range through quartiles. The drawing of a box plot relies on five key numbers: the minimum value (min), the lower quartile (Q1), the median (M), the upper quartile (Q3), and the maximum value (max). These values are collectively referred to as the "five-number summary". The IQR (interquartile range) is the difference between the upper quartile (Q3) and the lower quartile (Q1). Generally, data points below Q1-1.5IQR and above Q3+1.5IQR are considered outliers.

[0096] When analyzing characteristic parameters, first calculate the Q1 and Q3 of the data set, then calculate the IQR. Next, determine the lower limit (Q1-1.5IQR) and the upper limit (Q3+1.5IQR). Any characteristic parameter measurement below the lower limit or above the upper limit is considered abnormal. This method can intuitively identify outliers in the data set and determine the degree of data dispersion and bias in the data set. In this way, it can effectively analyze whether the characteristic parameters are normal, thereby helping to identify potential problems in the production process.

[0097] S4, when the characteristic parameter is not normal, obtain the production parameter set of the product with abnormal characteristic parameter, and mark it as an abnormal production parameter set.

[0098] S5, according to the abnormal production parameter set and the normal production parameter set in the database, analyze the process parameters that cause the characteristic parameter to be abnormal through a cause-and-effect diagram and / or multilevel mean and / or correlation analysis, and define it as a first abnormal process parameter set. Only one or more of these analysis methods can be used, and the present application does not make specific limitations.

[0099] A cause-and-effect diagram (also known as a fishbone diagram or Ishikawa diagram) is used to visually represent how various factors interact and lead to a particular problem or effect. In process parameter analysis, a cause-and-effect diagram can help identify and organize various potential causes that may lead to abnormal characteristic parameters. By drawing a cause-and-effect diagram, the relationship between each process parameter and the characteristic parameter can be systematically analyzed to identify key process parameters that may cause abnormalities.

[0100] Multilevel mean analysis is a statistical method used to analyze the hierarchical structure and mean changes in data. This method can help identify trends and patterns in data to determine which changes in process parameters may cause abnormalities in characteristic parameters. Through multilevel mean analysis, data can be stratified to identify mean changes at different levels, and then determine which changes in process parameters have the greatest impact on characteristic parameters.

[0101] Combined with the cause-effect diagram and the multi-layer mean analysis, the process parameters causing the abnormal characteristic parameters can be effectively identified and analyzed. The cause-effect diagram provides an intuitive view of the cause-effect relationship, and the multi-layer mean analysis provides data-driven statistical support, and the combination of the two can more comprehensively understand and solve the process parameter abnormality problems.

[0102] In addition, it can also be assisted to judge which production parameters are most likely to cause the capacitor leakage current to be unqualified by calculating the Pearson correlation coefficient, the Spearman's Rank Correlation Coefficient, the Kendall's Tau Coefficient, the Partial Correlation Coefficient, the Point-Biserial Correlation Coefficient and the Biserial Correlation Coefficient.

[0103] S6, input the abnormal production parameter set of the abnormal capacitor into the fault cause analysis model to obtain the process parameters causing the abnormal characteristic parameters, and define the process parameters as a second abnormal process parameter set. Preferably, step S6 specifically includes steps S61 to S62.

[0104] S61, obtain the maximum value and the minimum value of each production parameter, and normalize each abnormal production parameter of the abnormal capacitor according to the maximum value and the minimum value to obtain fault input data of the fault cause analysis model.

[0105] S62, input the fault input data into the fault cause analysis model to obtain the process parameters causing the abnormal characteristic parameters, and define the process parameters as a second abnormal process parameter set.

[0106] S7, obtain the repeated process parameters in the first abnormal process parameter set and the second abnormal process parameter set, and mark the process parameters as a final abnormal process parameter set.

[0107] S8, input the abnormal production parameter set and the final abnormal process parameter set into a fault solving model to obtain a suggested value of the abnormal production parameter. Specifically, the abnormal production parameter set also needs to be normalized before being input into the fault solving model.

