Capacitor failure rate prediction method based on artificial intelligence and computer equipment

Abstract

The invention discloses an artificial intelligence-based capacitor failure rate prediction method and computer equipment, and the method comprises the steps: carrying out the statistics of the production process parameters of a capacitor and the data of the product failure rate, selecting a feature parameter as the input of a to-be-constructed neural network model, taking the target variable as the failure rate, and carrying out the calculation of the product failure rate; z-Score standardization is carried out on the characteristic parameters and the failure rate data; a neural network model is constructed, each layer uses a ReLU function as an activation function, an output layer uses a linear activation function for prediction, an input layer uses L1 regularization to prevent overfitting, an Adam optimizer is used, and a mean square error is used as a loss function; splitting the data into a training set and a test set, and evaluating the performance of the constructed neural network model by using R2 scores; and predicting the failure rate of the capacitor according to the constructed neural network model. According to the method, the required neural network model can be established with less data volume, the model is updated in time according to the change of the failure rate, and the application scene of the model is expanded.

Description

Technical Field

[0001] The present invention belongs to the technical field of capacitor failure rate analysis, and in particular relates to a capacitor failure rate prediction method and computer equipment based on artificial intelligence. Background Art

[0002] Failure rate is an important indicator for measuring reliability. It means the number of failures that occur in a product per unit time after it has been working for time t. Mathematically, it is the limit of the conditional probability that a product that has not failed at time t will fail within a unit time of t+Δt.

[0003] The prediction and analysis of capacitor failure rate is crucial to improving its reliability and performance. At present, machine learning methods have shown great potential in capacitance extraction and supercapacitor performance prediction, which can improve the accuracy of prediction, time efficiency, and reduce costs. Different machine learning models, including artificial neural networks (ANN), support vector machines (SVM), decision trees (DT), random forests (RF), etc., are used to predict the performance of supercapacitors. These models can be trained based on input features such as structural properties of materials, synthesis conditions, operating conditions, etc. to predict capacitance and other performance indicators. However, the existing technology has the following disadvantages:

[0004] (1) Data requirements: Traditional methods often require a large amount of historical data to build models, which may not be accurate enough when the amount of data is small.

[0005] (2) Simplified assumptions: Traditional methods often assume that the failure rate is constant or follows a simple distribution (such as exponential distribution), which may not be applicable to all practical application scenarios.

[0006] (3) Lack of flexibility: Traditional methods are difficult to adapt to changes in the failure rate of equipment or systems under different usage conditions or at different life cycle stages. When equipment or systems change, traditional methods may need to re-collect data and recalculate failure rates, which may result in untimely updates of the prediction model. Summary of the invention

[0007] The object of the present invention is to provide a capacitor failure rate prediction method based on artificial intelligence, a computer device, a computer-readable storage medium and a computer program product to solve at least one problem existing in the prior art.

[0008] In order to achieve the above object, one aspect of the present invention provides a capacitor failure rate prediction method based on artificial intelligence, comprising:

[0009] Step S1, collecting data on production process parameters and product failure rates of capacitors, selecting characteristic parameters that affect product failure rates as inputs of a neural network model to be constructed, the target variable of the neural network model is failure rate, and performing Z-Score standardization on the characteristic parameters and failure rate data;

[0010] Step S2, construct a neural network model, including 1 input layer, 4 hidden layers, and 1 output layer. Each layer uses the ReLU function as the activation function. The output layer uses the linear activation function for prediction. The input layer applies L1 regularization to prevent overfitting. The Adam optimizer is used, and the mean square error is used as the loss function.

[0011] Step S3: Split the data into training and test sets using R 2 Scoring is used to evaluate the performance of the constructed neural network model. By repeating the training multiple times in a cycle, recording the R2 score each time and calculating the average of multiple times, a neural network model that meets the requirements is obtained.

[0012] Step S4, predicting the capacitor failure rate according to the constructed neural network model.

