One-time electronic tag service life prediction method based on big data analysis and artificial intelligence

Abstract

The invention provides a one-time electronic tag service life prediction method based on big data analysis and artificial intelligence, relates to the technical field of electronic tag service life prediction, and remarkably improves data quality by using a graph neural network and a generative adversarial network to perform accurate preprocessing on historical data and working condition data. And a life prediction model and a fault prediction model are respectively constructed by adopting a multiple linear regression model and a deep learning model, so that the precision and reliability of the models are enhanced. According to the method, through combination of optimization processing and multiple models, the service life and the fault risk of the electronic tag can be predicted more accurately, so that the practicability and the stability of the whole scheme are improved, and powerful support is provided for tag maintenance and decision-making in industrial and logistics management.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic tag life prediction, and in particular to a disposable electronic tag life prediction method based on big data analysis and artificial intelligence. Background Art

[0002] In modern industry and logistics management, disposable electronic tags are widely used in product tracking, inventory management, asset monitoring and other fields. Due to the complex and changeable working environment, the life and reliability of electronic tags have become an important issue. However, in the study of the life of electronic tags, there are still some significant shortcomings in traditional methods, such as the limitation of a single environmental stress and the lack of data support. Traditional methods usually only predict the life under a single or a few environmental stresses, and fail to fully consider the synergistic coupling of various stresses in a complex and changeable environment, resulting in the prediction results being difficult to accurately reflect the life changes of electronic tags in practical applications. At the same time, traditional methods often rely on limited experimental data or assumptions. These data are prone to introduce bias or lose useful information during the collection and processing process, resulting in insufficient accuracy in model training and prediction. The limited generalization ability further affects the accuracy and robustness of the prediction model.

[0003] Therefore, in order to solve the problems existing in traditional methods, it is necessary to provide a method that can efficiently process complex data and accurately predict the life and fault status of disposable electronic tags.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0005] The purpose of the present invention is to provide a disposable electronic tag life prediction method based on big data analysis and artificial intelligence to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: The life prediction method of disposable electronic tags based on big data analysis and artificial intelligence includes the following steps: S1: Take several groups of disposable electronic tags as samples, divide the samples into the first test group and the second test group for life test, collect historical data of all samples in the life test, and pre-process the historical data; S2: constructing a life prediction model and a fault prediction model, and using the historical data of the first test group and the second test group to train and optimize the life prediction model and the fault prediction model respectively; S3: Collect the working condition data of the disposable electronic tag in the actual working environment and pre-process the working condition data; S4: Substitute the preprocessed working condition data into the optimized life prediction model and fault prediction model in turn, make a preliminary prediction of the life and fault status of the disposable electronic tag, generate a risk index based on the preliminary prediction results, and make a final evaluation of the life of the disposable electronic tag based on the risk index.

[0007] Preferably, the historical data include the test temperature, test humidity, test pressure, and test amplitude during the test; the operating condition data include the working temperature, working humidity, working pressure, and working amplitude under the actual working environment; and the preprocessing of the historical data and operating condition data includes data cleaning, data denoising, missing value filling, and normalization.

[0008] Preferably, a preset A graph neural network with a graph convolutional layer is used to clean and denoise the historical data and working condition data. The output data of the graph neural network is expressed as: ; In the formula Representation Node In the The output data of the graph convolution layer, represents the activation function of the graph neural network, Representation Node The set of neighbor nodes of Represents the index of the neighbor node, represents the normalization coefficient, Indicates The weight matrix of the graph convolutional layer; Substitute the output data of the graph neural network into the preset adversarial network model to fill in the missing values. The adversarial network model is expressed as: ; In the formula Represents the loss function of the adversarial network model, where , Respectively represent the discriminator and generator of the adversarial network model, , They represent the distribution of the output data and the distribution of noise of the graph neural network respectively. represents noise, , They represent the expected value of the distribution of the graph neural network output data and the expected value of the distribution of noise respectively; After the missing values ​​are filled, the output data of the adversarial network model is normalized, and the calculation method is: ; In the formula represents normalized data, , , Respectively represent the output data of the adversarial network model, and the maximum and minimum values ​​of the output data of the adversarial network model; The preset label generation algorithm is then used to label the normalized historical data. According to the type of fault generated by the sample in the life test, different fault labels are generated to classify the historical data of each sample, and the fault labels are numbered and sorted based on the severity of the fault type.

