Factory execution management method and system based on new product NPI

Abstract

The invention discloses a new product NPI-based factory execution management method and system, and the method comprises the steps: collecting new product trial-production data which comprises process parameter value data, quality characteristic value data, material data and cost data; extracting the quality characteristic value data smaller than or equal to a quality characteristic threshold as optimal quality characteristic value data, and taking the optimal quality characteristic value data and the corresponding optimal process parameter value data as first data; data amplification processing is carried out on the process parameter value data to generate new process parameter value data, quality simulation prediction is carried out on the new process parameter value data to generate second data, and the second data comprises preliminary optimization process parameter value data and preliminary optimization quality characteristic value data; and according to the relationship between the first data and the second data, screening the process parameter value data closest to the first data in the second data as a new product production standard.

Description

Technical Field

[0001] The present invention relates to the technical field of factory product processing management, and in particular to a factory execution management method and system based on new product NPI. Background Art

[0002] In modern society, the manufacturing industry has ushered in a new wave of industrial revolution led by digital and intelligent technologies. Many domestic manufacturing companies have implemented information management systems such as enterprise resource planning (ERP) and manufacturing execution system (MES). However, due to objective factors such as the complexity of new product production processes and the limitations of production site conditions, information technology has not been widely used in new product manufacturing workshops. Therefore, when traditional new products are introduced into factories, the application of information and digital technologies is relatively lagging and the development speed is relatively slow.

[0003] At present, the automation and information level of product manufacturing enterprises has been rapidly improved. A large amount of production site data is stored in the information systems of enterprises and factories. As a result, the amount of internal data of enterprises and factories has exploded, but it has not been effectively converted into information and knowledge on which the enterprises rely for survival. For example, in the actual production scenario of metal casting, enterprises can now record and trace the entire process of new products, but the specific impact of the process parameters in each process on the quality of castings is not clear. The quantitative relationship between the process parameters involved in the production process of new products and the product quality characteristics has not been explored, and digital quality control and production quality control are not provided for the introduction of new products into the factory. The quality of new products cannot be predicted in advance when they are introduced into the factory. Summary of the invention

[0004] The technical problems solved by the present invention are: the quantitative relationship between the process parameters involved in the production process of new products and the product quality characteristics is not explored, digital quality control and production quality control are not provided for the new products introduced into the factory, and the quality of new products is not predicted in advance when the new products are introduced into the factory.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, a plant execution management method based on new product NPI comprises:

[0006] Step S100, collecting new product trial production data, the new product trial production data includes process parameter value data, quality characteristic value data, material data and cost data, extracting the quality characteristic value data less than or equal to the quality characteristic threshold as the optimal quality characteristic value data, and taking the optimal quality characteristic value data and the corresponding optimal process parameter value data as the first data;

[0007] Step S200: Perform data augmentation on the process parameter value data to generate new process parameter value data, and perform quality simulation prediction on the new process parameter value data to generate second data, where the second data includes preliminary optimized process parameter value data and preliminary optimized quality characteristic value data;

[0008] Step S300: According to the relationship between the first data and the second data, screen the process parameter value data in the second data that is closest to the first data as the production standard for new products.

[0009] Preferably, the new product trial production data further includes production data;

[0010] The production data includes production quantity, production cycle time, equipment efficiency, and product qualification rate;

[0011] The process parameter value data includes temperature data, equipment operation data, and processing time data;

[0012] The quality characteristic value data is the number of defective products in each batch of new products.

[0013] Preferably, step S100 specifically includes:

[0014] Step S101: Collect the original data of new product production, perform data cleaning on the original data, and delete duplicate data;

[0015] Step S102: Delete the abnormal data in the original data to generate new product trial production data;

[0016] Step S103: Extract the quality characteristic value data in the new product trial production data that is less than or equal to the quality characteristic threshold as the optimal quality characteristic value data;

[0017] Step S104: Extract the process parameter value data corresponding to the optimal quality characteristic value data as the optimal process parameter value data;

[0018] Step S105: Use the optimal process parameter value data and the optimal quality characteristic value data as the first data.