[0013] Preferably, in step S1, data normalization is performed as follows:

[0014]

[0015] Among them, Z is the original data point, u is the mean of the data set, σ is the standard deviation of the data set, and X is the standardized input feature vector.

[0016] Preferably, the expression of the neural network model constructed in step S2 is: y=W5(W4(ReLU(W3(ReLU(W2(ReLU(W1X+b1))+b2))+b3))+b4)+b5

[0017] Among them, W1, W2, W3, W4, and W5 are the weight matrices of each layer, b1, b2, b3, b4, and b5 are the bias vectors of each layer, ReLU represents the activation function, and y represents the output of the neural network model.

[0018] Preferably, in step S3, R is performed as follows 2 score:

[0019]

[0020] Among them, SS res represents the residual sum of squares, SS tot represents the total sum of squares, y i is the actual value of the ith observation, is the predicted value of the ith observation, is the mean of the actual values ​​and n is the number of observations.

[0021] Preferably, in step S1, the average printing thickness, casting thickness, slurry viscosity, slurry solid content, and sintering temperature are selected as characteristic parameters.

[0022] Another aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.

[0023] Another aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when executed by a processor.

[0024] Another aspect of the present invention provides a computer program product, comprising a computer program, which implements the steps of the above method when executed by a processor.

[0025] According to the reflectivity thermal imaging test method, computer equipment, computer-readable storage medium and computer program product of the above aspects of the present invention, it is possible to establish the required neural network model with a smaller amount of data, and update the model in time according to changes in failure rate, thereby expanding the application scenarios of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings used in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work:

[0027] Figure 1 is a flow chart of a capacitor failure rate prediction method based on artificial intelligence according to an embodiment of the present invention;

[0028] Figure 2 is a structural diagram of a neural network model according to an embodiment of the present invention;

[0029] Figure 3 It is a structural diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] An embodiment of the present invention provides a capacitor failure rate prediction method based on artificial intelligence. Figure 1 FIG. 1 is a flow chart of a capacitor failure rate prediction method based on artificial intelligence according to an embodiment of the present invention. Figure 1 As shown, the capacitor failure rate prediction method based on artificial intelligence in the embodiment of the present invention includes steps S1 to S4.

[0032] In step S1, data on production process parameters and product failure rates of capacitors are collected, and characteristic parameters are selected as inputs of a neural network model to be constructed.

[0033] The inventors of the present invention have found that the failure rate of capacitors is closely related to multiple process parameters, including slurry viscosity, slurry particle size, slurry solid content, casting thickness, ceramic tape weight loss rate, average printing thickness and sintering temperature.

[0034] The thickness of the tape casting directly affects the voltage resistance of the capacitor and is one of the key factors affecting the reliability of the capacitor. If the tape casting thickness is too thin, the capacitor's voltage resistance may be insufficient; if it is too thick, it may increase the size and weight of the capacitor, and also affect its performance and reliability.

[0035] High-viscosity ceramic slurry may cause uneven dispersion and coating of electrode materials, thus affecting the internal structure and performance of the capacitor and increasing the failure rate. Conversely, although low-viscosity slurry is conducive to coating uniformity, it may also cause excessive dispersion of electrode materials, affecting the cycle life and stability of the capacitor.

[0036] The particle size distribution of ceramic slurry directly affects the density of ceramic materials. Ceramic materials with uniform particle size distribution have a higher specific surface area, which is conducive to the diffusion and transmission of ions, thereby improving the voltage resistance and charge and discharge performance of the capacitor medium. If the slurry particles are too large or unevenly distributed, it may cause pore defects inside the capacitor medium, reduce the dielectric voltage resistance, and increase the failure rate.

[0037] The solid content of ceramic slurry is closely related to the stability and coating efficiency of the slurry. The higher the solid content, the greater the viscosity of the slurry, and the higher the coating efficiency may be, but it may also require higher equipment and have greater losses. The appropriate solid content can ensure that the slurry forms a uniform film during the coating process, thereby improving the reliability and stability of the capacitor.