[0108] The capacitor production monitoring method of the present application accurately determines the process parameters that cause capacitor defects through a fault cause analysis model, significantly improving the accuracy of problem diagnosis. This method can quickly identify problem parameters, thereby significantly reducing downtime analysis time and improving production line efficiency. By accurately and quickly finding the source of the problem and providing a solution, the solution effectively reduces economic losses caused by downtime.

[0109] In addition, the fault resolution model can learn and provide optimization suggestions for process parameters based on existing data, helping to improve product quality and production efficiency. By continuously monitoring and optimizing process parameters, the solution enhances the stability and reliability of production, reducing potential future quality problems. This not only improves customer satisfaction with the product, but also reduces production costs and improves production efficiency through intelligent means.

[0110] Based on the above embodiments, an optional embodiment of the present application is as shown in Fig. 4 The fault cause analysis model is an N-label independent classification model. The network structure of the fault cause analysis model includes an input layer, a first hidden layer, a first random inactivation layer, a second hidden layer, a second random inactivation layer, an output layer, and a conditional random field.

[0111] The input layer is provided with N neurons. Wherein, N is the number of production parameters, and each process parameter corresponds to a neuron.

[0112] The first hidden layer is provided with 4N neurons. The first hidden layer uses a ReLU activation function.

[0113] The inactivation rate of the first random inactivation layer is 0.2.

[0114] The second hidden layer is provided with 4N neurons, and the second hidden layer uses a ReLU activation function.

[0115] The inactivation rate of the second random inactivation layer is 0.2.

[0116] The output layer is provided with N neurons and N linear activation functions, respectively, to output the non-normalized predicted values of N production parameters. Wherein, each neuron corresponds to a production parameter

[0117] The conditional random field is provided with an NxN transition matrix to represent the transition probability between any two labels. Wherein, the conditional random field uses the Viterbi algorithm for decoding.

[0118] The first random inactivation layer and the second random inactivation layer are deleted after model training.

[0119] In the embodiment, the production parameters include the following 16 parameters: material number, production equipment, tank wet detection, wet detection voltage, wet detection capacity, tester number, energizing voltage, energizing process voltage, process current, energizing temperature, pure water conductivity, tank liquid room temperature conductivity, voltage, capacity, production team information, production personnel information, and leakage current. Therefore, N is 16. Each hidden layer has 64 neurons. The fault cause analysis model is constructed as a 16-label independent classification model.

[0120] The existing multi-label independent classification model has the following disadvantages in the existing multi-layer perceptron: 1. Without a large enough data set or a proper regularization method, overfitting is easy to occur, which leads to good performance of the model on the training set but poor generalization ability on the test set. 2. When processing high-dimensional input data, the computational complexity is high, and it is difficult to effectively learn the potential features in the data. 3. Label correlation is difficult to capture: in a multi-label classification task, there may be correlation or dependency between labels. The basic independent classification model is difficult to directly capture the complex relationship between these labels and cannot make accurate judgments.

[0121] Therefore, the embodiment of the present application adds a Dropout layer between the hidden layer and the output layer, thereby effectively reducing overfitting.

[0122] In addition, in general, the abnormality of the characteristic parameters of the product is not caused by one parameter, but by the joint action of several production parameters. Therefore, in the embodiment, a conditional random field layer is added after the output layer, so that the model considers the mutual influence between all labels. Further, it is more accurate to determine which production parameters jointly cause the product defect. Specifically, each neuron of the output layer is responsible for calculating the emission score (logits) of the corresponding label, and then inputting the conditional random field layer after linear activation function processing. The emission score reflects the possibility of each label under a given input.

[0123] Since determining whether the production parameter is the cause of the product defect is a multi-label classification problem, the embodiment of the present application designs a special CRF layer, called "label pair CRF". This CRF layer will construct an NxN transition matrix to represent the transition probability between any two labels. The elements in the transition matrix represent the probability or preference of a label transitioning to another label. For each sample, the CRF layer will calculate the most likely label combination according to the emission score output by the independent classification model and the transition matrix.

