Production quality intelligent management method based on AI training

Abstract

The invention discloses an intelligent production quality management method based on AI training, and particularly relates to the technical field of production quality management.The method comprises the steps that production quality management parameters required by regulations in the production process are stored and updated in real time, and a target production quality management database is constructed; then, production quality management parameters required in the production process are collected, a first target data set is obtained and processed to obtain a second target data set, a third target data set is obtained based on the second target data set, and the third target data set is used for training the AI model to obtain a second target data set; after the model training verification is passed, the production quality management data collected in real time is input into the AI model, and a prediction result is output; and finally carrying out man-machine interaction on the prediction result and the early warning signal. According to the method, a large amount of production quality related data is learned through the AI model, and the overall production quality management and control level is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of production quality management, and specifically to an intelligent production quality management method based on AI training. Background Art

[0002] In modern manufacturing, production quality management is the key to ensuring product compliance, improving customer satisfaction and reducing production costs. With the rapid development of artificial intelligence technology, AI models have advantages in data processing and analysis. By learning from historical data, they can predict quality problems that may occur in the production process.

[0003] With the rapid development of the manufacturing industry, production quality management has become increasingly important. Traditional production quality management methods mainly rely on manual inspection and statistics, which have problems such as low efficiency, poor accuracy, and slow response speed; key parameters in the production process are monitored through statistical methods (such as control charts) to ensure that they are within a controllable range; through sampling inspection, some products are randomly selected from the production batch for inspection to infer the quality of the entire batch of products.

[0004] However, traditional production quality management methods have many limitations in practical applications. For example, manual experience is not only inefficient, but also easily affected by subjective factors of inspectors, resulting in deviations in test results. Statistical methods cannot monitor the entire production process in real time and comprehensively, and there is a lag. Through sampling inspection, the sampling results may not fully represent the quality of the entire batch of products, which may lead to unqualified products entering the market. Moreover, the ability to analyze the large amount of data accumulated by enterprises in the production process is limited. Therefore, traditional production quality control methods can no longer meet the requirements of modern manufacturing for high-quality and high-efficiency production. There is an urgent need for an intelligent production quality management method based on AI training to improve the accuracy, real-time and intelligence level of production quality control. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an intelligent production quality management method based on AI training to solve the problems of low work efficiency, poor accuracy, limited data analysis capabilities, etc. raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a production quality intelligent management method based on AI training, comprising: S1: Based on the production management regulations of the target enterprise, the production quality management parameters required by the production process are stored and updated in real time through big data technology. The production quality management parameters include raw material management parameters, production process management parameters, equipment operation status management parameters and production environment management parameters, and the target production quality management database is constructed; S2: Based on S1, the required production quality management parameters in the production process are collected through sensor technology to obtain the first target data set for production quality management of each product; S3: Based on S2, the first target data set of each product production quality management is processed by feature engineering technology to obtain the second target data set of each product production quality management, and then the second target data set is screened to obtain the third target data set of each product production quality management; The second target data set for production quality management of each product in S3 includes second raw material management parameters, second production process management parameters, second equipment operation status management parameters and second production environment management parameters; The second raw material management parameter: The first raw material management parameter in the first target data set of each product production quality management is processed by feature engineering technology to obtain the raw material composition fluctuation C var , , N represents the number of times the raw material components are collected within the preset management cycle, c j represents the component value collected for the jth time, μ(c) represents the average value of the component value; the raw material batch difference B is obtained diff , , M represents the number of types in different batches, c I Indicates the composition value of batch I, c ref Indicates the reference value; obtains the raw material storage condition fluctuation S con , , T j and H j Represent the temperature and humidity collected for the jth time, T tar and H tar respectively represent reference values ​​of temperature and humidity; then obtain the second raw material management parameters, including raw material composition fluctuation, raw material batch difference and raw material storage condition fluctuation; S4: Based on S3, the third target data set of each product production quality management is input into the AI ​​model for training, and then the model is verified, and finally the verified AI model is output; S5: Through sensor technology, the production quality management data generated in the target production process collected in real time is input into the AI ​​model verified in S4, and the prediction results are output. The prediction results include the production product quality label number, 1 indicates qualified, 0 indicates unqualified, and an early warning is issued based on the result of the label number being 0; S6: Evaluate the prediction results and warning signals of S5, and conduct human-computer interaction on the results that are evaluated as abnormal.