[0038] The weight loss rate of the ceramic tape reflects the mass loss of the ceramic diaphragm during the sintering process. If the weight loss rate is too high, it may cause defects such as cracks or delamination in the ceramic diaphragm, thus affecting the performance and reliability of the capacitor.

[0039] The average printing thickness determines the thickness and uniformity of the electrode pattern. If the printing thickness is uneven, it may cause uneven distribution of the electric field inside the capacitor, increase leakage current and failure risk, and affect the performance and reliability of the capacitor.

[0040] Sintering temperature is one of the key factors affecting the performance of ceramic capacitors. Appropriate sintering temperature can ensure that the ceramic diaphragm and metal materials are tightly combined to form a ceramic body with high mechanical strength and excellent electrical properties. If the sintering temperature is too high or too low, it may cause defects such as delamination of the ceramic diaphragm or performance degradation, thus affecting the reliability and failure rate of the capacitor.

[0041] In the embodiment of the present invention, examples of the statistical data of the above process parameters and product failure rates are shown in Table 1 below.

[0042] Table 1 Data example of process parameters and product failure rate

[0043]

[0044] The present invention uses a neural network model to predict the failure rate, and selects the following five characteristic parameters as model inputs: average printing thickness, casting thickness, slurry viscosity, slurry solid content, and sintering temperature. The target variable is "failure rate % / 1000h".

[0045] The selected feature parameters and failure rate data are Z-Score standardized (standard deviation standardized). Through standardization, it can be ensured that different features affect the model training at the same scale, thereby improving the performance of the model. At the same time, because the failure rate value is too small, the failure rate is extracted to 1E-9 to ensure the accuracy of the value.

[0046]

[0047] Among them, Z is the original data point, u is the mean of the data set, σ is the standard deviation of the data set, and X is the standardized input feature vector. Standardize each feature parameter, sort the standardized data by column, and each row is a data.

[0048] In step S2, a neural network model is constructed, which contains 6 layers, including 1 input layer, 4 hidden layers, and 1 output layer. Each layer uses the ReLU function (Rectified Linear Unit) as the activation function, and the output layer uses the linear activation function for prediction. The input layer applies L1 regularization to prevent overfitting, the network uses the Adam optimizer, and the mean squared error (MSE) is used as the loss function.

[0049] Figure 2 is a structural diagram of a neural network model according to an embodiment of the present invention. Figure 2 As shown in the figure, the input layer has 256 neurons, the activation function is ReLU, and L1 regularization is used to prevent overfitting. The hidden layers have 128, 64, and 32 neurons, respectively, and the activation function is ReLU. The output layer has 1 neuron, and the activation function is a linear function, which is used for regression tasks.

[0050] In one embodiment, 750 data are split into a training set, a validation set, and a test set. 60% (450 data) are randomly selected as the training data set, 20% (150 data) as the validation set, and 20% (150 data) as the test set. The early stopping method is used to monitor the loss of the validation set. If the loss of the validation set no longer decreases, the training will end early to avoid overfitting.

[0051] By drawing the loss curve of the neural network model and comparing the loss changes of the neural network model on the training set and the validation set (Loss curve), it is confirmed that the model has converged.

[0052] In step S3, the model is evaluated and optimized. 2 Scoring is used to evaluate the performance of the model on the training set and the test set. 2 The score indicates how well the model fits the actual data. 2 The value measures the proportion of the total variability that is explained by the model. A value close to 1 indicates that the model is able to explain the variability in the data well.

[0053]

[0054] Among them, SS res is the residual sum of squares, calculated as SS tot is the total sum of squares, calculated as y i is the actual value of the ith observation. is the predicted value for the ith observation. is the mean of the actual values. n is the number of observations.

[0055] In order to reduce the contingency of a single training, in this embodiment, the training was repeated 5 times in a cycle, the R2 score of each time was recorded, and the average R2 score of the 5 results was calculated. 2 , as the final performance evaluation of the model. Average R 2 The score is 0.601, which is an ideal score. After each training, the test set is predicted and the results are visualized to compare the actual value with the predicted value.