[0019] Preferably, step S200 specifically includes:

[0020] The data augmentation processing method includes a first process and a second process. The process parameter value data randomly generates augmented data through the first process, and the augmented data is processed through the second process to generate new process parameter value data;

[0021] The quality simulation prediction includes quality prediction and quality verification. Quality prediction is performed on the new process parameter value data to generate corresponding quality characteristic values, and the corresponding quality characteristic values are put into one-to-one correspondence with the new process parameter value data to form process parameter groups. The process parameter groups with the quality characteristic values less than or equal to the quality characteristic threshold are screened for quality verification, and the process parameter groups that pass the verification are used as the second data.

[0022] Preferably, step S300 specifically includes:

[0023] Step S301: Use the preliminary optimization process parameter value data of the second data as the existing sample set, and the optimal process parameter value data of the first data as the original sample set. Calculate the Euclidean distance between the existing sample set and the original sample set. The calculation expression for the Euclidean distance between the existing sample set and the original sample set is:

[0024]

[0025] where d is the Euclidean distance, x is the existing sample set, y is the original sample set, x i is the i-th data of the existing sample set, y i is the i-th data of the original sample set, and n is the number of characteristics of the process parameters;

[0026] Step S302: Calculate the average value of the Euclidean distance between the existing sample set and the original sample set to generate the average Euclidean distance;

[0027] Step S303: Sort the average Euclidean distance in ascending order, and extract the sample with the smallest average Euclidean distance as the process parameter optimization result to generate the mass production standard for new products.

[0028] Preferably, the data augmentation processing method includes the first processing and the second processing:

[0029] Extract the process parameter values and quality characteristic values from the historical records of the factory area product production as the training data;

[0030] The first processing is: Input the training data into the neural network generator G of the GAN model. The generator G learns the feature distribution of the training data and generates realistic fake data;

[0031] The second processing is: Input the training data and the fake data as input sample data into the neural network discriminator D of the GAN model, evaluate the probability that each sample data comes from the real data space, and test the authenticity of the input sample data;

[0032] Through the alternating training of the generator G and the discriminator D and the error backpropagation mechanism, the generator G and the discriminator D are continuously iterated for n times, and the data augmentation method is completed.

[0033] Preferably, the quality simulation prediction includes constructing a Transformer quality prediction model to predict the quality of the new process parameter value data, and constructing an Adaboost algorithm quality verification model to verify the quality of the process parameter group.

[0034] Preferably, constructing a Transformer quality prediction model to predict the quality of the new process parameter value data includes: initializing the weight matrix and bias term parameters of the Transformer model, inputting the process parameter values into the Transformer model, and outputting the simulated quality characteristic values as the prediction results.

[0035] Preferably, constructing an Adaboost algorithm quality verification model to verify the quality of the process parameter group includes: using the Adaboost model to construct a strong classifier by integrating multiple weak classifiers, inputting the training data into the Adaboost model, dynamically adjusting the weight distribution of each sample data in the training data to generate sample weights, the Adaboost model assigns higher weights to the sample data with larger prediction errors, the Adaboost model assigns lower weights to the sample data with accurate predictions, generating a dataset with updated weights as the input for the new round of model training, and combining multiple model components obtained through iterative training in a weighted manner to form the strong classifier, generating the quality prediction model.

