A method, apparatus, electronic device, and storage medium for generating abnormal cigarette manufacturing process data.

Abstract

This invention discloses a method, apparatus, electronic device, and storage medium for generating abnormal cigarette manufacturing process data. The method includes: acquiring first cigarette manufacturing process data corresponding to multiple abnormal cigarettes in a target production batch, each first cigarette manufacturing process data carrying a first tag; determining a first process feature vector based on the first cigarette manufacturing process data; mapping the first tag to determine a first category feature vector; generating first noise information of a predetermined noise type; processing the first noise information based on a diffusion generation model using the first process feature vector, the first category feature vector, and a time step as first constraints to determine second noise information, the second noise information including first noise information to be adjusted from the first noise information; updating the first noise information based on the diffusion generation model and the second noise information to obtain target process features; and generating abnormal cigarette manufacturing process data based on the target process features, thus realizing the generation of abnormal cigarette manufacturing process data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, electronic device, and storage medium for generating abnormal cigarette manufacturing process data. Background Technology

[0002] As a core component of tobacco products, the quality and stability of cigarettes directly affect the consumer's smoking experience and the brand's market competitiveness. During cigarette production, factors such as batch differences in raw materials, fluctuations in equipment condition, and changes in environmental temperature and humidity inevitably lead to various types of cigarettes with quality abnormalities. In actual production, the proportion of cigarettes with quality abnormalities is usually low, and the number of abnormal samples is far less than that of normal samples. Therefore, it is necessary to generate a large amount of abnormal cigarette data based on the process data of cigarettes with quality abnormalities.

[0003] In existing technologies, methods such as random oversampling, generative adversarial networks (GANs), and synthetic minority class oversampling are commonly used to generate abnormal cigarette data. First, since abnormal cigarettes from different production batches have different cigarette manufacturing processes, the abnormal cigarette data generated using these methods cannot guarantee consistency with the actual manufacturing processes of the abnormal cigarettes in the corresponding production batches. Second, for abnormal cigarettes within the same production batch, which typically involve multiple cigarettes with simultaneously abnormal manufacturing processes, the abnormal cigarette data generated using the above methods suffers from low accuracy. Summary of the Invention

[0004] This invention provides a method, apparatus, electronic device, and storage medium for generating abnormal cigarette manufacturing process data, so as to achieve accurate generation of abnormal cigarette manufacturing process data.

[0005] According to one aspect of the present invention, a method for generating abnormal cigarette manufacturing process data is provided, comprising: Obtain the first cigarette process data corresponding to multiple abnormal cigarettes in the target production batch. The first cigarette process data carries a first tag, and the first tag represents the defect type of each abnormal cigarette in the target production batch. The first process feature vector is determined based on the first cigarette manufacturing process data; and the first category feature vector is determined by mapping the first label. Generate first noise information of the specified noise type; Using the first process feature vector, the first category feature vector, and the preset time step as the first constraint, the first noise information is processed based on the diffusion generation model to determine the second noise information, which includes the first noise information to be adjusted in the first noise information; the first noise information is updated in multiple iterations based on the diffusion generation model and the second noise information to obtain the target process features, and the iterations are determined based on the time step. Abnormal cigarette manufacturing process data are generated based on the target process characteristics.

[0006] According to another aspect of the present invention, an apparatus for generating abnormal cigarette manufacturing process data is provided, comprising: The cigarette manufacturing process data acquisition module is used to acquire the first cigarette manufacturing process data corresponding to multiple abnormal cigarettes in the target production batch. The first cigarette manufacturing process data carries a first tag, which represents the defect type of each abnormal cigarette in the target production batch. The feature vector determination module is used to determine a first process feature vector based on the first cigarette manufacturing data; and to map the first label to determine a first category feature vector. The first noise information generation module is used to generate first noise information of a specified noise type. The first noise information iteration module is used to process the first noise information based on the diffusion generation model, using the first process feature vector, the first category feature vector, and the preset time step as the first constraints, to determine the second noise information, which includes the first noise information to be adjusted in the first noise information; and to update the first noise information through multiple iterations based on the diffusion generation model and the second noise information to obtain the target process features, wherein the iteration rounds are determined based on the time step. The abnormal cigarette manufacturing process data generation module is used to generate abnormal cigarette manufacturing process data based on the target process characteristics.

[0007] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory that is communicatively connected to at least one processor; wherein, The memory stores a computer program that can be executed by at least one processor, such that the at least one processor is able to perform the method for generating abnormal cigarette manufacturing data according to any embodiment of the present invention.

[0008] According to another aspect of the present invention, a computer-readable storage medium is provided, which stores computer instructions for causing a processor to execute a method for generating abnormal cigarette manufacturing process data according to any embodiment of the present invention.

[0009] The technical solution of this invention achieves the acquisition of first cigarette process data corresponding to multiple abnormal cigarettes in a target production batch. Each first cigarette process data carries a first tag, which characterizes the defect type of each abnormal cigarette in the target production batch, thus providing comprehensive data support for subsequent analysis and processing. Based on the first cigarette process data, a first process feature vector is determined. Furthermore, the first tag is mapped to determine a first category feature vector, thus processing the first cigarette process data and the first tag, providing comprehensive and accurate data support for subsequent analysis and processing. First noise information of a defined noise type is generated, providing a data foundation for subsequent processing. Using the first process feature vector, the first category feature vector, and a preset time step as first constraints, the first noise information is processed based on a diffusion generation model to determine second noise information. The second noise information includes the first noise information containing a first undefined feature vector. Noise information is adjusted, and the first noise information is updated through multiple iterations based on the diffusion generation model and the second noise information to obtain the target process features. The iteration rounds are determined based on the time step, which achieves accurate determination of the target process features. This ensures consistency between the target process features and the first process feature vector corresponding to the target production batch, improving the authenticity and accuracy of the target process features and providing accurate data support for subsequent analysis and processing. Abnormal cigarette process data is generated based on the target process features, solving the problem of low accuracy in abnormal cigarette data generated by existing technologies. The abnormal cigarette process data is highly consistent with the first cigarette process data of abnormal cigarettes in the target production batch, and the defect types of the abnormal cigarettes are consistent with the defect types of abnormal cigarettes in the target production batch, improving the authenticity and accuracy of the abnormal cigarette process data and providing accurate and authentic data support for subsequent analysis and processing.

