Production data processing method and system based on visual interaction
By combining the Internet of Things, neural networks, reinforcement learning and generative adversarial network technologies, real-time monitoring and optimization of production data is achieved, and the problem of difficult prediction and improvement of traditional statistical methods is solved, and the production efficiency and quality are improved.
Patent Information
- Application Number
- CN202411629544.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-15
- Publication Date
- 2025-02-14
AI Technical Summary
The prior art is difficult to predict possible future problems in production lines through traditional statistical methods or provide improvement solutions, resulting in limited production efficiency and quality improvement.
Using production data processing methods based on visual interaction, through advanced technologies such as the Internet of Things, neural networks, reinforcement learning and generative adversarial networks, the production status is monitored in real time, the production, environment and historical data are integrated, probability scores are generated, and reinforcement learning algorithms are used and Generative adversarial networks are optimized.
Real-time monitoring and data-driven decision-making of the production process are realized, production efficiency and quality are improved, problems in the production process are discovered and solved in a timely manner, and potential production risks are predicted and prevented.
Smart Images
Figure CN119444136A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular to a production data processing method and system based on visual interaction. Background Art
[0002] With the development of new generation information technologies such as the Internet, Internet of Things, cloud computing, and artificial intelligence, more and more companies are beginning to focus on using these new technologies to achieve digital transformation. An important aspect of digital transformation is to use various sensors to collect various production data on the production line and analyze and mine them to discover potential value. Currently, most companies use traditional statistical methods to analyze production data. Although this method can help us understand the current production status, it cannot predict possible problems in the future or provide improvement plans. Summary of the invention
[0003] The purpose of the present invention is to provide a production data processing method and system based on visual interaction, which combines advanced technologies such as the Internet of Things, neural networks, reinforcement learning, and generative adversarial networks to monitor the production status in real time, make decisions in a data-driven manner, and ultimately achieve improved production efficiency and quality.
[0004] The purpose of the present invention can be achieved through the following technical solutions: The present application provides a production data processing method based on visual interaction, comprising the following steps: Obtain production data through IoT devices and record them; Use neural networks to analyze key features of production data and generate probability scores; The method comprises extracting a plurality of key features from the production data; generating a first probability score for each of the features using a first neural network; generating a second probability score for each of the features using a second neural network; calculating a probability score for each of the features based on both the first probability score and the second probability score; and providing the user with access to a selected portion of the production data; Integrate production data with environmental data and historical data to obtain a data view of the production process.
[0005] Furthermore, production data is obtained through Internet of Things devices, which specifically include an environment acquisition module and a device perception module. The environment acquisition module is used to collect temperature and humidity data during the production process, and the device perception module obtains equipment operation data images during the production process through a camera.
[0006] Furthermore, before fusing the production data with the environmental data and the historical data, the method further includes: classifying the production data; wherein the production data includes two types of data, represented as visual data and non-visual data respectively.
[0007] Furthermore, after the production data is classified, the visual data and the non-visual data are clustered and dimensionally reduced. Specifically, the clustering of the visual data and the non-visual data includes: Use Gaussian kernel function to calculate the similarity between visual data and non-visual data; Construct similarity moments based on similarity measures; Use the similarity matrix to calculate the Laplacian matrix and find its first k smallest non-zero eigenvectors; Project the original data onto these feature vectors to form a low-dimensional feature representation; Apply K-means on the low-dimensional feature space to determine the number and distribution of clusters.
[0008] Furthermore, the visual data and non-visual data are subjected to dimensionality reduction processing, specifically including: Use t-SNE to reduce the dimension of visual and non-visual data; Simulate the similarity of high-dimensional space in low-dimensional space and adjust the position of data points through the optimization process; After the optimization is completed, a dimensionality reduction representation of the data in two-dimensional or three-dimensional space is obtained.
[0009] Furthermore, the visual data and non-visual data after dimensionality reduction are spliced together, specifically including splicing the visual data and non-visual data after dimensionality reduction to form a comprehensive data set, and using a scatter plot to visualize the data after dimensionality reduction to intuitively display the distribution and structure of the data.