[0007] Technical effects and advantages of the present invention: 1. The present invention uses sensor technology to deploy raw material composition detection equipment, process data collection equipment, equipment operation status data collection equipment and environmental data collection equipment in the production process to collect corresponding data respectively; then through the powerful data analysis capabilities of AI, accurate and comprehensive data support is provided for the subsequent training of AI models; 2. The present invention uses AI models to learn a large amount of production quality-related data, which can accurately identify product quality problems, avoid the subjectivity and errors of manual inspection, and improve the accuracy and consistency of quality inspection; and monitor the production process in real time, promptly discover potential quality problems and issue early warnings, reduce the production of unqualified products, reduce production costs, and improve work efficiency; 3. The present invention uses AI technology to conduct in-depth analysis of quality problems, which can quickly and accurately identify the root causes of quality problems, achieve continuous improvement of production processes, provide strong support for solving quality problems, and then help enterprises optimize production processes, improve production efficiency, and enhance the overall production quality control level. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 It is a schematic diagram of the overall process of the present invention.

[0009] Figure 2 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0010] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0011] See also Figure 1 As shown, the present invention provides a production quality intelligent management system based on AI training, including a target production quality management database, a production quality management data acquisition module, a production quality management data feature extraction module, an AI model training module, a production quality intelligent management module and a production quality intelligent management human-computer interaction module.

[0012] The target production quality management database is connected to the production quality management data acquisition module, the production quality management data feature extraction module is respectively connected to the production quality management data acquisition module and the AI ​​model training module, and the production quality intelligent management module is respectively connected to the AI ​​model training module and the production quality intelligent management human-computer interaction module.

[0013] Target production quality management database: Based on the production management regulations of the target enterprise, through big data technology, the production quality management parameters required in the production process are stored and updated in real time, including raw material management parameters, production process management parameters, equipment operation status management parameters and production environment management parameters, to build a target production quality management database; Production quality management data acquisition module: Through sensor technology, the production quality management parameters required in the production process are collected, and the collected first target data set of production quality management of each product is transmitted to the production quality management data feature extraction module; Production quality management data feature extraction module: Through feature engineering technology, the first target data set of production quality management of each product is processed to obtain the second target data set of production quality management of each product, and then the second target data set is screened to obtain the third target data set of production quality management of each product, and transmitted to the AI ​​model training module; AI model training module: used to input the third target data set of each product production quality management into the AI ​​model for training, then verify the model, and transmit the verified AI model to the production quality intelligent management module; Production quality intelligent management module: used to input the production quality management data generated in the target production process collected in real time into the verified AI model, produce prediction results, and issue an early warning to the production quality intelligent management human-computer interaction module if the AI ​​model predicts unqualified results; Production quality intelligent management human-computer interaction module: Evaluate the prediction results and early warning signals of the production quality intelligent management module, and conduct human-computer interaction on the results evaluated as abnormal.

[0014] See also Figure 2As shown, an intelligent production quality management method based on AI training includes the following steps: S1: Based on the production management regulations of the target enterprise, the production quality management parameters required by the production process are stored and updated in real time through big data technology. The production quality management parameters include raw material management parameters, production process management parameters, equipment operation status management parameters and production environment management parameters, and a target production quality management database is constructed. S2: Based on S1, the production quality management parameters required by the production process are collected through sensor technology to obtain the first target data set for production quality management of each product. S3: Based on S2, the production quality of each product is collected through feature engineering technology. Manage the first target data set for processing to obtain the second target data set for production quality management of each product, and then screen the second target data set to obtain the third target data set for production quality management of each product. S4: Based on S3, input the third target data set for production quality management of each product into the AI ​​model for training, and then verify the model, and finally output the verified AI model. S5: Through sensor technology, the production quality management data generated in the target production process collected in real time is input into the verified AI model in S4, and the prediction results are output. S6: Evaluate the prediction results and warning signals of S5, and conduct human-computer interaction for the results evaluated as abnormal.