[0010] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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 effort.

[0012] Figure 1 This is a flowchart of a method for generating abnormal cigarette manufacturing process data provided in an embodiment of the present invention; Figure 2This is a flowchart of a training method for a diffusion generation model provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a noise generation module provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a device for generating abnormal cigarette manufacturing process data provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0013] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0014] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0015] Figure 1 This is a flowchart illustrating a method for generating abnormal cigarette manufacturing process data according to an embodiment of the present invention. This embodiment is applicable to situations involving the generation of abnormal cigarette manufacturing process data. The method can be executed by an abnormal cigarette manufacturing process data generation device, which can be implemented in hardware and / or software. This device can be configured in the electronic device provided in this embodiment of the invention. The electronic device can be a server, computer, or mobile terminal, such as a mobile phone or tablet computer. Figure 1 As shown, the method specifically includes the following steps: S110. Obtain the first cigarette process data corresponding to multiple abnormal cigarettes in the target production batch. The first cigarette process data carries a first tag, and the first tag represents the defect type of each abnormal cigarette in the target production batch.

[0016] The target production batch refers to the batch of cigarettes for which abnormal cigarette manufacturing process data is required. The target production batch can be selected according to needs; this invention does not impose any restrictions. An abnormal cigarette is one whose manufacturing process indicators deviate from the normal range during production. The first cigarette manufacturing process data refers to the manufacturing process indicators corresponding to the abnormal cigarette. Optionally, the first cigarette manufacturing process data includes physical process data and chemical process data; physical process data includes at least one of weight, draw resistance, ventilation rate, circumference, hardness, and length; chemical process data includes at least one of tar content, carbon monoxide content, and nicotine content. The first label is an identifier characterizing the defect type of each abnormal cigarette in the target production batch. Defect types include, but are not limited to, weight exceeding a preset weight range and draw resistance exceeding a preset draw resistance range. It should be noted that the first cigarette manufacturing process data in this invention includes the first cigarette manufacturing process data corresponding to abnormal cigarettes with both defect types. The first cigarette manufacturing process data can be obtained by detecting the manufacturing process indicators of abnormal cigarettes using a cigarette manufacturing process data detection device. The first cigarette manufacturing process data can also be obtained from a first cigarette manufacturing process database. For example, the unique identifier of the target production batch can be used to match the first cigarette process database, and the matched first cigarette process data can be used as the first cigarette process data corresponding to the target production batch. The first cigarette process database can store the first cigarette process data corresponding to different production batches.

[0017] Specifically, the process involves matching the unique identifier of the target production batch against the first cigarette process database, and then using the matched first cigarette process data as the first cigarette process data corresponding to the target production batch. This process enables the acquisition of the first cigarette process data and provides comprehensive data support for subsequent analysis and processing.

[0018] S120. Determine the first process feature vector based on the first cigarette manufacturing process data; and map the first label to determine the first category feature vector.

[0019] The first process feature vector represents the process distribution and correlation of abnormal cigarettes. The first process feature vector can be determined based on the process data of the first cigarette. For example, the process data of the first cigarette can be input into a trained first process feature vector determination model for processing to obtain the first process feature vector. The first process feature vector determination model includes, but is not limited to, a neural network model. Optionally, determining the first process feature vector based on the process data of the first cigarette includes: constructing an undirected weighted graph based on the process data of the first cigarettes corresponding to multiple cigarettes; and extracting process features from the undirected weighted graph to obtain the first process feature vector. The undirected weighted graph is a graph structure representing the degree of process similarity between abnormal cigarettes. The undirected weighted graph includes multiple nodes and edges. Nodes represent abnormal cigarettes, and edges represent the degree of process similarity between abnormal cigarettes. Different nodes can be connected by edges. The degree of process similarity between abnormal cigarettes can be determined based on the first cigarette process data of the abnormal cigarettes. For example, the first cigarette process data of the abnormal cigarettes can be input into a trained process similarity determination model for processing to obtain the degree of process similarity between abnormal cigarettes. The process similarity determination model includes, but is not limited to, a neural network model and a mathematical model. Process feature extraction from an undirected weighted graph can be performed by using a feature extraction model. The feature extraction model includes, but is not limited to, a neural network model. For example, the feature extraction model can be a graph convolutional network model.

[0020] The first category feature vector represents the defect type of the abnormal cigarette. The first category feature vector can be determined based on the first label. For example, the first label can be mapped to a preset vector dimension to obtain the first category feature vector. Mapping the first label can be done according to a first mapping model, which includes, but is not limited to, a neural network model. For example, the first mapping model sequentially includes a linear layer, an activation function layer, and another linear layer, where the activation function in the activation function layer is the SiLU activation function.

[0021] Specifically, the first cigarette manufacturing process data is input into a pre-trained first process feature vector determination model for processing to obtain the first process feature vector; the first label is mapped according to the first mapping model to a preset vector dimension to obtain the first category feature vector. This realizes the processing of the first cigarette manufacturing process data and the first label, providing comprehensive and accurate data support for subsequent analysis and processing.

[0022] S130, Generate first noise information of the set noise type.

[0023] The noise types include, but are not limited to, Gaussian noise, Laplace noise, uniform noise, and conditional noise. Optionally, the first noise information can be randomly generated based on Gaussian noise that follows a standard state distribution.

[0024] Specifically, the first noise information is randomly generated based on Gaussian noise that follows a standard state distribution, providing a data foundation for subsequent processing.