[0010] Furthermore, after fusing the production data with the environmental data and the historical data, the method further includes: using a reinforcement learning algorithm to model the production data so as to obtain an optimal control strategy.
[0011] Furthermore, after fusing the production data with the environmental data and the historical data, it also includes: reconstructing the production data using a generative adversarial network, by constructing a generative adversarial network, including two parts: a generator and a discriminator. The generator is used to generate new data samples, and the discriminator is used to judge the authenticity of the data samples.
[0012] A production data processing system based on visual interaction, comprising: Data acquisition module, used to obtain production data through IoT devices and record them; Among them, the production data is obtained through the Internet of Things device, and the Internet of Things device includes an environment acquisition module and an equipment perception module. The environment acquisition module is used to collect temperature and humidity data during the production process, and the equipment perception module obtains equipment operation data images during the production process through a camera; Data analysis module, which is used to analyze the key features of production data using neural networks and generate probability scores; analyze the key features of production data using neural networks wherein a plurality of key features are determined to be extracted from the production data; a first probability score for each of the features is generated using a first neural network; a second probability score for each of the features is generated using a second neural network; a probability score for each of the features is calculated based on both the first probability score and the second probability score; and access to a selected portion of the production data is provided to the user; Data fusion module, used to fuse production data with environmental data and historical data to obtain a data view of the production process; The model building module is used to model production data using reinforcement learning algorithms in order to obtain the optimal control strategy.
[0013] Furthermore, the system also includes: a production data reconstruction module, which is used to reconstruct the production data using a generative adversarial network.
[0014] The beneficial effects of the present invention are: By acquiring production data through IoT devices and using neural networks to analyze key features to generate probability scores, managers can monitor the production process in real time through a visual interface, view historical trends and key indicators, and make more data-driven decisions. This method improves the efficiency and quality of production monitoring and helps to promptly identify and solve problems in the production process. By deeply analyzing production data, including environmental data, the fusion of historical data, and processing visual and non-visual data using clustering and dimensionality reduction techniques, this method can reveal potential patterns and anomalies in the production process, which can not only help companies optimize production processes, improve output and quality, but also predict and prevent potential production risks; Combining reinforcement learning algorithms and generative adversarial networks to model and reconstruct production data can automatically discover the optimal control strategy and generate new data samples to optimize the quality of existing data, which has promoted the production process to a higher level of automation and intelligence, helping to reduce production costs and improve production efficiency and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] For better understanding and implementation, the technical solution of the present application is described in detail below with reference to the accompanying drawings.
[0016] Figure 1 A flowchart of a production data processing method based on visual interaction provided in Example 1 of the present application; Figure 2 A schematic diagram of a process for clustering visual data and non-visual data in a production data processing method based on visual interaction provided in Example 1 of the present application; Figure 3 A schematic diagram of the results of the production data processing system based on visual interaction provided in Example 2 of the present application. DETAILED DESCRIPTION
[0017] In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, exemplary embodiments will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are only examples of methods and systems consistent with some aspects of the present application as detailed in the attached claims.
[0018] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.
[0019] The specific implementation methods, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.
[0020] Example 1 See also Figure 1-Figure 2 ,This embodiment provides a production data processing method and system based on visual interaction, which combines advanced technologies such as the Internet of Things, neural networks, reinforcement learning, and generative adversarial networks to monitor the production status in real time, and make decisions in a data-driven manner, ultimately achieving improved production efficiency and quality.
[0021] The present invention provides a production data processing method based on visual interaction, comprising the following steps: S1. Obtain production data through IoT devices and record them; Among them, production data is obtained through Internet of Things devices, specifically including an environment acquisition module and an equipment perception module. The environment acquisition module is used to collect temperature and humidity data during the production process. The equipment perception module obtains equipment operation data images during the production process through a camera. The equipment operation data includes equipment operation, stop, failure, idle and other status information.
[0022] Further, the acquired production data is recorded, including sending the production data to an IoT gateway, adding a timestamp to each data point, and attaching a device identifier to the acquired data point of the device operation.