[0056] The expression of the constructed neural network model is:

[0057] y=W5(W4(ReLU(W3(ReLU(W2(ReLU(W1X+b1)))+b2))+b3))+b4)+b5

[0058] Where X is the normalized input feature vector. W1, W2, W3, W4, and W5 are the weight matrices of each layer. b1, b2, b3, b4, and b5 are the bias vectors of each layer. ReLU represents the activation function ReLU(x) = max(0, x). y represents the output of the neural network model.

[0059] In step S4, the capacitor failure rate is predicted based on the constructed neural network model.

[0060] The artificial intelligence-based capacitor failure rate prediction method of the embodiment of the present invention builds a reliability prediction model by statistically analyzing relevant data such as process parameters and product failure rates, and then tests process parameters in key process processes to conduct reliability evaluation of the batch of products in advance. This can not only reduce unnecessary testing work and costs, but also avoid product operation with defects caused by process control problems, reduce the possibility of accidents with poor quality of product batches, and avoid fatal failures.

[0061] Moreover, the artificial intelligence-based capacitor failure rate prediction method of the embodiment of the present invention can establish the required neural network model with a smaller amount of data, can update the model in time according to changes in the failure rate, and does not need to assume that the failure rate is constant or follows a simple distribution, thereby expanding the application scenarios of the model.

[0062] An embodiment of the present invention further provides a computer device, which may be a server, and its internal structure diagram may be as shown in FIG. Figure 3As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store operating parameter data of each framework. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the steps of the method of the embodiment of the present invention are implemented.

[0063] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0064] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method of the embodiment of the present invention are implemented.

[0065] An embodiment of the present invention further provides a computer program product, including a computer program, which implements the steps of the method of the embodiment of the present invention when executed by a processor.

[0066] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A capacitor failure rate prediction method based on artificial intelligence, characterized in that: include: Step S1, collecting data on production process parameters and product failure rates of capacitors, selecting characteristic parameters that affect product failure rates as inputs of a neural network model to be constructed, the target variable of the neural network model is failure rate, and performing Z-Score standardization on the characteristic parameters and failure rate data; Step S2, construct a neural network model, including 1 input layer, 4 hidden layers, and 1 output layer. Each layer uses the ReLU function as the activation function. The output layer uses the linear activation function for prediction. The input layer applies L1 regularization to prevent overfitting. The Adam optimizer is used, and the mean square error is used as the loss function. Step S3: Split the data into training and test sets using R 2 Scoring is used to evaluate the performance of the constructed neural network model. By repeating the training multiple times in a cycle, recording the R2 score each time and calculating the average of multiple times, a neural network model that meets the requirements is obtained. Step S4, predicting the capacitor failure rate according to the constructed neural network model.

2. The method according to claim 1, characterized in that In step S1, the data is normalized as follows: Among them, Z is the original data point, u is the mean of the data set, σ is the standard deviation of the data set, and X is the standardized input feature vector.

3. The method according to claim 2, characterized in that The expression of the neural network model constructed in step S2 is: y=W5(W4(ReLU(W3(ReLU(W2(ReLU(W1X+b1))+b2))+b3))+b4)+b5 Among them, W1, W2, W3, W4, and W5 are the weight matrices of each layer, b1, b2, b3, b4, and b5 are the bias vectors of each layer, ReLU represents the activation function, and y represents the output of the neural network model.

4. The method according to any one of claims 1 to 3, characterized in that In step S3, R is performed as follows 2 score: Among them, SS res represents the residual sum of squares, SS tot represents the total sum of squares, y i is the actual value of the ith observation, is the predicted value of the ith observation, is the mean of the actual values ​​and n is the number of observations.

5. The method according to any one of claims 1 to 3, characterized in that In step S1 , the average printing thickness, casting thickness, slurry viscosity, slurry solid content, and sintering temperature are selected as characteristic parameters.

6. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images