[0124] In industrial production and manufacturing environments, real-time monitoring and analysis of multiple parameters during production processes are crucial for ensuring product quality and improving production efficiency. Suppose we have 16 key production parameters, each of which can be one of the causes of abnormalities in the production process. The goal of this invention is to develop a model based on independent classification models and conditional random fields that can independently evaluate the 16 parameters and mark which parameters may be the root cause of abnormalities.

[0125] Below, a specific scheme for using a fault cause analysis model to independently determine whether 16 production parameters are abnormal cause parameters is listed.

[0126] Data preprocessing and feature selection engineering. Data collection: First, a large amount of historical data needs to be collected from various sensors and other monitoring devices on the production line. These data should include data during normal operation and data when known abnormalities occur. Feature selection: According to the knowledge of domain experts or through automated feature selection methods, select the features that best reflect the changes in production status. For example, production device number, enabling voltage, etc. Standardization / normalization: In order to ensure that the data input into the neural network has the same scale, standardization or normalization processing is usually performed on all features.

[0127] The 16 production parameters at each time point form the input vector The input layer of the fault cause analysis model is input, and the output layer of the fault cause analysis model outputs a 16-dimensional vector, corresponding to the possibility score (logits) of each parameter as an abnormal cause. These possibility scores are input as emission scores for each label into the conditional random field layer.

[0128] In practical applications, there may be some correlation between different production parameters. Therefore, the conditional random field is used to help capture this potential correlation, and then output the most likely abnormal cause combination.

[0129] Preferably, the loss function of the fault cause analysis model is:

[0130] .

[0131] where, is the set of all parameters to be optimized, is the total number of samples, is the maximum joint probability calculated by the Viterbi algorithm, represents the probability, is the target label, represents the th feature parameter measurement value.

[0132] Specifically, The weights of the input layer, the first hidden layer, the second hidden layer, and the output layer, and the elements of the transition matrix of the conditional random field are included. The loss function Can well improve the prediction accuracy of the multi-label independent classification model and simplify the gradient calculation. By accurately measuring the difference between the predicted probability distribution and the real distribution, the loss function Can guide the model to adjust the prediction result more finely and approximate the real label. In addition, the simplicity of the derivative calculation helps to speed up the model training, so that the model can converge faster during the training process. In addition, the loss function Enhances the robustness of the model to outliers and has wide applicability and flexibility. It can promote the model to focus on learning samples with large deviations by generating larger loss values, thereby improving the stability of the model.

[0133] On the basis of the above embodiment, in an optional embodiment of the present application, the fault cause analysis model training comprises steps A1 to A6.

[0134] A1, a large number of production parameter sets in the production process are collected. These data include product state labels of qualified or unqualified products, and production parameter labels that cause product unqualified.

[0135] A2, pre-process the collected data to obtain first training data. The pre-processing includes data cleaning, removing error records or outliers, and normalization processing.

[0136] A3, according to the network structure of the fault cause analysis model, a network model of the fault cause analysis model is constructed.

[0137] A4, randomly initialize the fault cause analysis model. The conditional random field needs to initialize the transition matrix.

[0138] A5, set the optimization algorithm and the loss function. The optimization algorithm uses adaptive matrix estimation.

[0139] A6, according to the first training data, the optimization algorithm and the loss function Train the fault cause analysis model, and use the k-fold cross-validation technique to evaluate the model performance to obtain the trained fault cause analysis model.

[0140] The improved fault cause analysis model is improved based on a basic multi-label independent classification model. The model learns the complex nonlinear relationship between features through multiple hidden layers and nonlinear activation functions, which enhances the expression ability of the model and enables the model to capture subtle patterns in the data. Moreover, the CRF model optimizes the sequence prediction result by considering the dependency between adjacent elements in the sequence, improves the robustness of the prediction, and enables the model to more accurately process the dependency between labels in the multi-label independent classification task, thereby improving the overall classification performance.