[0036] In a second aspect, a factory area execution management system based on new product NPI includes a data acquisition module, a data augmentation module, a parameter prediction module, and a comparison and analysis module;

[0037] The data acquisition module is used to obtain the process parameter value data, material data, and cost data of the new product, count the number of defective products in the unit production batch as the quality characteristic value data, and obtain the first data, where the first data includes the optimal quality characteristic data and the corresponding process parameter value data;

[0038] The data augmentation module is used to perform data augmentation on the process parameter value data to obtain new process parameter value data;

[0039] The parameter prediction module is used to perform quality simulation prediction on the new process parameter value data to obtain the preliminary optimized process parameters and the corresponding quality characteristic value data after prediction. Among them, the quality simulation prediction is used to find the optimal quality characteristic value data according to the new process parameter values, and obtain the prediction analysis result information as the second data;

[0040] The comparison and analysis module is used to compare and screen the parameters of the first data obtained by the data acquisition module and the second data obtained by the prediction and analysis module, and obtain the most suitable process parameter value data in the second data as the new product quality standard.

[0041] The present invention discloses the following technical effects: By using the data amplification process on the new product trial production data collected by factory machines, deeply explore the richness of process parameters of new products, explore the quantitative relationship between each process parameter and quality characteristics, and obtain the optimized process parameters of new products to achieve precise control of production parameters. In addition, by constructing a quality simulation and prediction model of the product, the quality characteristics of product batches can be predicted, improving product quality and production efficiency, thereby further reducing the production cost of products, and providing digital quality control and production quality control for the introduction of new products into the factory. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the basic process of a factory area execution management method based on new product NPI provided by an embodiment of the present invention.

[0043] Figure 2 It is a schematic diagram of the Adaboost algorithm process in the quality verification model of a factory area execution management method based on new product NPI provided by an embodiment of the present invention.

[0044] Figure 3 It is a schematic diagram of the basic process of a factory area execution management system based on new product NPI provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be made in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments.

[0046] Embodiment 1, referring to Figures 1 - 3 the figure, an embodiment of the present invention provides a factory area execution management method based on new product NPI, including:

[0047] Step S100, collecting new product trial production data, the new product trial production data includes process parameter value data, quality characteristic value data, material data, and cost data, extracting the quality characteristic value data less than or equal to the quality characteristic threshold as the optimal quality characteristic value data, and using the optimal quality characteristic value data and the corresponding optimal process parameter value data as the first data;

[0048] Step S200: Perform data augmentation on the process parameter value data to generate new process parameter value data, and conduct quality simulation prediction on the new process parameter value data to generate second data, where the second data includes preliminary optimized process parameter value data and preliminary optimized quality characteristic value data.

[0049] Step S300: According to the relationship between the first data and the second data, screen out the process parameter value data in the second data that is closest to the first data as the production standard for new products.

[0050] The new product trial production data also includes production data.

[0051] The production data includes production quantity, production cycle time, equipment efficiency, and product pass rate.

[0052] The process parameter value data includes temperature data, equipment operation data, and processing time data.

[0053] The quality characteristic value data is the number of defective products in each batch of new products.

[0054] In this embodiment, the new product trial production data collected by factory machines is processed using data augmentation to deeply explore the richness of the process parameters of new products, explore the quantitative relationship between each process parameter and the quality characteristics, and obtain the optimized process parameters of new products to achieve precise control of production parameters. In addition, by constructing a quality simulation prediction model for products, the quality characteristics of product batches can be predicted, improving product quality and production efficiency, thereby further reducing the production cost of products and providing digital quality control and production quality control for the introduction of new products into the factory.

[0055] Step S100 specifically includes:

[0056] Step S101: Collect the original data of new product production, perform data cleaning on the original data, and delete duplicate data.

[0057] Step S102: Delete the abnormal data in the original data to generate new product trial production data.

[0058] Step S103: Extract the quality characteristic value data in the new product trial production data that is less than or equal to the quality characteristic threshold as the optimal quality characteristic value data.

[0059] Step S104: Extract the process parameter value data corresponding to the optimal quality characteristic value data as the optimal process parameter value data.

[0060] Step S105: Use the optimal process parameter value data and the optimal quality characteristic value data as the first data.