[0025] S140. Using the first process feature vector, the first category feature vector, and the preset time step as the first constraint, the first noise information is processed based on the diffusion generation model to determine the second noise information, which includes the first noise information to be adjusted in the first noise information; the first noise information is updated in multiple iterations based on the diffusion generation model and the second noise information to obtain the target process features, and the iterations are determined based on the time step.

[0026] The first constraint is the conditional information used to constrain the diffusion generation model. The first constraint includes a first process feature vector, a first category feature vector, and a time step. The time step is a progress indicator for generating production process indicators. The time step can be pre-set. The diffusion generation model is the structure used to process the first noise information. The diffusion generation model includes, but is not limited to, a neural network model. Optionally, the diffusion generation model is a conditional diffusion model. The second noise information is the data obtained by processing the first noise information through the diffusion generation model. The second noise information includes the first noise information to be adjusted within the first noise information. The first noise information to be adjusted is the noise information in the first noise information that needs to be adjusted. The second noise information can be determined based on the first constraint, the diffusion generation model, and the first noise information. For example, the first constraint and the first noise information can be input into a trained diffusion generation model for processing to obtain the second noise information. The target process feature is the process feature vector required by the user. The dimension of the target process feature is consistent with the dimension of the first process feature vector. The target process feature can be obtained by updating the first noise information through multiple iterations based on the diffusion generation model and the second noise information. For example, the diffusion generation model, the second noise information, the first noise information, and the iteration rounds can be input into a trained iterative processing model for processing to obtain the target process features. The iterative processing model includes, but is not limited to, a neural network model. The iteration rounds can be determined based on the time steps. There is a corresponding relationship between the iteration rounds and the time steps, including but not limited to a proportional relationship.

[0027] Specifically, the first process feature vector, the first category feature vector, and the preset time step are used as the first constraints. The first constraints and the first noise information are input into the trained diffusion generation model for processing to obtain the second noise information. The diffusion generation model, the second noise information, the first noise information, and the iteration rounds are input into the trained iterative processing model for processing to obtain the target process features. This achieves accurate determination of the target process features, ensuring consistency between the target process features and the first process feature vector corresponding to the target production batch. This improves the authenticity and accuracy of the target process features and provides accurate data support for subsequent analysis and processing.

[0028] Based on the above embodiments, before inputting the first constraint condition into the diffusion generation model for processing, the method further includes: mapping the time step to a corresponding time step embedding. For example, a trained time step embedding model can be used to perform high-dimensional mapping on the time step, mapping the time step to a time step embedding of a preset dimension. The time step embedding model includes, but is not limited to, a neural network model. Optionally, the time step embedding model is a multilayer perceptron. The time step embedding model sequentially includes a linear layer, an activation function layer, and another linear layer, where the activation function in the activation function layer is the SiLU activation function.

[0029] Based on the above embodiments, before inputting the first constraint condition into the diffusion generation model for processing, the method further includes: converting the first constraint condition into a preset constraint vector. For example, the time step embedding and the first process feature vector can be element-wise added to obtain a fused feature, and the fused feature and the first category feature vector can be concatenated along the channel dimension to obtain the constraint vector.

[0030] Optionally, the update process for each iteration is as follows: determine the noise information to be updated in the current iteration, which includes the first noise information or the updated noise data obtained after the update processing of the previous iteration; process the noise information to be updated in the current iteration based on the diffusion generation model to obtain the third noise information, which includes the second noise information to be adjusted in the noise information to be updated; process the third noise information and the noise information to be updated based on the iterative update model to obtain the updated noise information corresponding to the current iteration.

[0031] The noise information to be updated refers to the noise information that needs to be updated in the current iteration. The third noise information is the data obtained by processing the noise information to be updated in the current iteration through a diffusion generation model. The third noise information includes the second noise information to be adjusted within the noise information to be updated. The second noise information to be adjusted is the noise information in the noise information to be updated that needs to be adjusted. The iterative update model is the model for updating the noise information to be updated. The iterative update model includes, but is not limited to, neural network models and mathematical models. The updated noise information is the data obtained after updating in the current iteration. The updated noise information corresponding to the current iteration can be obtained by processing the third noise information and the noise information to be updated according to the iterative update model. For example, the third noise information and the noise information to be updated can be input into the iterative update model for processing to obtain the updated noise information corresponding to the current iteration.

[0032] It should be noted that the noise information to be updated in the first iteration is the first noise information; the noise information to be updated in subsequent iterations is the updated noise data obtained from the previous iteration.

[0033] It should also be noted that the first constraint condition is applied across multiple iterations. That is, in each iteration, the diffusion generation model needs to input the first constraint condition into the diffusion generation model while processing the noise information to be updated in the current iteration.

[0034] Specifically, the first noise information is used as the noise information to be updated in the first iteration round, and the updated noise data obtained from the previous iteration round is used as the noise information to be updated in the current iteration round. The noise information to be updated in the current iteration round is input into the diffusion generation model for processing to obtain the third noise information. The third noise information and the noise information to be updated are input into the iterative update model for processing to obtain the updated noise information corresponding to the current iteration round, thus realizing the update of the iteration round and providing accurate data support for subsequent iterative updates.

[0035] For example, the formula for calculating updated noise information is as follows: ; in, This represents the update noise information corresponding to the (t-1)th iteration, which is the update noise information corresponding to the current iteration. This represents the cumulative noise scheduling coefficient corresponding to the (t-1)th iteration. This represents the update noise information corresponding to the t-th iteration, which is the update noise information corresponding to the previous iteration. This represents the third noise information corresponding to the t-th iteration.

[0036] S150. Generate abnormal cigarette manufacturing process data based on the target process characteristics.