[0009] Preferably, the logic for constructing the life prediction model and the failure prediction model is: A life prediction model for disposable electronic tags is built based on a multivariate linear regression model, and a fault prediction model for disposable electronic tags is built based on a deep learning model and a softmax function. The historical data of the first test group and the second test group are divided into the first training set, the first validation set and the second training set, and the second validation set in a ratio of 8:2 respectively; The life prediction model is trained and optimized using the first training set and the first validation set, and the fault prediction model is trained and optimized using the second training set and the second validation set.

[0010] Preferably, the life prediction model is expressed as: ; In the formula represents the predicted life of a disposable electronic tag, represents the model parameters of the life prediction model, , , , represent temperature, humidity, pressure and amplitude respectively, Indicates the preset first margin.

[0011] Preferably, the construction logic of the fault prediction model is: The normalized data is input into the deep learning model to calculate its unnormalized score, which is expressed as: ; Then substitute the unnormalized score into the softmax function to convert it into a predicted probability distribution, which is expressed as: ; ; In the formula Indicates A one-time electronic tag is The predicted probability within the fault label, Indicates A one-time electronic tag is The unnormalized scores within the fault labels, Indicates the number of fault labels, Indicates The predicted probability distribution of a disposable electronic tag.

[0012] Preferably, when training the life prediction model and the fault prediction model, the loss is calculated according to the loss function of the life prediction model and the fault prediction model, wherein: The loss function of the life prediction model is expressed as: ; The loss function of the fault prediction model is expressed as: ; In the formula , represent the losses of the life prediction model and the failure prediction model respectively, , are the number of samples in the first training set and the second training set respectively, , Respectively represent The predicted and actual lifespan of a disposable electronic tag. Indicates A one-time electronic tag is The actual probability of a fault label; The back propagation algorithm is used to calculate the gradient of the loss function of the life prediction model and the fault prediction model respectively, and then the gradient descent method is used to update the life prediction model and the fault prediction model. The first validation set and the second validation set are respectively substituted into the life prediction model and the fault prediction model for optimization. When the loss of the life prediction model Less than the preset first threshold , and the loss of the fault prediction model Less than the preset second threshold When , the optimization of life prediction model and failure prediction model is considered complete.

[0013] Preferably, the logic of risk assessment for disposable electronic tags is: Substitute the working condition data of the disposable electronic tag into the life prediction model and the fault prediction model respectively to obtain the corresponding predicted life and predicted probability distribution; The risk index is generated based on the predicted life span and the predicted probability distribution, and is calculated as follows: ; In the formula Indicates The risk index of a disposable electronic tag; Compare the risk index with the preset risk threshold, and when it meets: , think that The risk of a disposable electronic tag is low, and the predicted life span is long; , think that The disposable electronic tag is medium risk and the predicted life expectancy is medium; , think that A disposable electronic tag is high risk and has a short predicted lifespan; In the formula , Respectively represent the upper and lower limits of the risk threshold, and .

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention uses graph neural networks and generative adversarial networks to accurately preprocess historical data and operating condition data, significantly improving data quality. Multivariate linear regression models and deep learning models are used to construct life prediction and failure prediction models, respectively, enhancing the accuracy and reliability of the models. This method combines optimization processing with multiple models to more accurately predict the life and failure risk of electronic tags, thereby improving the practicality and stability of the overall solution and providing strong support for tag maintenance and decision-making in industrial and logistics management. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0016] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0017] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0018] Example: See also Figure 1 , the present invention provides a technical solution: The life prediction method of disposable electronic tags based on big data analysis and artificial intelligence includes the following steps: S1: Take several groups of disposable electronic tags as samples, divide the samples into a first test group and a second test group for life test, collect historical data of all samples in the life test, and pre-process the historical data.