[0023] S2. Use neural networks to analyze key features of production data and generate probability scores; Specifically, using a neural network to analyze key features of production data includes: extracting a plurality of key features from the production data; using a first neural network to generate a first probability score for each of the features; using a second neural network to generate a second probability score for each of the features; calculating a probability score for each of the features based on both the first probability score and the second probability score; and providing the user with access to a selected portion of the production data. wherein the selected portion is selected based on the probability score for one of the features.
[0024] Using neural networks to analyze the key features of production data can effectively identify and evaluate the importance of each key feature in the production process, generate corresponding probability scores, and provide users with the most relevant data based on these scores, thereby optimizing production monitoring and decision-making and improving production efficiency and quality.
[0025] Furthermore, the first probability score of each feature is generated, and the neural network structure of the multilayer perceptron is selected. The multilayer perceptron (MLP) is a classic feedforward neural network structure suitable for processing structured data. The specific implementation steps are: Input layer: receives the extracted key feature vector. Assuming there are n key features, the dimension of the input layer is n. Hidden layer: contains several hidden layers, each hidden layer consists of several neurons, and each neuron is fully connected to all neurons in the previous layer; Output layer: Generates the first probability score of each feature. The number of neurons in the output layer is the same as the number of input features, and each neuron corresponds to the probability score of a feature.
[0026] Further, a second probability score for each feature is generated, and a CNN model is selected. By training a CNN model, the second probability score can be generated according to the following steps, including: Model prediction probability: For a given input data, the model outputs the probability of each category, which is obtained by applying the softmax function to the output layer of the model; Sort and select the second highest probability: For each input data, sort the probability scores predicted by the model and then select the second highest one.
[0027] Based on the first probability score and the second probability score, a probability score for each of the features is calculated, a new neural network model is trained, and the same input data and labels are used. For each input feature, the probability score of its classification is calculated by the new model, and the second highest probability score of each feature is extracted.
[0028] Specifically, the production data processing method includes: Obtain various sensor data at the production site through IoT devices, such as temperature and humidity sensors, infrared detectors, etc. Then, convert these raw data into a standard format and upload them to the cloud server for analysis and processing; Secondly, machine learning algorithms are used to automatically identify abnormal situations and issue alarms to remind relevant personnel to take timely countermeasures. Finally, the analysis results are presented to managers in the form of charts so that they can better understand the current production status and formulate corresponding countermeasures.
[0029] Specifically, a visual interface is also provided for users to view real-time monitoring images, historical trend curves and other related information.
[0030] S3. Integrate production data with environmental data and historical data to obtain a data view of the production process.
[0031] Furthermore, before fusing the production data with the environmental data and the historical data, the acquired equipment operation data image is preprocessed, including denoising, contrast adjustment, standardization, etc., and key visual features are extracted from the preprocessed image data, and key non-visual features are extracted from the sensor data; As a preferred solution, before fusing the production data with the environmental data and the historical data, the method may further include: classifying the production data; wherein the production data includes two types of data, represented as visual data and non-visual data respectively.
[0032] Furthermore, after the production data is classified, it also includes: clustering and dimensionality reduction processing of visual data and non-visual data.
[0033] Specifically, clustering visual data and non-visual data includes: S21, using Gaussian kernel function to calculate the similarity between visual data and non-visual data; S22, constructing a similarity moment based on the similarity measure; S23, using the similarity matrix to calculate the Laplace matrix, and find its first k smallest non-zero eigenvectors; S24, projecting the original data onto these feature vectors to form a low-dimensional feature representation; S25. Apply K-means on the low-dimensional feature space to determine the number and distribution of clusters.
[0034] Use t-SNE to reduce the dimension of visual and non-visual data; Simulate the similarity of high-dimensional space in low-dimensional space and adjust the position of data points through the optimization process; After the optimization is completed, a dimensionality reduction representation of the data in two-dimensional or three-dimensional space is obtained.
[0035] Furthermore, the visual data and non-visual data after dimensionality reduction are spliced.
[0036] Specifically, it involves splicing the dimensionality-reduced visual data and non-visual data to form a comprehensive data set, and using scatter plots to visualize the dimensionality-reduced data to intuitively display the distribution and structure of the data.