[0037] The abnormal cigarette manufacturing process data refers to the production process indicators that meet the target production batch. The defect type corresponding to the abnormal cigarette manufacturing process data is the same as the defect type of the abnormal cigarettes in the target production batch. The abnormal cigarette manufacturing process data can be processed based on the target process characteristics. For example, the target process characteristics can be input into a trained feature mapping model for processing to obtain the abnormal cigarette manufacturing process data. The feature mapping model includes, but is not limited to, neural network models, such as multilayer perceptrons.

[0038] Specifically, the target process features are input into a trained feature mapping model for processing to obtain abnormal cigarette process data. This process generates abnormal cigarette process data, which is highly consistent with the first cigarette process data of abnormal cigarettes in the target production batch. The defect types of abnormal cigarettes are consistent with the defect types of abnormal cigarettes in the target production batch, thus improving the authenticity and accuracy of the abnormal cigarette process data and providing accurate and authentic data support for subsequent analysis and processing.

[0039] The technical solution of this embodiment acquires first cigarette process data corresponding to multiple abnormal cigarettes in a target production batch. Each first cigarette process data carries a first tag, which represents the defect type of each abnormal cigarette in the target production batch. This acquisition of first cigarette process data provides comprehensive data support for subsequent analysis and processing. Based on the first cigarette process data, a first process feature vector is determined. Furthermore, the first tag is mapped to determine a first category feature vector, thus processing the first cigarette process data and the first tag, providing comprehensive and accurate data support for subsequent analysis and processing. First noise information of a defined noise type is generated, providing a data foundation for subsequent processing. Using the first process feature vector, the first category feature vector, and a preset time step as first constraints, the first noise information is processed based on a diffusion generation model to determine second noise information. The process includes the first noise information to be adjusted in the first noise information. Based on the diffusion generation model and the second noise information, the first noise information is updated in multiple iterations to obtain the target process features. The iteration rounds are determined based on the time step, which realizes the accurate determination of the target process features. This ensures that the target process features and the first process feature vector corresponding to the target production batch are consistent, improving the authenticity and accuracy of the target process features and providing accurate data support for subsequent analysis and processing. Abnormal cigarette process data is generated based on the target process features. The abnormal cigarette process data is highly consistent with the first cigarette process data of abnormal cigarettes in the target production batch. The defect types of abnormal cigarettes are consistent with the defect types of abnormal cigarettes in the target production batch, improving the authenticity and accuracy of the abnormal cigarette process data and providing accurate and authentic data support for subsequent analysis and processing.

[0040] Figure 2 This is a flowchart of a training method for a diffusion generation model provided in an embodiment of the present invention. For example... Figure 2 As shown, the method specifically includes the following steps: S210. Construct the initial diffusion generation model to be trained.

[0041] The initial diffusion generation model includes, but is not limited to, a neural network model. For example, the initial diffusion generation model may be a conditional diffusion model. Optionally, the initial diffusion generation model may include a noise addition module and a noise generation module.

[0042] S220. Obtain the second cigarette process data corresponding to multiple abnormal cigarettes in different production batches. The second cigarette process data carries a second tag, which represents the defect type of each abnormal cigarette in different production batches.

[0043] The second cigarette manufacturing process data is the dataset used for training the initial diffusion generation model. The second cigarette manufacturing process data corresponding to multiple abnormal cigarettes from different production batches constitute the training dataset. This data includes physical and chemical process data. Physical process data includes at least one of weight, draw resistance, ventilation rate, circumference, hardness, and length; chemical process data includes at least one of tar content, carbon monoxide content, and nicotine content. The second label identifies the defect type of the abnormal cigarette. Defect types include, but are not limited to, weight exceeding a preset weight range and draw resistance exceeding a preset draw resistance range. It should be noted that the second cigarette manufacturing process data in this invention includes second cigarette manufacturing process data corresponding to abnormal cigarettes with both defect types. This data can be obtained from a second cigarette manufacturing process database. For example, a preset number of second cigarette manufacturing process data corresponding to abnormal cigarettes from different production batches can be selected from the database as needed. The database can store second cigarette manufacturing process data corresponding to abnormal cigarettes from multiple production batches.

[0044] Specifically, based on the requirements, a preset number of abnormal cigarettes from the production batches are selected from the second cigarette process database to obtain the corresponding second cigarette process data, providing comprehensive data support for the subsequent training of the initial diffusion generation model.

[0045] S230. Based on the second cigarette process data corresponding to multiple cigarettes in different production batches, determine the statistical characteristics of the second cigarette process data corresponding to multiple cigarettes in each production batch.

[0046] The statistical features are information that characterizes the distribution pattern and fluctuation level of the second cigarette manufacturing process data for each production batch. Optionally, the statistical features include the mean and standard deviation. The statistical features for each production batch can be calculated by taking the mean and standard deviation of each production process indicator within that batch.

[0047] Specifically, the mean and standard deviation of each production process indicator in each production batch are statistically analyzed as the statistical characteristics of that production batch, providing comprehensive data support for subsequent analysis and processing.

[0048] S240. Based on the second cigarette process data corresponding to multiple cigarettes in different production batches, determine the second process feature vector corresponding to the second cigarette process data of multiple cigarettes in each production batch; and perform mapping processing on the second label to obtain the second category feature vector.

[0049] The second process feature vector represents the process distribution and correlation of abnormal cigarettes. The second process feature vector can be determined based on the second cigarette process data. For example, the second cigarette process data can be input into a trained second process feature vector determination model for processing to obtain the second process feature vector. The second process feature vector determination model includes, but is not limited to, a neural network model. Optionally, the determination process of the second process feature vector is as follows: construct an undirected weighted graph corresponding to the second cigarette process data based on the second cigarette process data corresponding to multiple cigarettes; extract process features from the undirected weighted graph corresponding to the second cigarette process data to obtain the second process feature vector. It should be noted that the determination process of the second process feature vector can be the same as the determination process of the first process feature vector. The second category feature vector represents the defect type of abnormal cigarettes. The second category feature vector can be determined based on the second label. For example, the second label can be mapped according to a second mapping model to obtain a binary feature vector. The second mapping model includes, but is not limited to, a neural network model. For example, the second mapping model sequentially includes a linear layer, an activation function layer, and another linear layer, where the activation function in the activation function layer is the SiLU activation function.