[0019] The historical data include the test temperature, test humidity, test pressure, and test amplitude during the test. The preprocessing of the historical data includes data cleaning, data denoising, missing value filling, and normalization.

[0020] During preprocessing, use the preset A graph neural network with a graph convolutional layer is used to clean and denoise the historical data and working condition data. The output data of the graph neural network is expressed as: ; In the formula Representation Node In the The output data of the graph convolution layer, represents the activation function of the graph neural network, Representation Node The set of neighbor nodes of Represents the index of the neighbor node, represents the normalization coefficient, Indicates The weight matrix of the graph convolution layer is a node that represents an observed state or data point of a disposable electronic tag in a life test or actual working environment. For example, a node can represent a collection of environmental parameters such as temperature, humidity, pressure, vibration, etc. at a certain moment. These nodes form a graph structure to capture and represent the complex relationship between environmental factors. After being processed by the graph neural network, the accuracy and consistency of the data are improved, providing more reliable input for subsequent life and failure prediction. Specifically, assuming that at time point The environmental parameters collected at this time are temperature ,humidity ,pressure and amplitude , then the corresponding node feature can be expressed as a vector , the vector is input into GNN through the graph structure, and then After layer convolution, the denoised features can be obtained .

[0021] Graph neural networks, or GNNs, can use the graph structure information of data to effectively capture complex associations and patterns in the data, so they perform well when processing multi-dimensional and unstructured data. Compared with traditional denoising methods, GNNs can more accurately retain useful information and remove noise.

[0022] Substitute the output data of the graph neural network into the preset adversarial network model to fill in the missing values. The adversarial network model is expressed as: ; In the formula Represents the loss function of the adversarial network model, where , Respectively represent the discriminator and generator of the adversarial network model, , They represent the distribution of the output data and the distribution of noise of the graph neural network respectively. represents noise, , They represent the expected value of the distribution of the graph neural network output data and the expected value of the distribution of noise respectively.

[0023] In the adversarial network model, that is, GAN, This item calculates the expected value of the logarithm of the output probability of the discriminator for the input data of the graph neural network, that is, the real data. One calculation is the expected value of the logarithm of the output probability of the discriminator for the noise, that is, the generated data. The goal of the discriminator is to maximize the sum of these two parts so that the discriminator can more accurately identify real data and generated data, while the goal of the generator is to minimize This part makes the probability that the generated data is considered as real data by the discriminator as high as possible, thereby generating more realistic data to deceive the discriminator. GAN achieves missing value filling by generating adversarial methods, which can generate complementary information that is closer to real data. Compared with simple mean filling or interpolation methods, GAN's generator and discriminator are optimized at the same time, which improves the accuracy of filling.

[0024] After the missing values ​​are filled, the output data of the adversarial network model is normalized, and the calculation method is: ; In the formula represents normalized data, , , They respectively represent the output data of the adversarial network model, as well as the maximum and minimum values ​​of the output data of the adversarial network model. Combined with the normalization processing after the adversarial network output, the stability and consistency of the data when inputting the model are ensured, which improves the effect of subsequent model training.

[0025] Here is an example to illustrate, vector When using graph neural networks, Layer convolution to obtain denoised features , and then the denoised features Input into the adversarial network model to fill in missing values. Assume that the data after missing value filling is , then at this time It is equivalent to the output data of the adversarial network model, which is equivalent to the normalization calculation formula , the data after normalization can be expressed as .

[0026] The preset label generation algorithm is then used to label the normalized historical data. According to the fault type generated by the sample in the life test, different fault labels are generated to classify the historical data of each sample, and the fault labels are numbered and sorted based on the severity of the fault type. Fault labels are automatically generated and numbered, and sorted based on the severity of the fault type. This can effectively support the training of subsequent classification models and improve the model's ability to recognize different fault types.