[0141] Based on the above-mentioned embodiments, in an optional embodiment of the present application, as shown in Fig. 5 The network structure of the fault resolution model includes an input layer, a mask input layer, an encoder layer, a decoder layer, a second mask application layer, and an output layer.

[0142] The input layer corresponds to N production parameters and is provided with N neurons. The input layer uses a linear activation function.

[0143] The mask input layer is provided with N neurons to receive a mask vector corresponding to the input data. Each element is 0 or 1, where 0 indicates that the corresponding production parameter is marked as abnormal, and 1 indicates normal.

[0144] The encoder layer includes connected in sequence The first mask application layer, the first hidden layer, the second hidden layer, and the bottleneck layer. The first mask application layer is used to multiply the data of the input layer with the mask vector, ensuring that all parameters marked as abnormal (i.e., positions with 0 in the mask) are set to zero, while normal parameters remain unchanged. The first hidden layer is provided with 128 neurons and uses a ReLU activation function. The second hidden layer is provided with 64 neurons and uses a ReLU activation function. The bottleneck layer is provided with 32 neurons and uses a ReLU activation function.

[0145] The decoder layer includes a third hidden layer and a fourth hidden layer. The third hidden layer is provided with 64 neurons and uses a Leaky ReLU activation function. The fourth hidden layer is provided with 128 neurons and uses a Leaky ReLU activation function.

[0146] The second mask application layer is used to apply the mask vector, ensuring that only those parameters marked as abnormal are updated by the model, and other parameters remain unchanged.

[0147] The output layer is provided with N neurons for reconstructing the input data, and uses a linear activation function to output the recommended value of the production parameter that maintains the original scale.

[0148] Specifically, the encoder compresses the input data (multiple production parameters) into a low-dimensional representation, called latent representation. The decoder then reconstructs the original data from this latent representation. During the training process, the autoencoder learns how to reconstruct the input data with the smallest error. When an abnormal production parameter occurs, the data set containing the abnormal parameter is input into the trained autoencoder, which attempts to reconstruct the abnormal parameter from the representations of other normal parameters in the latent space, thereby obtaining an estimate of the normal value of the abnormal parameter.

[0149] Example: Suppose we have 5 production parameters, namely energizing temperature, energizing voltage, process current, production equipment number and energizing process voltage. When the energizing voltage is abnormal, the other 4 normal parameters (temperature, humidity, flow rate, pH) containing the abnormal pressure value are input into the trained autoencoder together with the abnormal pressure value. The autoencoder finds the most matching representation in the latent space according to the normal data pattern learned in the past, and then outputs the estimated values of all parameters including the pressure parameter through the decoder, where the estimated value of the pressure parameter is a guess of its normal value.

[0150] Preferably, the loss function of the fault resolution model is :

[0151]

[0152] where N is the number of production parameters, represents the measurement value of the th feature parameter, is the corresponding mask vector, is the corresponding value of reconstructed or estimated by the fault resolution model.

[0153] Specifically, the loss function can help the autoencoder learn to reconstruct the input data, even if some input values are missing or damaged, to restore the original data as much as possible.

[0154] On the basis of the above embodiment, in an optional embodiment of the present application, the fault resolution model training comprises the following steps B1 to step B5.

[0155] B1, obtain historical data during normal production to train the fault resolution model to learn the correct parameter pattern.

[0156] B2, pre-process the collected data to obtain the second training data. The pre-processing includes standardization or normalization processing to scale all features to the same range.

[0157] B3, a mask vector is created, and an N-dimensional mask vector is constructed for each input sample. Wherein, the element is 0, indicating that the corresponding production parameter is marked as abnormal, and 1 indicating normal. In the training phase, all mask vectors are set to all 1, because we want the model to learn to reconstruct all input features.