[0015] S1: Based on the production management regulations of the target enterprise, the production quality management parameters required by the production process are stored and updated in real time through big data technology. The production quality management parameters include raw material management parameters, production process management parameters, equipment operation status management parameters and production environment management parameters, and the target production quality management database is constructed; What needs to be specifically explained in this embodiment is the production management regulations of the target automobile manufacturing enterprise. The raw material management parameters required in the production process include raw material composition, raw material batches, raw material storage conditions, etc.; production process management parameters include temperature, pressure, speed, time, etc.; equipment operation status management parameters include equipment status (such as operation, shutdown, failure), equipment parameters (such as current, voltage, vibration frequency, temperature) and production environment management parameters include temperature and humidity, air quality, light intensity, noise level, etc.

[0016] S2: Based on S1, the production quality management parameters required in the production process are collected through sensor technology to obtain the first target data set for production quality management of each product. The collection process includes the following steps: S2.1: Sensor deployment: In the target production process, component detection equipment (such as component analyzer) shall be installed at the raw material inlet according to regulations to collect raw material component data; process data collection equipment (such as temperature sensor, pressure sensor, speed sensor, etc.) and equipment operation status data collection equipment (such as voltage and current sensor, control center recorder equipment in shutdown state, fault state, etc.) shall be installed on each production equipment to collect production process management parameters and equipment operation status management parameters respectively; environmental data collection equipment (such as temperature and humidity sensor) shall be installed in the production area to collect production environment management parameters; S2.2: In the target historical production data, the production data of each production product is collected from each sensor according to the preset collection frequency (for example, the process management parameters are collected once every 1 minute, and the raw material management parameters are collected once every 5 minutes), and the first target data set for production quality management of each product is obtained. The first target management data set includes the first raw material management parameter, the first production process management parameter, the first equipment operation status management parameter and the first production environment management parameter of each production product; What needs to be specifically explained in this embodiment is that the first raw material management parameters include raw material composition, raw material batches, and raw material storage conditions, and the raw material storage conditions include storage temperature and humidity; the first production process management parameters include the types of process parameters and the data of each type; the first equipment operation status management parameters include the type of equipment, equipment operation time, equipment failure time, number of equipment failures, equipment energy consumption, and equipment maintenance interval; the first production environment management parameters include production environment temperature and humidity, PM2.5 and VOC (volatile organic compound concentration) concentrations, light intensity, and noise value.

[0017] It should be specifically noted in this embodiment that the collected data has been cleaned and screened to remove noise and outliers, so as to improve the quality and availability of the data.