[0061] In this embodiment, the quality characteristic threshold is preferably 5. There are 8 cases of characteristic data for the process parameter values. The quality characteristic value data is the number of defective products in every 100 batches of products. The minimum value of the quality characteristic value is 5 and the maximum value is 15. The process parameter value data is the characteristic variable, and the quality characteristic value data is the target variable. By performing operations such as data cleaning on the original data, complete process parameter value data and quality characteristic value data can be extracted to provide more accurate data for quality simulation prediction.

[0062] Step S200 specifically includes:

[0063] The data augmentation processing method includes a first process and a second process. The process parameter value data undergoes the first process to randomly generate augmented data, and the augmented data undergoes the second process to generate new process parameter value data;

[0064] Quality simulation prediction includes quality prediction and quality verification. Quality prediction is performed on the new process parameter value data to generate corresponding quality characteristic values, and the corresponding quality characteristic values are paired with the new process parameter value data one by one to form a process parameter group. The process parameter groups with quality characteristic values less than or equal to the quality characteristic threshold are screened for quality verification, and the process parameter groups that pass the verification are used as the second data.

[0065] In this embodiment, the purpose of performing data augmentation processing on the process parameter value data is to expand the data set. By generating new samples, the original data set is expanded, the diversity and quantity of the samples are increased, and the success rate of optimizing the process parameter values is improved.

[0066] Step S300 specifically includes:

[0067] Step S301: Take the preliminary optimization process parameter value data of the second data as the existing sample set, and the optimal process parameter value data of the first data as the original sample set. Calculate the Euclidean distance between the existing sample set and the original sample set. The calculation expression for the Euclidean distance between the existing sample set and the original sample set is:

[0068]

[0069] where d is the Euclidean distance, x is the existing sample set, y is the original sample set, x i is the i-th data of the existing sample set, y i is the i-th data of the original sample set, and n is the number of characteristics of the process parameters;

[0070] Step S302: Calculate the average value of the Euclidean distance between the existing sample set and the original sample set to generate the average Euclidean distance;

[0071] Step S303: Sort the average Euclidean distances in ascending order, extract the sample with the smallest average Euclidean distance as the optimized result of process parameters, and generate the mass production standard for new products.

[0072] In this embodiment, the number of feature quantities of process parameters is 8. Both the existing sample set and the original sample set have 8 features. Calculate the Euclidean distances between the existing sample set and the original sample set, calculate the average Euclidean distances of the 8 features, sort the averages in ascending order, and use the first 3 pieces of data in the existing sample set as the mass production standard for new products.

[0073] The data augmentation processing method includes a first process and a second process:

[0074] Extract the process parameter values and quality characteristic values from the historical records of product production in the factory area as training data;

[0075] The first process is: Input the training data into the neural network generator G of the GAN model. The generator G learns the feature distribution of the training data and generates realistic fake data;

[0076] The second process is: Use the training data and the fake data as input sample data and input them into the neural network discriminator D of the GAN model to evaluate the probability that each sample data comes from the real data space and test the authenticity of the input sample data;

[0077] Through the alternating training and error backpropagation mechanism of the generator G and the discriminator D, continuously iterate the generator G and the discriminator D. The number of iterations is n times, and the data augmentation method is completed.

[0078] In this embodiment, set the number of iterative samples input to the generator to 4. Each iteration will use 4 real samples for training and generate 4 fake samples. Use the 4 real samples and the 4 fake samples as input data and input them into the discriminator. After receiving the input, the discriminator will output a probability value through two fully connected layers, which represents the possibility that the input data is real.

[0079] Quality simulation prediction includes constructing a Transformer quality prediction model to predict the quality of new process parameter value data, and constructing an Adaboost algorithm quality verification model to verify the quality of process parameter groups.

[0080] Constructing a Transformer quality prediction model to predict the quality of new process parameter value data includes: Initializing the weight matrix and bias term parameters of the Transformer model, inputting the process parameter values into the Transformer model, and outputting the simulated quality characteristic values as the prediction results.