[0158] B4, constructing a fault solving model according to the network structure of the fault solving model.

[0159] B5, training the fault solving model according to the second training data and the loss function , using the Adam optimizer to train the fault solving model, and obtaining the trained fault solving model.

[0160] Specifically, by obtaining historical data during normal production and preprocessing, the fault solving model can learn the correct parameter pattern and accelerate convergence. Specifically, step B1 ensures that the model is trained based on data under normal operating conditions, so as to accurately capture the correlation between normal production parameters. And step B2 processes the data by standardization or normalization, so that all features are in the same scale range, which not only speeds up the training, but also improves the stability and generalization ability of the model.

[0161] Creating a mask vector and setting it to all 1 in the training phase, combined with constructing a model according to the network structure and training using the Adam optimizer, ensures that the model can learn the reconstruction of input features comprehensively. Steps B3 to B5 work together to enable the model to focus on learning all input features during training, while using an efficient optimization algorithm to improve training efficiency, and finally obtain a well-trained, high-performance fault solving model.

[0162] Embodiment two, the application provides a capacitor production monitoring device, which comprises a data recording module, a data monitoring module, a data analysis module, an abnormal data acquisition module, a first correlation analysis module, a second correlation analysis module, a comparison module and a suggestion module.

[0163] The data recording module is used to obtain a set of production parameters in the capacitor production process. The set of production parameters includes a set of process parameters that can be controlled in the production process and a set of characteristic parameters of the product itself.

[0164] The data monitoring module is used to monitor whether the characteristic parameters of the product are normal based on the pre-set parameter monitoring rules according to the characteristic parameters.

[0165] The data analysis module is used to analyze whether the characteristic parameters of the product are normal by box plot and IQR analysis regularly and in batches according to the characteristic parameters.

[0166] An abnormal data acquisition module is configured to acquire a production parameter set of a product with an abnormal characteristic parameter when the characteristic parameter is abnormal, and mark the production parameter set as an abnormal production parameter set.

[0167] A first correlation analysis module is configured to analyze process parameters causing the characteristic parameter to be abnormal by a cause-effect diagram and / or a multi-layer mean and / or a correlation analysis according to the abnormal production parameter set and a normal production parameter set in a database, and define the process parameters as a first abnormal process parameter set.

[0168] A second correlation analysis module is configured to input the abnormal production parameter set of the abnormal capacitor into a fault cause analysis model, acquire process parameters causing the characteristic parameter to be abnormal, and define the process parameters as a second abnormal process parameter set.

[0169] A comparison module is configured to acquire repeated process parameters in the first abnormal process parameter set and the second abnormal process parameter set, and mark the process parameters as a final abnormal process parameter set.

[0170] A suggestion module is configured to input the abnormal production parameter set and the final abnormal process parameter set into a fault solution model, and acquire a suggested value of the abnormal production parameter.

[0171] In a third aspect, the present application provides a capacitor production monitoring device, which comprises a processor, a memory, and a computer program stored in the memory. The computer program can be executed by the processor to implement the capacitor production monitoring method according to any one of the first aspect.

[0172] In a fourth aspect, the present application provides a computer readable storage medium, which comprises a stored computer program. When the computer program is running, the computer readable storage medium controls a device where the computer readable storage medium is located to execute the capacitor production monitoring method according to any one of the first aspect.

[0173] In several embodiments provided by the embodiments of the present application, it should be understood that the disclosed apparatus and method can also be implemented by other manners. The apparatus and method embodiments described above are only illustrative, for example, the flowcharts and block diagrams in the drawings show the possible implementation architecture, function and operation of the apparatus, method and computer program product according to the embodiments of the present application. In this regard, each block in the flowchart or block diagram can represent a module, a program segment or a part of code, which includes one or more executable instructions for implementing the specified logic function. It should also be noted that in some alternative implementation manners, the functions noted in the blocks can also occur in different order from that noted in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes they can also be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for executing the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0174] In addition, each functional module in the embodiments of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0175] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part of the prior art that makes a contribution or the part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, an electronic device, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes. It should be noted that in this document, the terms “include”, “contain” or any other variant thereof are intended to cover non-exclusive inclusion, so that the processes, methods, articles or devices that include a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement “including a” does not exclude the presence of another identical element in the process, method, article or device that includes the element.