[0018] S2.3: Obtain the target production data collection management index Q through big data analysis technology co , , C i It represents the coverage rate of the i-th type of data. The data coverage rate = the actual amount of data collected / the theoretical maximum amount of data. i represents the sampling rate of the i-th type of data, the data sampling rate = actual acquisition frequency / theoretical maximum acquisition frequency, i = 1 represents the first raw material management parameter, i = 2 represents the first production process management parameter, i = 3 represents the first equipment operation status management parameter and i = 4 represents the first production environment management parameter. The target production data acquisition management index Q co Compare with the threshold value, if it is less than the threshold value, repeat S2.2, otherwise it means that the target production data collection management is good; S3: Based on S2, the first target data set of each product production quality management is processed by feature engineering technology to obtain the second target data set of each product production quality management, the second target data set includes the second raw material management parameter, the second production process management parameter, the second equipment operation status management parameter and the second production environment management parameter, and then the second target data set is screened to obtain the third target data set of each product production quality management, including the following steps: S3.1: Through feature engineering technology, the first raw material management parameter in the first target data set of each product production quality management is processed to obtain the raw material composition fluctuation C var , , N represents the number of times the raw material components are collected within the preset management cycle, c j represents the component value collected for the jth time, μ(c) represents the average value of the component value; the raw material batch difference B is obtained diff , , M represents the number of types of different batches, such as the 2025-1-23 batch and the 2025-1-24 batch, c I Indicates the composition value of batch I, c ref Indicates the reference value; obtains the raw material storage condition fluctuation S con , , T j and H j Represent the temperature and humidity collected for the jth time, T tar and H tar respectively represent reference values ​​of temperature and humidity; then obtain the second raw material management parameters, including raw material composition fluctuation, raw material batch difference and raw material storage condition fluctuation; S3.2: Through feature engineering technology, the first production process management parameter in the first target data set of each product production quality management is processed to obtain the fluctuation P of each process parameter. var , , x j represents the process parameter value collected for the jth time, μ(x) represents the average value of the process parameter; the deviation P of each process parameter is obtained dev , , x j and x tar Respectively represent the process parameter value and reference value collected for the jth time; then obtain the second production process management parameters, including the fluctuation of each process parameter and the deviation of each process parameter; S3.3: Through feature engineering technology, the first equipment operation status management parameter in the first target data set of each product production quality management is processed to obtain the cumulative operation time t of each equipment within the preset management cycle. run; Get the failure frequency f of each device freq , f freq =n_g / (t g +t run ), n_g represents the number of equipment failures, t g Indicates the failure time; obtain the energy consumption fluctuation E of each device var , , e j represents the equipment energy consumption value collected for the jth time. The equipment energy consumption value refers to the electric energy obtained from the equipment current, voltage and running time. μ(e) represents the average value of the equipment energy consumption. The maintenance interval M of each equipment is obtained. int , , Δt m represents the time interval of the mth equipment maintenance, and J represents the number of times the equipment maintenance interval is collected; then the second equipment operation status management parameters are obtained, including the cumulative operation time of each equipment, the failure frequency of each equipment, the energy consumption fluctuation of each equipment, and the maintenance interval of each equipment; S3.4: Through feature engineering technology, the first production environment management parameter in the first target data set of each product production quality management is processed to obtain the production environment temperature and humidity fluctuation PE var , , T_h j and H_h j Respectively represent the temperature and humidity of the jth collection of the production environment, T_h p and H_h p Respectively represent the average value of temperature collection and the average value of humidity collection; get the air quality index A qua , , VOC i Indicates the concentration collected for the i-th time, PM2.5 i represents the concentration collected for the i-th