[0050] Specifically, the second cigarette manufacturing process data is input into a trained second process feature vector determination model for processing to obtain the second process feature vector; the second label is mapped according to the second mapping model to obtain a binary feature vector, thus realizing the processing of the second cigarette manufacturing process data and the second label, providing comprehensive and accurate data support for subsequent analysis and processing.

[0051] S250. Based on statistical features, the second process feature vector, the second category feature vector, and the time step, the model parameters of the initial diffusion generation model are adjusted to obtain a trained diffusion generation model.

[0052] The process involves adjusting the model parameters of the initial diffusion generation model. This can be achieved by progressively adding pre-defined types of noise information to the second process feature vector using the noise addition module in the initial diffusion generation model, resulting in noise-carrying second process feature vectors at different time steps. The statistical features, second process feature vectors, second category feature vectors, and time steps are then fused to obtain fusion conditions. These fusion conditions, time steps, and noise-carrying second process feature vectors are input into the noise generation module in the initial diffusion generation model for noise prediction, yielding the noise information that needs adjustment in the noise-carrying second process feature vectors. A loss function is calculated based on the added noise information and the noise information that needs adjustment in the noise-carrying second process feature vectors. The model parameters of the initial diffusion generation model are then adjusted based on the loss function until a trained diffusion generation model is obtained. This process trains the diffusion generation model and provides a model foundation for generating abnormal cigarette manufacturing process data.

[0053] Optionally, the initial diffusion generation model is adjusted based on statistical features, a second process feature vector, a second category feature vector, and a time step to obtain a trained diffusion generation model. This includes: a noise addition module adding noise to the second process feature vector based on the time step, and determining fourth noise information based on the added noise; a noise generation module processing the fourth noise information using the time step, the second process feature vector, and the second category feature vector as second constraints to obtain fifth noise information, which includes the third noise information to be adjusted from the fourth noise information; a first loss function being generated based on the fourth noise information, the fifth noise information, and statistical features, and the model parameters of the noise generation module being adjusted based on the first loss function; and the diffusion generation model being determined based on the noise generation module.

[0054] The fourth noise information is the noise information added to the second process feature vector. The second constraint is the set of conditions that constrain the initial diffusion generation model during the training phase. The second constraint includes the time step, the second process feature vector, and the second category feature vector. The fifth noise information is the data obtained by processing the noise information added to the second process feature vector through the noise generation module. The fifth noise information includes the third noise information to be adjusted in the fourth noise information. The third noise information to be adjusted is the noise information in the fourth noise information that needs to be adjusted. The fifth noise information can be determined based on the second constraint, the noise generation module, and the fourth noise information. For example, the fourth noise information and the second constraint can be input into the noise generation module for processing to obtain the fifth noise information. The first loss function is information characterizing the performance of the initial diffusion generation model. The first loss function can be determined based on the fourth noise information, the fifth noise information, and statistical features. For example, the fourth noise information, the fifth noise information, and statistical features can be input into the trained first loss function determination model for processing to obtain the first loss function. The first loss function determination model includes, but is not limited to, neural network models and mathematical models. Adjusting the model parameters of the noise generation module based on the first loss function can be done with the goal of minimizing the first loss function. Determining the diffusion generation model based on the noise generation module can be achieved by using the noise generation module with adjusted model parameters as the diffusion generation model.

[0055] Specifically, the noise addition module adds noise to the second process feature vector based on the time step, and the noise information added to the second process feature vector is used as the fourth noise information. Using the time step, the second process feature vector, and the second category feature vector as the second constraint, the fourth noise information and the second constraint are input into the noise generation module for processing to obtain the fifth noise information. The fourth noise information, the fifth noise information, and statistical features are input into the trained first loss function determination model for processing to obtain the first loss function. The model parameters of the noise generation module are adjusted with the goal of minimizing the first loss function. The noise generation module with adjusted model parameters is used as the diffusion generation model, which realizes the training of the diffusion generation model, improves the accuracy of the diffusion generation model, and provides accurate model support for subsequent analysis and processing.

[0056] Based on the above embodiments, the noise addition module adds noise to the second process feature vector based on time steps, including: gradually adding noise to the second process feature vector using a cosine scheduling strategy. For example, the calculation formula for the second category feature vector after adding noise is as follows: ; ; in, This represents the second-class feature vector after noise has been added at the current time step t; Represents the cumulative noise scheduling coefficient; This represents the feature vector of the second process; represents standard Gaussian noise; t represents the current time step; T represents the time step; s represents the scaling factor, which is preset and can be set to 0.008.

[0057] Based on the above embodiments, the noise generation module includes at least one graph convolutional network unit. The noise generation module may include one graph convolutional network unit, or it may include three graph convolutional network units. The number of graph convolutional network units in the noise generation module is selected according to requirements, and this invention is not limited thereto. For example, see [link to example]. Figure 3 , Figure 3 This is a schematic diagram of a noise generation module provided in an embodiment of the present invention. The noise generation module includes a first graph convolutional network unit, a second graph convolutional network unit, and a third graph convolutional network unit. A batch normalization layer and a SiLU activation function layer are included between the first graph convolutional network unit and the second graph convolutional network unit, and a batch normalization layer and a SiLU activation function layer are included between the second graph convolutional network unit and the third graph convolutional network unit.

[0058] Optionally, generating a first loss function based on fourth noise information, fifth noise information, and statistical features includes: generating a second loss function based on fourth noise information and fifth noise information; generating a third loss function based on mean, standard deviation, and fifth noise information; and determining the first loss function based on the second loss function and the third loss function.