[0037] It is integrated into the entire production data processing flow through clustering and dimensionality reduction steps, which not only enhances the data analysis capabilities, but also provides deeper insights through visual interaction, helping users make more data-driven decisions.
[0038] Furthermore, after fusing the production data with the environmental data and the historical data, the method further includes: S4, using a reinforcement learning algorithm to model the production data in order to obtain an optimal control strategy.
[0039] Furthermore, after fusing the production data with the environmental data and the historical data, the method further includes: S5, reconstructing the production data using a generative adversarial network.
[0040] Specifically, the production data is modeled using a reinforcement learning algorithm to construct a reward function with probability score input, including Define production optimization goals: clearly define the goals that need to be optimized, such as increasing output, reducing scrap rate, reducing energy consumption, etc. Collect production data: Collect data related to the production process, including input raw materials, production time, equipment status, output quality, etc.; Calculate probability scores: Use historical data and statistical models to calculate the success probability or quality probability score for each production process; Design a reward function: Based on the production optimization goal and the probability score, design a reward function. The reward function should reflect the achievement of the production goal.
[0041] Specifically expressed as: ,in, , is the weight parameter, which is adjusted according to the specific optimization goal. is the success probability, F is the quality score, and C is the cost.
[0042] Specifically, we use a generative adversarial network (GAN) to reconstruct production data, including Construct a generative adversarial network: It includes two parts: the generator and the discriminator. The generator is used to generate new data samples, and the discriminator is used to judge the authenticity of the data samples. Data preprocessing: Preprocess existing production data to ensure that the data format and quality are suitable for input into GAN.
[0043] Retraining the GAN model includes initializing the parameters of the generator and the discriminator. The generator generates new data samples, the discriminator determines the authenticity of the samples, and updates the generator based on the feedback from the discriminator. The training process is as follows: in, is the distribution of the real data, is the distribution of input noise, D and G are the parameters of the discriminator and generator respectively, is the log-likelihood of the true data, Fake data samples.
[0044] After training is complete, use the generator to generate more data samples, or to improve the quality of existing data.
[0045] The embodiment of the present invention obtains production data through the Internet of Things devices, and collects temperature and humidity data in the production environment and the operating status of the equipment through the environment acquisition module and the equipment perception module. These data are recorded and sent to the Internet of Things gateway, with timestamps and device identifiers attached to ensure the accuracy and traceability of the data.
[0046] Next, a neural network is used to analyze the key features of the acquired production data and generate probability scores. Specifically, a multi-layer perceptron (MLP) is used to generate the first probability score for each feature, a convolutional neural network (CNN) is used to generate the second probability score, and finally a new neural network is used to calculate the comprehensive probability score. These scores help users identify and evaluate key features in the production process, thereby optimizing production monitoring and decision-making.
[0047] The method then fuses the production data with environmental and historical data. Prior to this, the data was preprocessed and classified, and the visual and non-visual data were clustered and reduced in dimension respectively, so that their structure can be better displayed and analyzed when the data is fused. Through clustering and dimensionality reduction, the data is presented in a more intuitive form, helping users to deeply understand the data distribution and trends in the production process through a visual interface.
[0048] After data fusion, the method also applies reinforcement learning algorithms and generative adversarial networks (GANs) for further optimization. Reinforcement learning is used to model production data and find the optimal control strategy based on the defined reward function to achieve production goals, such as increasing output or reducing energy consumption. Generative adversarial networks provide support for further data analysis and production optimization by generating new samples or optimizing the quality of existing data.
[0049] Embodiment 2: This embodiment provides a production data processing system based on visual interaction, such as Figure 2 As shown, it is applied to a production data processing method based on visual interaction, including: The data acquisition module is used to obtain production data through the Internet of Things devices and record them; wherein, obtaining production data through the Internet of Things devices specifically includes an environment acquisition module and a device perception module. The environment acquisition module is used to collect temperature and humidity data during the production process. The device perception module obtains equipment operation data images during the production process through a camera. The equipment operation data includes equipment operation, stop, fault, idle and other status information.