[0027] In this step, the preset label generation algorithm can be implemented by using a combination model of the K-Means clustering analysis method and a custom sorting algorithm. The K-Means clustering analysis method can group similar fault features in historical data, and apply a clustering algorithm to identify naturally distributed groups in the data after normalization. Each group represents a fault type, and finally each data is assigned to a specific clustering group, thereby preliminarily identifying possible fault types and obtaining corresponding fault labels. In the above embodiment, it is assumed that the normalized data set is , ,in represents the number of samples in the data set. The K-Means algorithm is executed on the data set. The number of clusters is set to 3. Three data are randomly selected as the initial cluster centers, which are recorded as , calculate the Euclidean distance between each data and the cluster center , subscript Represents the index of the cluster center, and the calculation method is expressed as: ; In the formula Indicates the index of the dimension. In this embodiment, since there are four features: temperature, humidity, pressure and amplitude, the number of dimensions is 4. Each data is assigned to the nearest cluster center, that is, a cluster classification is completed, and then each cluster is updated, and the cluster center is updated to the mean value of all data in the cluster. The calculation method is expressed as: ; In the formula Represents the data set in the cluster. Repeat the cluster classification and cluster center update for iteration. When the cluster center no longer changes significantly, it is considered to be converged, and the cluster label corresponding to each sample, that is, the fault label, can be output.

[0028] The custom sorting algorithm can use numerical scores or weighted rules to evaluate the severity of each fault. Each fault label is assigned a severity score and numbered in order from low to high. The specific numerical scores or weighted rules can be obtained by combining expert scoring and domain knowledge. Assuming that the fault labels include C1 high temperature, C2 unstable humidity, and C3 sensor drift, the corresponding severity scores are S1=3, S2=2, and S3=1, respectively, indicating the severity of the three faults. Then the fault labels can be sorted according to the severity score and then renumbered according to the sorting.

[0029] Here is a simple example to illustrate. Suppose a batch of disposable electronic tags are undergoing life tests. Due to sensor failure, some data is missing, and the collected data is interfered by environmental noise. Traditional methods may use the mean to fill in the missing values ​​and perform denoising through simple filtering. This may introduce bias in the filling process and lose valid information during denoising. In this invention, accurate data preprocessing and missing value filling ensure the high quality of input data, which can improve the performance of the life prediction model and the fault prediction model, thereby improving the accuracy and robustness of the prediction model; and effectively remove noise and fill in missing values ​​in the preprocessing stage, so that the model can maintain stable performance when dealing with data in different environments and working conditions, thereby enhancing the generalization ability of the model.

[0030] S2: Construct a life prediction model and a fault prediction model, and use the historical data of the first test group and the second test group to train and optimize the life prediction model and the fault prediction model respectively.

[0031] The logic for building the life prediction model and the failure prediction model is: A life prediction model for disposable electronic tags is built based on a multivariate linear regression model, and a fault prediction model for disposable electronic tags is built based on a deep learning model and a softmax function. The historical data of the first test group and the second test group are divided into the first training set, the first validation set and the second training set, and the second validation set in a ratio of 8:2 respectively; The life prediction model is trained and optimized using the first training set and the first validation set, and the fault prediction model is trained and optimized using the second training set and the second validation set.

[0032] The samples are divided into two groups, each used for different model training, making full use of data resources and improving model pertinence. The optimized models can be adjusted and expanded according to different application scenarios, so that the solution can maintain good performance in a variety of environments. Compared with the traditional method of using a single model for unified prediction, the present invention can take into account the needs of both life and failure, thereby more accurately predicting the life and failure type of electronic tags, and supporting more effective decision-making and maintenance plans.

[0033] The life prediction model is expressed as: ; In the formula represents the predicted life of a disposable electronic tag, represents the model parameters of the life prediction model, , , , represent temperature, humidity, pressure and amplitude respectively, It represents the preset first margin, and the specific size of the first margin can be set according to the user's experience in model training.