[0059] The second loss function is a parameter characterizing the accuracy of the noise generation module. The second loss function includes, but is not limited to, mean squared error. The second loss function can be determined based on the fourth and fifth noise information. For example, the fourth and fifth noise information can be input into a trained model for determining the second loss function to obtain the second loss function. This model includes, but is not limited to, neural network models and mathematical models. The third loss function characterizes the deviation of the fifth noise information. The third loss function can be determined based on the mean, standard deviation, and the fifth noise information. For example, the mean, standard deviation, and the fifth noise information can be input into a trained model for determining the third loss function to obtain the third loss function. This model includes, but is not limited to, neural network models and mathematical models. The first loss function can be determined based on the second and third loss functions. For example, a weighted sum between the second and third loss functions can be calculated, and this weighted sum is used as the first loss function.

[0060] Specifically, the fourth and fifth noise information are input into the trained second loss function determination model for processing to obtain the second loss function; the mean, standard deviation, and fifth noise information are input into the trained third loss function determination model for processing to obtain the third loss function; the weighted sum between the second and third loss functions is calculated, and the weighted sum between the second and third loss functions is used as the first loss function, providing data support for adjusting the model parameters of the initial diffusion generation model.

[0061] Based on the above embodiments, a third loss function is generated based on the mean, standard deviation, and fifth noise information, including: determining a first boundary range and a second boundary range based on the mean and standard deviation; and determining the third loss function based on the first boundary range, the second boundary range, and the fifth noise information.

[0062] The first boundary range and the second boundary range represent the offset range of the fifth noise information. The first boundary range can be a soft boundary range, and correspondingly, the second boundary range can be a hard boundary range; or, the first boundary range can be a hard boundary range, and correspondingly, the second boundary range can be a soft boundary range. This embodiment of the invention uses an example where the first boundary range is a hard boundary range and the second boundary range is a soft boundary range. The mean and standard deviation are respectively input into the first boundary range determination model and the second boundary range determination model for processing to obtain the first boundary range and the second boundary range. The first boundary range determination model and the second boundary range determination model include, but are not limited to, neural network models and mathematical models. The third loss function can also be determined based on the first boundary range, the second boundary range, and the fifth noise information. For example, the first boundary range, the second boundary range, and the fifth noise information are input into a trained third loss function determination model for processing to obtain the third loss function. It should be noted that the first boundary range includes hard boundary ranges corresponding to multiple cigarette manufacturing process data, and the second boundary range includes soft boundary ranges corresponding to multiple cigarette manufacturing process data.

[0063] For example, the formula for calculating the first loss function is as follows: ; in, Represents the first loss function; Indicates the fifth noise information; Indicates the fourth noise information; This represents the second loss function, which is the mean square error between the fifth and fourth noise information. Represents the third loss function; The weight coefficients representing the third loss function can be pre-set, for example, 0.1. Third loss function The calculation formula is as follows: ; in, This represents the minimum boundary value corresponding to the i-th cigarette manufacturing data within the hard boundary range; This represents the offset of the hard boundary range of the i-th cigarette manufacturing data; This represents the maximum boundary value corresponding to the i-th cigarette manufacturing data within the hard boundary range; This represents the offset of the soft boundary range of the i-th cigarette manufacturing process data; Indicates the first boundary range; This indicates the range of the second boundary. and It can be determined based on variance, for example , . and The calculation formula is as follows: ; ; in, This represents the mean value corresponding to the process data of the i-th cigarette. This represents the standard deviation of the process data for the i-th cigarette. This represents the minimum value of the tolerance parameter corresponding to the i-th cigarette manufacturing process data; This represents the maximum value of the tolerance parameter corresponding to the i-th cigarette manufacturing process data; in an embodiment of the present invention, and Dynamic adjustments can be made using the gradient descent method.

[0064] The technical solution of this embodiment involves constructing an initial diffusion generation model to be trained; acquiring second cigarette process data corresponding to multiple abnormal cigarettes in different production batches, each second cigarette process data carrying a second label, which characterizes the defect type of each abnormal cigarette in different production batches, providing comprehensive data support for the subsequent training of the initial diffusion generation model; determining the statistical characteristics of the second cigarette process data corresponding to multiple cigarettes in each production batch based on the second cigarette process data corresponding to multiple cigarettes in different production batches, providing comprehensive data support for subsequent analysis and processing; and further determining the statistical characteristics of the second cigarette process data corresponding to multiple cigarettes in each production batch based on the second... The process data of cigarette manufacturing is used to determine the second process feature vector corresponding to the second process data of multiple cigarettes in each production batch, and to map the second label to obtain the second category feature vector. This process of processing the second cigarette manufacturing data and the second label provides comprehensive and accurate data support for subsequent analysis and processing. Based on statistical features, the second process feature vector, the second category feature vector, and the time step, the model parameters of the initial diffusion generation model are adjusted to obtain a trained diffusion generation model. This process of training the diffusion generation model improves the accuracy of the diffusion generation model and provides an accurate model foundation for the generation of abnormal cigarette manufacturing data.

[0065] Figure 4 This is a schematic diagram of a device for generating abnormal cigarette manufacturing process data according to an embodiment of the present invention. Figure 4 As shown, the device includes a cigarette manufacturing process data acquisition module 310, a feature vector determination module 320, a first noise information generation module 330, a first noise information iteration module 340, and an abnormal cigarette manufacturing process data generation module 350.

[0066] The system includes: a cigarette manufacturing process data acquisition module 310, used to acquire first cigarette manufacturing process data corresponding to multiple abnormal cigarettes in a target production batch, each first cigarette manufacturing process data carrying a first label, the first label representing the defect type of each abnormal cigarette in the target production batch; a feature vector determination module 320, used to determine a first process feature vector based on the first cigarette manufacturing process data; and to map the first label to determine a first category feature vector; a first noise information generation module 330, used to generate first noise information of a set noise type; a first noise information iteration module 340, used to process the first noise information based on a diffusion generation model with the first process feature vector, the first category feature vector, and a preset time step as first constraints, to determine second noise information, the second noise information including the first noise information to be adjusted in the first noise information; and to update the first noise information through multiple iterations based on the diffusion generation model and the second noise information to obtain target process features, the iteration rounds being determined based on the time step; and an abnormal cigarette manufacturing process data generation module 350, used to generate abnormal cigarette manufacturing process data based on the target process features.