[0081] In this example, the training data is split into a training set and a test set in a ratio of 8:2. The process parameter values have 8 features and belong to long sequence data. A Transformer model is constructed with a feature dimension of 8, a learning rate of 0.0001, 200 training epochs, an internal dimension of 64, 4 multi-head attention heads, 3 model layers, and an output layer dimension of 1. The training set and the test set are input into the model with the set parameters for training, and the mean squared error (MSE) loss function is set to evaluate the model performance. The calculation formula for the MSE loss function is:

[0082]

[0083] where MSE is the loss value, m is the number of training samples, and g i is the true value of the i-th sample; y i is the observed value of the i-th sample. During this period, the loss value of the training set continued to be around 5. When the number of training epochs increased to 150, the loss values of the training set and the test set were basically stable at around 0.4, and the quality prediction model training was completed. The advantage of Transformer is its parallel computing ability, which can process all elements in the sequence simultaneously and is suitable for the long sequence data in this example.

[0084] The overall flowchart of the Adaboost algorithm is as Figure 2 shown. To construct an Adaboost algorithm quality verification model for quality verification of process parameter groups: Using the Adaboost model, by integrating multiple weak classifiers, a strong classifier is constructed. The training data is input into the Adaboost model, and the weight distribution of each sample data in the training data is dynamically adjusted to generate sample weights. The Adaboost model assigns higher weights to sample data with larger prediction errors and lower weights to sample data with accurate predictions. The dataset with updated weights is generated as the input for the new round of model training. Multiple model components obtained through iterative training are combined in a weighted manner to form a strong classifier, generating a quality prediction model.

[0085] In this embodiment, the quality verification model uses the Adaboost algorithm to construct a strong classifier by integrating multiple weak classifiers. The data input into the model is the training data, and the training data is saved in the form of D = {(x1, y1), (x2, y2), …, (x N , y N )}, and the number of weak classifiers is set to M. The weight of the training data distribution is initialized as W m . Assuming there are N samples in the training data, the same weight is assigned to each sample, and the weight is 1 / N. Then the weight distribution in the initial state of the training data is Train a weak classifier \(G(x)\) using weighted training samples, and calculate the weighted error rate \(e\) of the weak classifier on the training samples. m (x), where the weighted error rate \(e\) m is calculated by the following formula: m

[0086]

[0087] Among them, \(e_m\) m is the weighted error rate of the \(m\)-th weak classifier, \(w_{mi}\) m,i is the training data distribution weight of the \(i\)-th sample in the \(m\)-th weak classifier, \(G_m(x_i)\) m (x_i i ) is the prediction result of the weak classifier \(G_m\) m on the sample \(x_i\) i , \(I(G_m(x_i)\neq y_i)\) m (x_i i )\neq y_i i ) is an indicator function, which takes the value of 1 when \(G_m(x_i)\neq y_i\) m (x_i i )\neq y_i i , and 0 otherwise. \(y_i\) i is the true label of the \(i\)-th sample;

[0088] Calculate the weight coefficient \(\alpha_m\) of the weak classifier \(G_m(x)\). The formula for calculating the weight coefficient \(\alpha_m\) of the weak classifier \(G_m(x)\) is: m (x) m , and the weight coefficient \(\alpha_m\) of the weak classifier \(G_m(x)\) m (x) m is calculated by the following formula:

[0089]

[0090] Update the weights of the next training samples according to the weight coefficients of the prediction sequence to obtain the update formula. The formula for the update formula \(w_{m + 1,i}\) is: m+1,i

[0091]

[0092] Among them, \(Z_m\) m is the normalization factor, \(G_m(x)\) is the prediction result, and \(y\) is the expected result; m

[0093] Integrate the weak classifier \(G_m(x)\) and combine the weak classifiers to obtain the final strong classifier (prediction function) \(F(x)\). The formula for the strong classifier is: m

[0094]

[0095] So far, the quality verification model training is completed. At this time, the quality verification model can perform quality verification based on the process parameter sets output by the quality prediction model, and the industrial parameter sets that pass the verification are saved as the second data, providing data reference for finding high-quality process parameters in the follow-up.