[0034] The construction logic of the fault prediction model is: The normalized data is input into the deep learning model to calculate its unnormalized score, which is expressed as: ; The unnormalized score represents the score of a specific sample in a specific category during forward propagation. It is the model's internal evaluation value of whether the sample belongs to this category. By performing a softmax transformation on these scores, the predicted probability of each sample in each category can be obtained.

[0035] Then substitute the unnormalized score into the softmax function to convert it into a predicted probability distribution, which is expressed as: ; ; In the formula Indicates A one-time electronic tag is The predicted probability within the fault label, Indicates A one-time electronic tag is The unnormalized scores within the fault labels, Indicates the number of fault labels, Indicates The predicted probability distribution of a disposable electronic tag.

[0036] Here is a simple example to illustrate. Suppose there are 3 types of fault labels, which are marked as 1, 2, and 3 according to severity. The corresponding fault types are poor contact, partial damage, and overall damage, respectively. , , This means that the probability of poor contact, partial damage, and overall damage of sample No. 1 is 10%, 70%, and 20% respectively. The predicted probability distribution of sample No. 1 can be expressed as .

[0037] The loss function of the life prediction model is expressed as: ; The loss function of the fault prediction model is expressed as: ; In the formula , represent the losses of the life prediction model and the failure prediction model respectively, , are the number of samples in the first training set and the second training set respectively, , Respectively represent The predicted and actual lifespan of a disposable electronic tag. Indicates A one-time electronic tag is The actual probability of a fault label.

[0038] When training the life prediction model and the fault prediction model, the loss is calculated according to the loss function of the life prediction model and the fault prediction model, where: The back propagation algorithm is used to calculate the gradient of the loss function of the life prediction model and the fault prediction model respectively, and then the gradient descent method is used to update the life prediction model and the fault prediction model. The first validation set and the second validation set are respectively substituted into the life prediction model and the fault prediction model for optimization. When the loss of the life prediction model Less than the preset first threshold , and the loss of the fault prediction model Less than the preset second threshold When , it is considered that the life prediction model and the fault prediction model are optimized. The first threshold and the second threshold here can be set according to the user's experience in model training.

[0039] Life prediction is a multivariate regression task, and its goal is to explore the impact of environmental factors such as temperature, humidity, pressure, and amplitude on the life of disposable electronic tags. Therefore, a multivariate linear regression model is used to construct a life prediction model, and the root mean square error is used as the loss function of the life prediction model to minimize the gap between the predicted life and the actual life. It can be understood that the temperature, humidity, pressure, and amplitude in the life prediction model are a generalized concept, that is, when the input data is historical data, the temperature, humidity, pressure, and amplitude correspond to the test temperature, test humidity, test pressure, and test amplitude in the historical data; when the data is working condition data, the temperature, humidity, pressure, and amplitude correspond to the working temperature, working humidity, working pressure, and working amplitude in the working condition data.

[0040] Fault prediction is a classification task, so the fault prediction model is constructed by coupling the deep learning model with the softmax function, and the cross entropy loss is used as the loss function of the fault prediction model to maximize the consistency between the predicted probability distribution and the true probability distribution.

[0041] In this step, by applying multivariate linear regression to life prediction, combined with deep learning models and softmax functions for fault prediction, the accuracy and reliability of prediction can be significantly improved. Separately training and optimizing models for different tasks ensures the comprehensiveness and adaptability of the solution, thereby enhancing the practicality and effectiveness of the overall solution.

[0042] S3: Collect the working condition data of the disposable electronic tag in the actual working environment and pre-process the working condition data.

[0043] The working condition data includes the working temperature, working humidity, working pressure, and working amplitude in the actual working environment. Similar to the preprocessing method of historical data, the preprocessing of working condition data also includes data cleaning, data denoising, missing value filling, and normalization.

[0044] S4: Substitute the preprocessed working condition data into the optimized life prediction model and fault prediction model in turn, make a preliminary prediction of the life and fault status of the disposable electronic tag, generate a risk index based on the preliminary prediction results, and make a final evaluation of the life of the disposable electronic tag based on the risk index.