[0067] The technical solution of this embodiment, through a cigarette process data acquisition module, acquires first cigarette process data corresponding to multiple abnormal cigarettes in a target production batch. Each first cigarette process data carries a first tag, which characterizes the defect type of each abnormal cigarette in the target production batch, thus achieving the acquisition of first cigarette process data and providing comprehensive data support for subsequent analysis and processing. Through a feature vector determination module, a first process feature vector is determined based on the first cigarette process data, and the first tag is mapped to determine a first category feature vector, thus achieving the processing of the first cigarette process data and the first tag, providing comprehensive and accurate data support for subsequent analysis and processing. Through a first noise information generation module, first noise information of a set noise type is generated. Through a first noise information iteration module, using the first process feature vector, the first category feature vector, and a preset time step as first constraints, the first noise information is processed based on a diffusion generation model to determine a second noise. The information, the second noise information, includes the first noise information to be adjusted in the first noise information. Based on the diffusion generation model and the second noise information, the first noise information is updated in multiple iterations to obtain the target process features. The iteration rounds are determined based on the time step, which realizes the accurate determination of the target process features. This ensures that the target process features and the first process feature vector corresponding to the target production batch are consistent, improving the authenticity and accuracy of the target process features and providing accurate data support for subsequent analysis and processing. Through the abnormal cigarette process data generation module, abnormal cigarette process data is generated based on the target process features. This realizes the generation of abnormal cigarette process data. The abnormal cigarette process data is highly consistent with the first cigarette process data of abnormal cigarettes in the target production batch. The defect types of abnormal cigarettes are consistent with the defect types of abnormal cigarettes in the target production batch, improving the authenticity and accuracy of the abnormal cigarette process data and providing accurate and authentic data support for subsequent analysis and processing.

[0068] Based on the above embodiments, optionally, the first noise information iteration module 340 is further configured to: determine the noise information to be updated in the current iteration round, the noise information to be updated including the first noise information or the updated noise data obtained by the update processing of the previous iteration round; process the noise information to be updated in the current iteration round based on the diffusion generation model to obtain the third noise information, the third noise information including the second noise information to be adjusted in the noise information to be updated; process the third noise information and the noise information to be updated based on the iterative update model to obtain the updated noise information corresponding to the current iteration round.

[0069] Optionally, the first cigarette manufacturing data includes physical manufacturing data and chemical manufacturing data; the physical manufacturing data includes at least one of weight, draw resistance, ventilation rate, circumference, hardness and length; the chemical manufacturing data includes at least one of tar content, carbon monoxide content and nicotine content.

[0070] Optionally, the feature vector determination module 320 is also used to: construct an undirected weighted graph based on the first cigarette process data corresponding to multiple cigarettes; and extract process features from the undirected weighted graph to obtain the first process feature vector.

[0071] Optionally, the device further includes a diffusion generation model training module, used for: constructing an initial diffusion generation model to be trained; acquiring second cigarette process data corresponding to multiple abnormal cigarettes in different production batches, wherein the second cigarette process data carries a second label, and the second label represents the defect type of each abnormal cigarette in different production batches; determining the statistical characteristics of the second cigarette process data corresponding to multiple cigarettes in each production batch based on the second cigarette process data corresponding to multiple cigarettes in different production batches; determining the second process feature vector corresponding to the second cigarette process data corresponding to multiple cigarettes in each production batch based on the second cigarette process data corresponding to multiple cigarettes in different production batches; mapping the second label to obtain a second category feature vector; and adjusting the model parameters of the initial diffusion generation model based on the statistical characteristics, the second process feature vector, the second category feature vector, and the time step to obtain a trained diffusion generation model.

[0072] Optionally, the initial diffusion generation model includes a noise addition module and a noise generation module; the diffusion generation model training module is further used for: the noise addition module adding noise to the second process feature vector based on the time step, and determining the fourth noise information based on the added noise; the noise generation module processing the fourth noise information with the time step, the second process feature vector, and the second category feature vector as the second constraint conditions to obtain the fifth noise information, which includes the third noise information to be adjusted in the fourth noise information; generating a first loss function based on the fourth noise information, the fifth noise information, and statistical features, adjusting the model parameters of the noise generation module based on the first loss function; and determining the diffusion generation model based on the noise generation module.

[0073] Optionally, statistical features include the mean and standard deviation; the diffusion generation model training module is also used to: generate a second loss function based on the fourth and fifth noise information; generate a third loss function based on the mean, standard deviation, and fifth noise information; and determine a first loss function based on the second and third loss functions.

[0074] The device for generating abnormal cigarette process data provided in the embodiments of the present invention can execute the method for generating abnormal cigarette process data provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0075] Figure 5This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0076] like Figure 5 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0077] Multiple components in electronic device 10 are connected to input / output (I / O) interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of monitors, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0078] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as methods for generating abnormal cigarette manufacturing process data.

[0079] In some embodiments, the method for generating abnormal cigarette manufacturing process data may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via read-only memory (ROM) 12 and / or communication unit 19. When the computer program is loaded into random access memory (RAM) 13 and executed by processor 11, one or more steps of the method for generating abnormal cigarette manufacturing process data described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the method for generating abnormal cigarette manufacturing process data by any other suitable means (e.g., by means of firmware).