[0096] Example 2, referring to Figure 3 This is another embodiment of the present invention. Different from the first embodiment, it provides a factory execution management system based on new product NPI, including a data acquisition module, a data amplification module, a parameter prediction module, and a comparison and analysis module;

[0097] The data acquisition module is used to obtain the process parameter value data, material data, and cost data of the new product, count the number of defective products in a unit production batch as the quality characteristic value data, and obtain the first data. The first data includes the optimal quality characteristic data and the corresponding process parameter value data;

[0098] The data amplification module is used to amplify the process parameter value data to obtain new process parameter value data;

[0099] The parameter prediction module is used to perform quality simulation prediction on the new process parameter value data to obtain the preliminary optimized process parameters and the corresponding quality characteristic value data after prediction. Among them, the quality simulation prediction is used to find the optimal quality characteristic value data according to the new process parameter value, and obtain the prediction analysis result information as the second data;

[0100] The comparison and analysis module is used to compare and screen the parameters of the first data obtained by the data acquisition module and the second data obtained by the prediction analysis module, and obtain the most conforming process parameter value data in the second data as the new product quality standard.

[0101] In this embodiment, through the further description of the factory execution management system for new product NPI, it is divided into a data acquisition module, a data amplification module, a parameter prediction module, and a comparison and analysis module. These modules can more accurately represent the system operation tasks, clearly represent the roles of the first data and the second data obtained in the data processing, and provide a clear direction for the introduction of new products into factory production.

[0102] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disc. These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device, and the instruction device implements the functions specified in one process Figure 1 one process or multiple processes and / or boxes Figure 1 or the functions specified in multiple boxes.

[0103] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A factory area execution management method based on new product NPI, characterized in that, Including: Step S100: Collect new product trial production data, where the new product trial production data includes process parameter value data, quality characteristic value data, material data, and cost data. Extract the quality characteristic value data that is less than or equal to the quality characteristic threshold as the optimal quality characteristic value data, and use the optimal quality characteristic value data and the corresponding optimal process parameter value data as the first data. Step S200: Perform data augmentation processing on the process parameter value data to generate new process parameter value data. Conduct quality simulation prediction on the new process parameter value data to generate the second data, where the second data includes preliminarily optimized process parameter value data and preliminarily optimized quality characteristic value data. Step S300: According to the relationship between the first data and the second data, screen the process parameter value data in the second data that is closest to the first data as the new product production standard.

2. The method for plant execution management based on new product NPI according to claim 1, wherein: The new product trial production data further includes production data. The production data includes production quantity, production cycle time, equipment efficiency, and product qualification rate. The process parameter value data includes temperature data, equipment operation data, and processing time data. The quality characteristic value data is the number of defective products in each batch of new products.

3. The method for plant execution management based on new product NPI according to claim 2, wherein Step S100 specifically includes: Step S101: Collect the original data of new product production, perform data cleaning on the original data, and delete duplicate data. Step S102: Delete the abnormal data in the original data to generate new product trial production data. Step S103: Extract the quality characteristic value data that is less than or equal to the quality characteristic threshold in the new product trial production data as the optimal quality characteristic value data. Step S104: Extract the process parameter value data corresponding to the optimal quality characteristic value data as the optimal process parameter value data. Step S105: Use the optimal process parameter value data and the optimal quality characteristic value data as the first data.

4. A plant execution management method based on new product NPI according to claim 3, characterized in that, Step S200 specifically includes: The data augmentation processing method includes the first processing and the second processing. The process parameter value data generates augmented data through the first processing, and the augmented data is subjected to the second processing to generate new process parameter value data. The quality simulation prediction includes quality prediction and quality verification. Perform quality prediction on the new process parameter value data to generate corresponding quality characteristic values, and pair the corresponding quality characteristic values with the new process parameter value data one by one to form process parameter groups. Screen the process parameter groups whose quality characteristic values are less than or equal to the quality characteristic threshold for quality verification, and use the process parameter groups that pass the verification as the second data.