[0050] Further, recording the acquired production data includes sending the production data to an IoT gateway, adding a timestamp to each data point, and attaching a device identifier to the acquired data point of the device operation; A data analysis module, for analyzing key features of production data using a neural network and generating a probability score; specifically, analyzing key features of production data using a neural network includes: extracting a plurality of key features from the production data; generating a first probability score for each of the features using a first neural network; generating a second probability score for each of the features using a second neural network; calculating a probability score for each of the features based on both the first probability score and the second probability score; and providing the user with access to a selected portion of the production data.
[0051] The data fusion module is used to fuse production data with environmental data and historical data to obtain a data view of the production process.
[0052] Furthermore, the system also includes: a model building module for modeling production data using a reinforcement learning algorithm to obtain an optimal control strategy.
[0053] Furthermore, the system also includes: a production data reconstruction module, which is used to reconstruct the production data using a generative adversarial network.
[0054] This embodiment is based on a visually interactive production data processing system. By analyzing and processing production data, it can help users understand the status of the entire production line more intuitively, so as to adjust production plans and optimize resource allocation in a timely manner. In addition, the use of reinforcement learning networks and generative adversarial networks can effectively optimize the production process and improve production efficiency and quality. The reinforcement learning network uses probability scores to design reward functions, and the generative adversarial network reconstructs production data. It can also automatically detect potential quality risks and put forward improvement suggestions, which helps to reduce waste and improve production efficiency.
[0055] Embodiment three: This embodiment provides a computer-readable storage medium, which stores instructions. When the instructions are executed on a computer, the computer executes the method described above.
[0056] The main function of the present invention is to be applied in the field of industrial big data, especially in production data processing.
[0057] Embodiment 4 This embodiment is different from the first embodiment in that a deep reinforcement learning strategy is applied to optimize the production process. DQN is an algorithm that combines deep learning and Q learning (a reinforcement learning method) and is used to solve decision-making problems with high-dimensional observation space.
[0058] Specifically expressed as: Define the state space: determine the states in the production process, which will be used as inputs to the DQN. The states may include machine status, current production, raw material inventory, etc. Define the action space: determine the actions that can be taken in a given state. The actions can be to adjust production parameters, start or stop a machine, etc. Design reward function: The reward function is used to evaluate the effect of each action and guide DQN to learn how to take actions to maximize the cumulative reward; Initialize the Q network and the target network, including the Q network is a deep neural network used to predict the expected return of a given state and action, and the target network is a network with the same structure as the Q network, whose parameters are regularly updated during training to stabilize the training process; Experience replay: Use the experience replay buffer to store the experience (state, action, reward, new state) of the agent's interaction with the environment. By randomly sampling experience from the buffer for training, the temporal correlation between the data can be broken.
[0059] Update the Q network: Update the parameters of the Q network according to the Bellman equation. The Bellman equation is the core of Q learning and is used to estimate the value of state-action pairs. The update rules are as follows: in, is the predicted value of the current state-action pair, is the learning rate, r is the immediate reward after executing the action, is a discount factor used to balance immediate rewards and future rewards, is the maximum predicted value of the next state.
[0060] Update the target network: To reduce instability during training, the parameters of the Q network are periodically copied to the target network, meaning after a certain time interval or number of updates; Training strategy: Through interaction with the environment, the agent continuously optimizes its strategy and selects actions that maximize the cumulative reward.
[0061] This embodiment defines the state space (such as machine status, current output, raw material inventory, etc.) and action space (such as adjusting production parameters, starting or stopping machines, etc.); designs reward functions to evaluate the effects of different actions; initializes the Q network and target network for predicting the expected return of state-action pairs; uses the experience replay buffer to store and randomly sample training experience to break the time correlation of data; updates the parameters of the Q network based on the Bellman equation, and periodically copies these parameters to the target network to stabilize the training process; and finally, through interaction with the environment, optimizes the strategy to select actions that maximize the cumulative reward. This process intuitively displays the decision path in the production process through a visual interface, helping users to easily understand and adjust production strategies, thereby improving production efficiency and quality.
[0062] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.