[0176] The terminology used in the description of the embodiments herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the description of the embodiments and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0177] It should be understood that the term "and / or" as used herein merely describes associated objects, and can exist in three forms, for example, A and / or B can mean that A exists alone, A and B exist together, or B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects.

[0178] Depending on the context, the word "if" as used herein can be interpreted to mean "when" or "upon" or "in response to determining" or "in response to detecting." Similarly, the phrase "if it is determined" or "if [a stated condition or event] is detected" can be interpreted to mean "upon determining" or "in response to determining" or "upon detecting [the stated condition or event]" or "in response to detecting [the stated condition or event]."

[0179] The "first / second" mentioned in the embodiments are only to distinguish similar objects, and do not represent a specific order of the objects. Understandably, the "first / second" can be interchanged in a specific order or sequence as allowed. It should be understood that the objects distinguished by "first / second" can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0180] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method of monitoring production of capacitors, characterized by, The method comprises the following steps: acquiring a production parameter set in a production process of a capacitor; wherein the production parameter set comprises a process parameter set that can be controlled in a production process and a characteristic parameter set of the product itself; monitoring whether the characteristic parameter is normal based on a pre-set parameter monitoring rule according to the characteristic parameter; periodically and batch-wise analyzing whether the characteristic parameter of the product is normal through a box plot and IQR analysis according to the characteristic parameter; when the characteristic parameter is not normal, acquiring the production parameter set of the product with the abnormal characteristic parameter and marking it as an abnormal production parameter set; analyzing the process parameter that causes the characteristic parameter to be abnormal and defining it as a first abnormal process parameter set through a cause-effect diagram and / or multilayer mean and / or correlation analysis according to the abnormal production parameter set and a normal production parameter set in a database; inputting the abnormal production parameter set of the abnormal capacitor into a fault cause analysis model to acquire the process parameter that causes the characteristic parameter to be abnormal and defining it as a second abnormal process parameter set; acquiring the repeated process parameter in the first abnormal process parameter set and the second abnormal process parameter set and marking it as a final abnormal process parameter set; inputting the abnormal production parameter set and the final abnormal process parameter set into a fault solving model to acquire the recommended value of the abnormal production parameter.

2. The method of claim 1, wherein The fault cause analysis model is an N-label independent classification model; the network structure of the fault cause analysis model comprises: an input layer provided with N neurons; wherein N is the number of production parameters, and each production parameter corresponds to a neuron; a first hidden layer provided with 4N neurons; the first hidden layer uses a ReLU activation function; a first random inactivation layer, the inactivation rate of the first random inactivation layer is 0.2; a second hidden layer provided with 4N neurons, the second hidden layer uses a ReLU activation function; a second random inactivation layer, the inactivation rate of the second random inactivation layer is 0.2; an output layer provided with N neurons and N linear activation functions, respectively, to output the non-normalized predicted value of N production parameters; wherein each neuron corresponds to a production parameter; a conditional random field provided with an NxN transition matrix to represent the transition probability between any two labels; wherein the conditional random field uses a Viterbi algorithm for decoding; wherein the first random inactivation layer and the second random inactivation layer are deleted after the model is trained; the input layer, the first hidden layer, the second hidden layer and the output layer adopt a full connection structure, and each neuron in each layer is connected with all the neurons of the previous layer.