5. A factory area execution management method based on new product NPI according to claim 4, characterized in that, Step S300 specifically includes: Step S301: Use the preliminarily optimized process parameter value data of the second data as the existing sample set, and the optimal process parameter value data of the first data as the original sample set. Calculate the Euclidean distance between the existing sample set and the original sample set. The calculation expression for the Euclidean distance between the existing sample set and the original sample set is: where d is the Euclidean distance, x is the existing sample set, y is the original sample set, x i is the i-th data of the existing sample set, y i is the i-th data of the original sample set, and n is the number of features of the process parameters; Step S302: Calculate the average value of the Euclidean distance between the existing sample set and the original sample set to generate the average Euclidean distance. Step S303: Sort the average Euclidean distances in ascending order, extract the sample with the smallest average Euclidean distance as the optimized result of the process parameters, and produce the mass production standard for new products.

6. The plant execution management method based on new product NPI according to claim 5, characterized in that: The data augmentation processing method includes a first process and a second process: Extract the process parameter values and quality characteristic values from the historical records of product production in the factory area as training data; The first process is: input the training data into the neural network generator G of the GAN model, and the generator G learns the feature distribution of the training data and generates realistic fake data; The second process is: input the training data and the fake data as input sample data into the neural network discriminator D of the GAN model, evaluate the probability that each sample data comes from the real data space, and test the authenticity of the input sample data; Through the alternating training and error backpropagation mechanism of the generator G and the discriminator D, continuously iterate the generator G and the discriminator D for n times, and the data augmentation method is completed.

7. The factory area execution management method based on new product NPI according to claim 6, characterized in that: The quality simulation prediction includes constructing a Transformer quality prediction model to predict the quality of the new process parameter value data, and constructing an Adaboost algorithm quality verification model to verify the quality of the process parameter group.

8. The plant execution management method based on new product NPI according to claim 7, characterized in that: Constructing a Transformer quality prediction model to predict the quality of the new process parameter value data includes: initializing the weight matrix and bias term parameters of the Transformer model, inputting the process parameter values into the Transformer model, and outputting the simulated quality characteristic values as the prediction results.

9. A factory area execution management method based on new product NPI according to claim 8, characterized in that: Constructing an Adaboost algorithm quality verification model to verify the quality of the process parameter group includes: using the Adaboost model, by integrating multiple weak classifiers, constructing a strong classifier, inputting the training data into the Adaboost model, dynamically adjusting the weight distribution for each sample data of the training data to generate sample weights, the Adaboost model assigns higher weights to the sample data with larger prediction errors, the Adaboost model assigns lower weights to the sample data with accurate predictions, generates a dataset with updated weights as the input for the new round of model training, and combines multiple model components obtained through iterative training in a weighted manner to form the strong classifier, generating the quality prediction model.

10. A factory area execution management system based on new product NPI, which is implemented based on the factory area execution management method based on new product NPI described in any one of claims 1-9, and is characterized in that, It includes a data acquisition module, a data augmentation module, a parameter prediction module, and a comparative analysis module; The data acquisition module is used to obtain the process parameter value data, material data, and cost data of new products, count the number of defective products in the unit production batch as the quality characteristic value data, and obtain the first data, where the first data includes the optimal quality characteristic data and the corresponding process parameter value data; The data augmentation module is used to perform data augmentation on the process parameter value data to obtain new process parameter value data; The parameter prediction module is used to perform quality simulation prediction on the new process parameter value data, and obtain the preliminarily optimized process parameters and corresponding quality characteristic value data after prediction. Among them, the quality simulation prediction is used to find the optimal quality characteristic value data according to the new process parameter values, and obtain the prediction analysis result information as the second data; The comparison and analysis module is used to compare and screen the parameters of the first data obtained by the data acquisition module and the second data obtained by the prediction analysis module, and obtain the process parameter value data that best meets the requirements in the second data as the new product quality standard.

Images