[0080] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0081] The computer program used for generating abnormal cigarette manufacturing process data to implement the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are performed. The computer program can be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0082] This invention also provides a computer-readable storage medium storing computer instructions for causing a processor to execute a method for generating abnormal cigarette manufacturing process data, the method comprising: The process involves: acquiring first cigarette process data corresponding to multiple abnormal cigarettes in a target production batch, each carrying a first label representing the defect type of each abnormal cigarette in the target production batch; determining a first process feature vector based on the first cigarette process data; mapping the first label to determine a first category feature vector; generating first noise information of a predetermined noise type; processing the first noise information using a diffusion generation model with the first process feature vector, the first category feature vector, and a preset time step as first constraints to determine second noise information, which includes first noise information to be adjusted from the first noise information; updating the first noise information through multiple iterations based on the diffusion generation model and the second noise information to obtain target process features, with the iteration rounds determined based on the time step; and generating abnormal cigarette process data based on the target process features.

[0083] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0084] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0085] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0086] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0087] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0088] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method of generating anomalous cigarette process data, the method comprising: include: Obtain the first cigarette process data corresponding to multiple abnormal cigarettes in the target production batch. The first cigarette process data carries a first tag, and the first tag represents the defect type of each abnormal cigarette in the target production batch. The first process feature vector is determined based on the first cigarette manufacturing data; And map the first label to determine the first category feature vector; Generate first noise information of the specified noise type; Using the first process feature vector, the first category feature vector, and the preset time step as the first constraint, the first noise information is processed based on the diffusion generation model to determine the second noise information, wherein the second noise information includes the first noise information to be adjusted in the first noise information; Based on the diffusion generation model and the second noise information, the first noise information is updated in multiple iterations to obtain the target process features, wherein the iterations are determined based on the time step. The abnormal cigarette manufacturing process data is generated based on the target process characteristics.

2. The method according to claim 1, characterized in that, The update process for each iteration is as follows: Determine the noise information to be updated in the current iteration round, wherein the noise information to be updated includes the first noise information or the updated noise data obtained after the update processing of the previous iteration round; Based on the diffusion generation model, the noise information to be updated in the current iteration round is processed to obtain the third noise information, which includes the second noise information to be adjusted in the noise information to be updated. The third noise information and the noise information to be updated are processed based on the iterative update model to obtain the updated noise information corresponding to the current iteration round.

3. The method according to claim 1, characterized in that, The first cigarette manufacturing process data includes physical process data and chemical process data; The physical process data includes at least one of weight, suction resistance, ventilation rate, circumference, hardness, and length; The chemical process data includes at least one of the following: tar content, carbon monoxide content, and nicotine content.

4. The method according to claim 1, characterized in that, The step of determining the first process feature vector based on the first cigarette manufacturing process data includes: An undirected weighted graph is constructed based on the first cigarette manufacturing data corresponding to the multiple cigarettes; The process feature vector is obtained by extracting process features from the undirected weighted graph.

5. The method according to claim 1, characterized in that, The training process of the diffusion generation model is as follows: Construct an initial diffusion generation model to be trained; Obtain second cigarette process data corresponding to multiple abnormal cigarettes in different production batches. The second cigarette process data carries a second tag, which represents the defect type of each abnormal cigarette in the different production batches. Based on the second cigarette process data corresponding to multiple cigarettes in different production batches, the statistical characteristics of the second cigarette process data corresponding to multiple cigarettes in each production batch are determined. Based on the second cigarette process data corresponding to multiple cigarettes in different production batches, a second process feature vector corresponding to the second cigarette process data of multiple cigarettes in each production batch is determined; and the second label is mapped to obtain a second category feature vector. Based on the statistical features, the second process feature vector, the second category feature vector, and the time step, the model parameters of the initial diffusion generation model are adjusted to obtain the trained diffusion generation model.

6. The method according to claim 5, characterized in that, The initial diffusion generation model includes a noise addition module and a noise generation module; The step of adjusting the model parameters of the initial diffusion generation model based on the statistical features, the second process feature vector, the second category feature vector, and the time step to obtain the trained diffusion generation model includes: The noise addition module adds noise to the second process feature vector based on the time step, and determines the fourth noise information based on the added noise; The noise generation module processes the fourth noise information using the time step, the second process feature vector, and the second category feature vector as second constraints to obtain the fifth noise information, which includes the third noise information to be adjusted in the fourth noise information. A first loss function is generated based on the fourth noise information, the fifth noise information, and the statistical features; and the model parameters of the noise generation module are adjusted based on the first loss function. The diffusion generation model is determined based on the noise generation module.

7. The method according to claim 6, characterized in that, The statistical characteristics include the mean and standard deviation; The generation of the first loss function based on the fourth noise information, the fifth noise information, and the statistical features includes: A second loss function is generated based on the fourth noise information and the fifth noise information; A third loss function is generated based on the mean, the standard deviation, and the fifth noise information; The first loss function is determined based on the second loss function and the third loss function.

8. A device for generating abnormal cigarette manufacturing process data, characterized in that, include: The cigarette manufacturing process data acquisition module is used to acquire the first cigarette manufacturing process data corresponding to multiple abnormal cigarettes in the target production batch. The first cigarette manufacturing process data carries a first tag, and the first tag represents the defect type of each abnormal cigarette in the target production batch. The feature vector determination module is used to determine a first process feature vector based on the first cigarette manufacturing data; and to map the first label to determine a first category feature vector. The first noise information generation module is used to generate first noise information of a specified noise type. The first noise information iteration module is used to process the first noise information based on the diffusion generation model with the first process feature vector, the first category feature vector and the preset time step as the first constraint conditions to determine the second noise information, wherein the second noise information includes the first noise information to be adjusted in the first noise information; Based on the diffusion generation model and the second noise information, the first noise information is updated in multiple iterations to obtain the target process features, wherein the iterations are determined based on the time step. An abnormal cigarette manufacturing process data generation module is used to generate abnormal cigarette manufacturing process data based on the target process characteristics.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the method for generating abnormal cigarette manufacturing process data according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method for generating abnormal cigarette manufacturing process data as described in any one of claims 1-7.

Images