[0045] The logic of risk assessment for disposable electronic tags is: Substitute the working condition data of the disposable electronic tag into the life prediction model and the fault prediction model respectively to obtain the corresponding predicted life and predicted probability distribution; The risk index is generated based on the predicted life span and the predicted probability distribution, and is calculated as follows: ; In the formula Indicates The risk index of a disposable electronic tag; Compare the risk index with the preset risk threshold, and when it meets: , think that The risk of a disposable electronic tag is low, and the predicted life span is long; , think that The disposable electronic tag is medium risk and the predicted life expectancy is medium; , think that A disposable electronic tag is high risk and has a short predicted lifespan; In the formula , Respectively represent the upper and lower limits of the risk threshold, and .

[0046] From the calculation formula of the risk index, it can be seen that the risk index is proportional to the severity of the fault of the disposable electronic tag and inversely proportional to the predicted life of the disposable electronic tag, reflecting the ability of the disposable electronic tag to work stably under actual working conditions. The larger the risk index, the greater the probability of failure, the shorter the normal use time, and the weaker the ability to work stably under actual working conditions; conversely, the smaller the risk index, the smaller the probability of failure, the longer the normal use time, and the stronger the ability to work stably under actual working conditions.

[0047] In summary, the present invention uses graph neural networks and generative adversarial networks to accurately preprocess historical data and operating condition data, significantly improving data quality. Multivariate linear regression models and deep learning models are used to construct life prediction and fault prediction models, respectively, enhancing the accuracy and reliability of the models. This method combines optimization processing with multiple models to more accurately predict the life and failure risk of electronic tags, thereby improving the practicality and stability of the overall solution and providing strong support for tag maintenance and decision-making in industrial and logistics management.

[0048] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0049] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0050] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0051] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A disposable electronic tag life prediction method based on big data analysis and artificial intelligence, characterized in that: The specific steps include: S1: Take several groups of disposable electronic tags as samples, divide the samples into the first test group and the second test group for life test, collect historical data of all samples in the life test, and pre-process the historical data; S2: constructing a life prediction model and a fault prediction model, and using the historical data of the first test group and the second test group to train and optimize the life prediction model and the fault prediction model respectively; S3: Collect the working condition data of the disposable electronic tag in the actual working environment and pre-process the working condition data; S4: Substitute the preprocessed working condition data into the optimized life prediction model and fault prediction model in turn, make a preliminary prediction of the life and fault status of the disposable electronic tag, generate a risk index based on the preliminary prediction results, and make a final evaluation of the life of the disposable electronic tag based on the risk index.

2. The method for predicting the life of a disposable electronic tag based on big data analysis and artificial intelligence according to claim 1 is characterized in that: The historical data include the test temperature, test humidity, test pressure, and test amplitude during the test; the operating condition data include the working temperature, working humidity, working pressure, and working amplitude under the actual working environment; the preprocessing of the historical data and operating condition data includes data cleaning, data denoising, missing value filling, and normalization processing.

3. The method for predicting the life of a disposable electronic tag based on big data analysis and artificial intelligence according to claim 2 is characterized in that: Use the preset A graph neural network with a graph convolutional layer is used to clean and denoise the historical data and working condition data. The output data of the graph neural network is expressed as: ； In the formula Representation Node In the The output data of the graph convolution layer, represents the activation function of the graph neural network, Representation Node The set of neighbor nodes of Represents the index of the neighbor node, represents the normalization coefficient, Indicates The weight matrix of the graph convolutional layer; Substitute the output data of the graph neural network into the preset adversarial network model to fill in the missing values. The adversarial network model is expressed as: ； In the formula Represents the loss function of the adversarial network model, where , Respectively represent the discriminator and generator of the adversarial network model, , They represent the distribution of the output data and the distribution of noise of the graph neural network respectively. represents noise, , They represent the expected value of the distribution of the graph neural network output data and the expected value of the distribution of noise respectively; After the missing values ​​are filled, the output data of the adversarial network model is normalized, and the calculation method is: ； In the formula represents normalized data, , , Respectively represent the output data of the adversarial network model, and the maximum and minimum values ​​of the output data of the adversarial network model; The preset label generation algorithm is then used to label the normalized historical data. According to the type of fault generated by the sample in the life test, different fault labels are generated to classify the historical data of each sample, and the fault labels are numbered and sorted based on the severity of the fault type.