3. The method of claim 2, wherein The loss function of the failure cause analysis model is: ； in, For the set of all parameters to be optimized, It is the total number of samples, The maximum joint probability is calculated using the Viterbi algorithm. Represents probability, For target tags, Indicates the first Measured values ​​of each characteristic parameter; The fault cause analysis model training comprises the following steps: collecting a large number of production parameter sets in a production process of a product; wherein the data includes a product state label of whether the product produced is qualified or unqualified and a production parameter label causing the product to be unqualified; preprocessing the collected data to acquire first training data; wherein the preprocessing includes data cleaning to remove error records or outliers and normalization processing; constructing a network model of the fault cause analysis model according to the network structure of the fault cause analysis model; Randomly initialize the fault cause analysis model; wherein, the conditional random field needs to initialize the transition matrix; Set the optimization algorithm and loss function; wherein, the optimization algorithm uses adaptive matrix estimation; According to the first training data, the optimization algorithm and the loss function The fault cause analysis model is trained, and the model performance is evaluated using the k-fold cross-validation technique to obtain the trained fault cause analysis model.

4. The method of claim 1, wherein Input the abnormal production parameter set of the abnormal capacitor into the fault cause analysis model to obtain the process parameters that cause the characteristic parameters to be abnormal, which are defined as the second abnormal process parameter set, specifically including: Obtain the maximum and minimum values of each production parameter, and normalize each abnormal production parameter of the abnormal capacitor according to the maximum and minimum values to obtain the fault input data of the fault cause analysis model; Input the fault input data into the fault cause analysis model to obtain the process parameters that cause the characteristic parameters to be abnormal, which are defined as the second abnormal process parameter set.

5. The method of claim 1, wherein The network structure of the fault solving model includes: The input layer is provided with N neurons corresponding to N production parameters, and uses a linear activation function; The mask input layer is provided with N neurons to receive a mask vector corresponding to the input data; wherein, each element is 0 or 1, wherein 0 indicates that the corresponding production parameter is marked as abnormal, and 1 indicates normal; Encoder layer, comprising The decoder layer includes a third hidden layer and a fourth hidden layer; wherein, the third hidden layer is provided with 64 neurons and uses a Leaky ReLU activation function; the fourth hidden layer is provided with 128 neurons and uses a Leaky ReLU activation function; a first mask application layer, a first hidden layer, a second hidden layer and a bottleneck layer; wherein the first mask application layer is used to multiply the data of the input layer with a mask vector, ensuring that only the parameters marked as abnormal are set to zero, and the normal parameters remain unchanged; the first hidden layer is provided with 128 neurons and uses a ReLU activation function; the second hidden layer is provided with 64 neurons and uses a ReLU activation function; the bottleneck layer is provided with 32 neurons and uses a ReLU activation function; The second mask application layer is used to apply the mask vector to ensure that only the parameters marked as abnormal are updated by the model, and other parameters remain unchanged; The output layer is provided with N neurons for reconstructing the input data, and uses a linear activation function to output the recommended values of the production parameters that remain in the original scale. The fault solving model training includes the following steps:

6. The method of claim 1, wherein The loss function of the failure resolution model is: Where N is the number of production parameters, Indicates the first Each characteristic parameter measurement value, for The corresponding mask vector, For fault resolution model reconstruction or estimation The corresponding value; Obtain historical data during normal production to train the fault solving model to learn the correct parameter pattern; Preprocess the collected data to obtain second training data; wherein, the preprocessing includes standardization or normalization processing to scale all features to the same range; Create a mask vector, an N-dimensional mask vector is constructed for each input sample; wherein, an element of 0 indicates that the corresponding production parameter is marked as abnormal, and 1 indicates normal; during the training phase, all mask vectors are set to all 1s because we want the model to learn to reconstruct all input features; Construct the fault solving model according to the network structure of the fault solving model; The parameter monitoring rules include monitoring reference values, rule one, rule two, rule three, rule four, and rule five; According to the second training data and a loss function , training the fault resolution model using an Adam optimizer to obtain a trained fault resolution model.