time; the light noise intensity fluctuation L is obtained var , , l j represents the light intensity collected for the jth time, μ(l) represents the average light intensity; the average noise level N is obtained lev , , n i represents the noise value collected for the i-th time; then the second production environment management parameters are obtained, including the temperature and humidity fluctuation of the production environment, the air quality index, the light noise intensity fluctuation and the average noise level; S3.5: Feature verification: First, screen the second target data set for production quality management of each product, and use the existing correlation analysis method (such as Pearson correlation coefficient, Spearman rank correlation coefficient, chi-square test, etc.) to calculate the correlation coefficient value between each feature in each type of parameter and the corresponding product production quality label number, wherein the qualified production quality label is number 1 and the unqualified one is number 0; then compare the correlation coefficient value of each feature in each type of parameter with the correlation coefficient threshold corresponding to each type of parameter, and if it is less than the threshold, remove the feature; then update the features in such parameters, otherwise there is no removal operation, and obtain the third target data set for production quality management of each product, including the third raw material management parameters, the third production process management parameters, the third equipment operation status management parameters and the third production environment management parameters. For example, the correlation coefficient between raw material composition fluctuation and production quality is 0.65, the correlation coefficient between raw material batch difference and production quality is 0.50, the correlation coefficient between raw material storage condition fluctuation and production quality is 0.2, and the parameter threshold corresponding to the raw material parameter is 0.3. If the correlation coefficient between raw material storage condition fluctuation and production quality is less than the threshold, it will be removed; S4: Based on S3, the third target data set of each product production quality management is input into the AI ​​model for training, and then the model is verified, and finally the verified AI model is output. The AI ​​model training includes the following steps: S4.1: Construct sample data set: Match the third target data set of production quality management of each product with the corresponding quality label to obtain each sample data, each sample data includes the third target data set of production quality management of each product and the corresponding quality label; then construct a sample data set from each sample data, and divide the constructed sample data set into a training data set n_tra and a test data set n_tes. Usually, the training data set accounts for 70%, the test data set accounts for 20%, and the verification data set accounts for 10%; finally, the sample data construction quality index SDI is obtained through big data analysis technology. , ba represents the balance of quality labels, that is, the ratio of qualified labels to unqualified labels. If SDI is less than the threshold, repeat step S4.1, otherwise go to S4.2; S4.2: Input the training data set in the sample data set constructed in S4.1 into the AI ​​model (such as convolutional neural network CNN, recurrent neural network, etc. RNN) for training, calculate the output of the model through forward propagation, and compare it with the actual quality label; then use the mean square error loss function to calculate the error L(θ), , M represents the number of training samples, f(Np)I represents the output result, f(Np) 0I Represents the true result; then adjust the θ value through the gradient descent algorithm to solve the minimum value of the loss function, and then monitor the convergence of the model. Its convergence model is: , where K represents the number of model iterations, α is the model learning rate, and when the value of the loss function remains unchanged, the training is completed; finally, the trained AI model is output, and the parameters of the AI ​​model are recorded, including the network structure, the number of model iterations, and the model learning rate; S4.3: Verify the AI ​​model trained by S4.2 through the test set, obtain the verification accuracy PRE and recall REC of the test set, and obtain the F1 score of the test set. If the F1 score is greater than the threshold, it means that the AI ​​model has passed the verification. Otherwise, repeat S4.2. The accuracy rate indicates the ratio of the number of samples correctly predicted by the model to the total number of samples in the test set. The recall rate indicates the ratio of the number of samples predicted by the model to the number of samples that are actually positive. What needs to be specifically explained in this embodiment is that the network structure of the AI ​​model includes the number of network layers, the number of nodes in each layer, the activation function, the loss function, etc.; for CNN, there are also designed convolutional layers, pooling layers, fully connected layers, etc.; for RNN, there are also recurrent layers, attention mechanisms, etc.