4. The method for predicting the life of a disposable electronic tag based on big data analysis and artificial intelligence according to claim 3 is characterized in that: The logic for building the life prediction model and the failure prediction model is: A life prediction model for disposable electronic tags is built based on a multivariate linear regression model, and a fault prediction model for disposable electronic tags is built based on a deep learning model and a softmax function. The historical data of the first test group and the second test group are divided into the first training set, the first validation set and the second training set, and the second validation set in a ratio of 8:2 respectively; The life prediction model is trained and optimized using the first training set and the first validation set, and the fault prediction model is trained and optimized using the second training set and the second validation set.

5. The method for predicting the life of a disposable electronic tag based on big data analysis and artificial intelligence according to claim 4 is characterized in that: The life prediction model is expressed as: ； In the formula represents the predicted life of a disposable electronic tag, represents the model parameters of the life prediction model, , , , represent temperature, humidity, pressure and amplitude respectively, Indicates the preset first margin.

6. The method for predicting the life of a disposable electronic tag based on big data analysis and artificial intelligence according to claim 5 is characterized in that: The construction logic of the fault prediction model is: The normalized data is input into the deep learning model to calculate its unnormalized score, which is expressed as: ； Then substitute the unnormalized score into the softmax function to convert it into a predicted probability distribution, which is expressed as: ； ； In the formula Indicates A disposable electronic tag is The predicted probability within the fault label, Indicates A disposable electronic tag is The unnormalized scores within the fault labels, Indicates the number of fault labels, Indicates The predicted probability distribution of a disposable electronic tag.

7. The method for predicting the life of a disposable electronic tag based on big data analysis and artificial intelligence according to claim 6 is characterized in that: When training the life prediction model and the fault prediction model, the loss is calculated according to the loss function of the life prediction model and the fault prediction model, where: The loss function of the life prediction model is expressed as: ； The loss function of the fault prediction model is expressed as: ； In the formula , represent the losses of the life prediction model and the failure prediction model respectively, , are the number of samples in the first training set and the second training set respectively, , Respectively represent The predicted and actual lifespan of a disposable electronic tag. Indicates A disposable electronic tag is The actual probability of a fault label; The back propagation algorithm is used to calculate the gradient of the loss function of the life prediction model and the fault prediction model respectively, and then the gradient descent method is used to update the life prediction model and the fault prediction model. The first validation set and the second validation set are respectively substituted into the life prediction model and the fault prediction model for optimization. When the loss of the life prediction model Less than the preset first threshold , and the loss of the fault prediction model Less than the preset second threshold When , the optimization of life prediction model and failure prediction model is considered complete.

8. The method for predicting the life of a disposable electronic tag based on big data analysis and artificial intelligence according to claim 7 is characterized in that: The logic of risk assessment for disposable electronic tags is: Substitute the working condition data of the disposable electronic tag into the life prediction model and the fault prediction model respectively to obtain the corresponding predicted life and predicted probability distribution; The risk index is generated based on the predicted life span and the predicted probability distribution, and is calculated as follows: ； In the formula Indicates The risk index of a disposable electronic tag; Compare the risk index with the preset risk threshold, and when it meets: , think that The risk of a disposable electronic tag is low, and the predicted life span is long; , think that The disposable electronic tag is medium risk and the predicted life expectancy is medium; , think that A disposable electronic tag is high risk and has a short predicted lifespan; In the formula , Respectively represent the upper and lower limits of the risk threshold, and .

Images