7. The method of claim 1, wherein Rule one, if the feature parameter measurement value is greater than the upper limit of the parameter specification or less than the lower limit of the parameter specification, it is considered abnormal; The monitoring reference value comprises a parameter benchmark, a parameter specification upper limit, a parameter specification lower limit, a parameter control upper limit and a parameter control lower limit. The parameter control upper limit is the parameter benchmark plus 1.5 times the deviation threshold. The parameter control lower limit is the parameter benchmark minus 1.5 times the deviation threshold. The deviation threshold is calculated according to historical data and the parameter benchmark. The deviation threshold is calculated according to historical data and the parameter benchmark The calculation model of the deviation threshold is: , wherein, is the number of characteristic parameter measurement values in the historical data, represents the characteristic parameter measurement value, is the parameter benchmark. Rule two, if five or more consecutive characteristic parameter measurement values are greater than or less than the parameter reference, it is determined that the characteristic parameters are abnormal from the fifth point; Rule three, if five or more consecutive characteristic parameter measurement values are continuously increasing or decreasing, it is determined that the group of characteristic parameter measurement values that are continuously increasing or decreasing are abnormal; ​ Rule four, if the six consecutive characteristic parameter measurement values are alternately greater than and less than the parameter reference value, and the value greater than the parameter reference value exceeds the upper limit of the parameter control, and the value less than the parameter reference value is less than the lower limit of the parameter control, then the set of characteristic parameter measurement values is determined to be abnormal; Rule five, if the multiple consecutive characteristic parameter measurement values appear periodically, then the set of characteristic parameter measurement values is determined to be abnormal; The process parameters of the process parameter set include: material number, production equipment, tank wet detection, wet detection voltage, wet detection capacity, tester number, energizing voltage, energizing process voltage, process current, energizing temperature, pure water conductivity, tank liquid room temperature conductivity, voltage, capacity, production team information, and production personnel information. The characteristic parameters of the characteristic parameter set include: leakage current.

8. A device for monitoring production of capacitors, characterized by Comprise: a data recording module configured to obtain a production parameter set in a capacitor production process; wherein the production parameter set comprises a process parameter set that can be controlled in the production process and a characteristic parameter set of the product itself; a data monitoring module configured to monitor whether the characteristic parameters of the product are normal based on a pre-set parameter monitoring rule according to the characteristic parameters; a data analysis module configured to analyze whether the characteristic parameters of the product are normal through box plot and IQR analysis on a regular basis and in batches according to the characteristic parameters; an abnormal data obtaining module configured to obtain a production parameter set of a product with abnormal characteristic parameters when the characteristic parameters are abnormal, and mark the production parameter set as an abnormal production parameter set; a first correlation analysis module configured to analyze process parameters that cause the characteristic parameters to be abnormal by a cause-effect diagram and / or multilayer mean and / or correlation analysis according to the abnormal production parameter set and a normal production parameter set in a database, and define the process parameters as a first abnormal process parameter set; a second correlation analysis module configured to input the abnormal production parameter set of the abnormal capacitor into a fault cause analysis model, obtain process parameters that cause the characteristic parameters to be abnormal, and define the process parameters as a second abnormal process parameter set; a comparison module configured to obtain repeated process parameters in the first abnormal process parameter set and the second abnormal process parameter set, and mark the process parameters as a final abnormal process parameter set; a suggestion module configured to input the abnormal production parameter set and the final abnormal process parameter set into a fault solving model, and obtain a suggested value of the abnormal production parameter set.

9. A capacitor production monitoring device, characterized in that, The computer readable storage medium comprises a stored computer program, wherein the computer readable storage medium controls the device where the computer readable storage medium is located to execute the capacitor production monitoring method according to any one of claims 1 to 7 when the computer program runs.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium comprises a stored computer program, wherein the computer readable storage medium controls the device where the computer readable storage medium is located to execute the capacitor production monitoring method according to any one of claims 1 to 7 when the computer program runs.

Images