[0019] S5: Through sensor technology, the production quality management data generated in the target production process collected in real time is input into the AI ​​model verified in S4, and the prediction results are output. The prediction results include the production product quality label number, 1 indicates qualified, 0 indicates unqualified, and an early warning is issued based on the result of the label number being 0; S6: Evaluate the prediction results and warning signals of S5, and conduct human-computer interaction for the results evaluated as abnormal. First, obtain the number of products n_s, the total number Tn_s, the number of correct warnings n_al, and the total number of warnings Tn_al whose AI model prediction results are consistent with the actual quality labels, and obtain the production quality intelligent management index PQMI. , a1 and a2 represent the corresponding weight coefficients, for example, a1=0.7 and a2=0.3; then compare the production quality intelligent management index PQMI with the threshold. If it is greater than or equal to the threshold, it means that the production quality intelligent management is good, otherwise it means that the production quality intelligent management is abnormal, reminding the production management personnel to pay attention and take corresponding measures. For example, if it is found that the model predicts inaccurately in some cases or there are other problems, timely feedback can be given to the model developers; if it is found that it is a problem with the production management data, feedback will be given to the production management personnel to adjust the process parameters, replace equipment parts, etc.

[0020] Secondly: In the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other; Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A production quality intelligent management method based on AI training, characterized by: include: S1: Based on the production management regulations of the target enterprise, the production quality management parameters required by the production process are stored and updated in real time through big data technology. The production quality management parameters include raw material management parameters, production process management parameters, equipment operation status management parameters and production environment management parameters, and the target production quality management database is constructed; S2: Based on S1, the required production quality management parameters in the production process are collected through sensor technology to obtain the first target data set for production quality management of each product; S3: Based on S2, the first target data set of each product production quality management is processed by feature engineering technology to obtain the second target data set of each product production quality management, and then the second target data set is screened to obtain the third target data set of each product production quality management; The second target data set for production quality management of each product in S3 includes second raw material management parameters, second production process management parameters, second equipment operation status management parameters and second production environment management parameters; The second raw material management parameter: The first raw material management parameter in the first target data set of each product production quality management is processed by feature engineering technology to obtain the raw material composition fluctuation C var , , N represents the number of times the raw material components are collected within the preset management cycle, c j represents the component value collected for the jth time, μ(c) represents the average value of the component value; the raw material batch difference B is obtained diff , , M represents the number of types in different batches, c I Indicates the composition value of batch I, c ref Indicates reference value; Get the raw material storage condition fluctuation S con , , T j and H j Represent the temperature and humidity collected for the jth time, T tar and H tar respectively represent reference values ​​of temperature and humidity; then obtain the second raw material management parameters, including raw material composition fluctuation, raw material batch difference and raw material storage condition fluctuation; S4: Based on S3, the third target data set of each product production quality management is input into the AI ​​model for training, and then the model is verified, and finally the verified AI model is output; S5: Through sensor technology, the production quality management data generated in the target production process collected in real time is input into the AI ​​model verified in S4, and the prediction results are output. The prediction results include the production product quality label number, 1 indicates qualified, 0 indicates unqualified, and an early warning is issued based on the result of the label number being 0; S6: Evaluate the prediction results and warning signals of S5, and conduct human-computer interaction on the results that are evaluated as abnormal.

2. According to claim 1, a production quality intelligent management method based on AI training is characterized by: Obtaining the first target data set for production quality management of each product in S2 comprises the following steps: S2.1: Sensor deployment: In the target production process, install component detection equipment at the raw material inlet according to regulations to collect raw material component data; install process data collection equipment and equipment operation status data collection equipment on each production equipment to collect production process management parameters and equipment operation status management parameters respectively; install environmental data collection equipment in the production area to collect production environment management parameters; S2.2: In the target historical production data, the production data of each production product is collected from each sensor according to the preset collection frequency to obtain a first target data set for production quality management of each product, wherein the first target management data set includes a first raw material management parameter, a first production process management parameter, a first equipment operation status management parameter, and a first production environment management parameter of each production product; S2.3: Obtain the target production data collection management index Q through big data analysis technology co , , C i It represents the coverage rate of the i-th type of data. The data coverage rate = the actual amount of data collected / the theoretical maximum amount of data. i represents the sampling rate of the i-th type of data, the data sampling rate = actual acquisition frequency / theoretical maximum acquisition frequency, i = 1 represents the first raw material management parameter, i = 2 represents the first production process management parameter, i = 3 represents the first equipment operation status management parameter and i = 4 represents the first production environment management parameter. The target production data acquisition management index Q co Compare with the threshold value, if it is less than the threshold value, repeat S2.2, otherwise it means that the target production data collection management is good.

3. According to claim 1, a production quality intelligent management method based on AI training is characterized in that: The second production process management parameter in S3: The first production process management parameter in the first target data set of each product production quality management is processed by feature engineering technology to obtain the fluctuation P of each process parameter. var , , x j represents the process parameter value collected for the jth time, μ(x) represents the average value of the process parameter; the deviation P of each process parameter is obtained dev , , x j and x tar They respectively represent the process parameter value and reference value collected for the jth time; then the second production process management parameters are obtained, including the fluctuation of each process parameter and the deviation of each process parameter.

4. According to claim 1, a production quality intelligent management method based on AI training is characterized in that: The second equipment operation status management parameter in S3: The first equipment operation status management parameter in the first target data set of each product production quality management is processed by feature engineering technology to obtain the cumulative operation time t of each equipment in the preset management cycle. run ; Get the failure frequency f of each device freq , f freq =n_g / (t g +t run ), n_g represents the number of equipment failures, t g Indicates the fault time; Get the energy consumption fluctuation E of each device var , , e j represents the equipment energy consumption value collected for the jth time. The equipment energy consumption value refers to the electric energy obtained from the equipment current, voltage and running time. μ(e) represents the average value of the equipment energy consumption. The maintenance interval M of each equipment is obtained. int , , Δt m represents the time interval of the mth equipment maintenance, and J represents the number of times the equipment maintenance interval is collected; then the second equipment operation status management parameters are obtained, including the cumulative operation time of each equipment, the failure frequency of each equipment, the energy consumption fluctuation of each equipment and the maintenance interval of each equipment.