Claims
1. A production data processing method based on visual interaction, characterized in that: The steps include: Obtain production data through IoT devices and record them; Use neural networks to analyze key features of production data and generate probability scores; The method comprises extracting a plurality of key features from the production data; generating a first probability score for each of the features using a first neural network; generating a second probability score for each of the features using a second neural network; calculating a probability score for each of the features based on both the first probability score and the second probability score; and providing the user with access to a selected portion of the production data; Integrate production data with environmental data and historical data to obtain a data view of the production process.
2. The production data processing method based on visual interaction according to claim 1 is characterized in that: Production data is obtained through Internet of Things devices, which specifically include an environment acquisition module and an equipment perception module. The environment acquisition module is used to collect temperature and humidity data during the production process, and the equipment perception module obtains equipment operation data images during the production process through a camera.
3. The production data processing method based on visual interaction according to claim 1 is characterized in that: Before fusing the production data with the environmental data and the historical data, the method further includes: classifying the production data; wherein the production data includes two types of data, which are respectively represented as visual data and non-visual data.
4. The production data processing method based on visual interaction according to claim 3 is characterized in that: After the production data is classified, the visual data and non-visual data are clustered and dimension reduction is performed. Specifically, the clustering of visual data and non-visual data includes: Use Gaussian kernel function to calculate the similarity between visual data and non-visual data; Construct similarity moments based on similarity measures; Use the similarity matrix to calculate the Laplacian matrix and find its first k smallest non-zero eigenvectors; Project the original data onto the feature vector to form a low-dimensional feature representation; Apply K-means on the low-dimensional feature space to determine the number and distribution of clusters.
5. The production data processing method based on visual interaction according to claim 4 is characterized in that: Perform dimensionality reduction on visual and non-visual data, including: Use t-SNE to reduce the dimension of visual and non-visual data; Simulate the similarity of high-dimensional space in low-dimensional space and adjust the position of data points through the optimization process; After the optimization is completed, a dimensionality reduction representation of the data in two-dimensional or three-dimensional space is obtained.
6. The production data processing method based on visual interaction according to claim 5 is characterized in that: The visual data and non-visual data after dimensionality reduction are spliced together, specifically including: splicing the visual data and non-visual data after dimensionality reduction to form a comprehensive data set, and using scatter plots to visualize the data after dimensionality reduction to intuitively display the distribution and structure of the data.
7. The production data processing method based on visual interaction according to claim 1 is characterized in that: After fusing the production data with the environmental data and the historical data, the method further includes: using a reinforcement learning algorithm to model the production data so as to obtain an optimal control strategy.
8. The production data processing method based on visual interaction according to claim 1 is characterized in that: After fusing the production data with the environmental data and the historical data, the method also includes: reconstructing the production data using a generative adversarial network by constructing a generative adversarial network, including two parts: a generator and a discriminator. The generator is used to generate new data samples, and the discriminator is used to judge the authenticity of the data samples.
9. A production data processing system based on visual interaction, applied to the production data processing method based on visual interaction according to any one of claims 1 to 8, characterized in that: include: Data acquisition module, used to obtain production data through IoT devices and record them; Among them, the production data is obtained through the Internet of Things device, and the Internet of Things device includes an environment acquisition module and an equipment perception module. The environment acquisition module is used to collect temperature and humidity data during the production process, and the equipment perception module obtains equipment operation data images during the production process through a camera; Data analysis module, which is used to analyze the key features of production data using neural networks and generate probability scores; analyze the key features of production data using neural networks wherein a plurality of key features are determined to be extracted from the production data; a first probability score for each of the features is generated using a first neural network; a second probability score for each of the features is generated using a second neural network; a probability score for each of the features is calculated based on both the first probability score and the second probability score; and access to a selected portion of the production data is provided to the user; Data fusion module, used to fuse production data with environmental data and historical data to obtain a data view of the production process; The model building module is used to model production data using reinforcement learning algorithms in order to obtain the optimal control strategy.
10. The production data processing system based on visual interaction according to claim 9, characterized in that: Also includes: The production data reconstruction module is used to reconstruct the production data using a generative adversarial network.