5. According to claim 1, a production quality intelligent management method based on AI training is characterized in that: The second production environment management parameter in S3: The first production environment management parameter in the first target data set of each product production quality management is processed by feature engineering technology to obtain the production environment temperature and humidity fluctuation PE var , , T_h j and H_h j Respectively represent the temperature and humidity of the jth collection of the production environment, T_h p and H_h p Respectively represent the average value of temperature collection and the average value of humidity collection; get the air quality index A qua , , VOC i Indicates the concentration collected for the i-th time, PM2.5 i represents the concentration collected for the i-th time; the light noise intensity fluctuation L is obtained var , , l j represents the light intensity collected for the jth time, μ(l) represents the average light intensity; the average noise level N is obtained lev , , n i represents the noise value collected for the i-th time; then the second production environment management parameters are obtained, including the temperature and humidity fluctuation of the production environment, the air quality index, the light noise intensity fluctuation and the average noise level.

6. The method for intelligent production quality management based on AI training according to claim 1, characterized in that: The third target data set of each product production quality management is obtained in S3: first, the second target data set of each product production quality management is screened, and the correlation coefficient value between each feature in each type of parameter and the corresponding product production quality label number is calculated using the existing correlation analysis method, and the production quality label is 1 if it is qualified and 0 if it is unqualified; then the correlation coefficient value of each feature in each type of parameter is compared with the correlation coefficient threshold value corresponding to each type of parameter, and if it is less than the threshold, the feature is eliminated; Then the features in such parameters are updated, otherwise there is no elimination operation, and the third target data set for production quality management of each product is obtained, including third raw material management parameters, third production process management parameters, third equipment operation status management parameters and third production environment management parameters.

7. The method for intelligent production quality management based on AI training according to claim 1, characterized in that: The AI ​​model verified in S4 includes the following steps: S4.1: Construct sample data set: Match the third target data set of production quality management of each product with the corresponding quality label to obtain each sample data, each sample data includes the third target data set of production quality management of each product and the corresponding quality label; then construct a sample data set from each sample data, and divide the constructed sample data set into a training data set n_tra and a test data set n_tes; finally, obtain the sample data construction quality index SDI through big data analysis technology, , ba represents the balance of quality labels, that is, the ratio of qualified labels to unqualified labels. If SDI is less than the threshold, repeat step S4.1, otherwise go to S4.2; S4.2: Input the training data set in the sample data set constructed in S4.1 into the AI ​​model for training, calculate the output of the model through forward propagation, and compare it with the real quality label; then use the mean square error loss function to calculate the error L(θ), , M represents the number of training samples, f(Np)I represents the output result, f(Np) 0I Represents the true result; then adjust the θ value through the gradient descent algorithm to solve the minimum value of the loss function, and then monitor the convergence of the model. Its convergence model is: , where K represents the number of model iterations, α is the model learning rate, and when the value of the loss function remains unchanged, the training is completed; finally, the trained AI model is output, and the parameters of the AI ​​model are recorded, including the network structure, the number of model iterations, and the model learning rate; S4.3: Verify the trained AI model output by S4.2 through the test set, obtain the verification accuracy PRE and recall REC of the test set, and obtain the F1 score of the test set. ,If the F1 score is greater than the threshold, it means that the AI ​​model verification has passed, otherwise repeat S4.

2.

8. The method for intelligent production quality management based on AI training according to claim 1, characterized in that: The results evaluated as abnormal in S6 are obtained by first obtaining the number of products n_s, the total number Tn_s, the number of correct warnings n_al, and the total number of warnings Tn_al whose AI model prediction results are consistent with the actual quality labels, and obtaining the production quality intelligent management index PQMI, , a1 and a2 represent the corresponding weight coefficients; then the production quality intelligent management index PQMI is compared with the threshold. If it is greater than or equal to the threshold, it means that the production quality intelligent management is good, otherwise it means that the production quality intelligent management is abnormal.

Images