Safety production control method and system for printing machine equipment

Abstract

The invention provides a printing machine equipment safety production control method and system, and relates to the technical field of printing machines, and the method comprises the steps: collecting the operation data of a printing machine group through a multi-mode sensor, extracting dynamic characteristics through a time sequence attention mechanism, and constructing a space-time incidence matrix to represent the coupling relation between the printing machines. And constructing a hierarchical deep learning model, integrating feature fusion, time sequence prediction and risk assessment, analyzing the health state of the equipment, calculating an overall safety risk index and identifying fault equipment. And generating an optimal scheduling scheme including an equipment start-stop sequence and a load distribution strategy based on a scheduling decision model of deep reinforcement learning in combination with a safety threshold, resource distribution and production efficiency constraints, and realizing safe and efficient scheduling of the printing machine group. Potential risks can be effectively identified, safety accidents are avoided, and the production efficiency is improved.

Description

Technical Field

[0001] The present invention relates to printing press technology, and in particular to a safety production control method and system for printing press equipment. Background Art

[0002] The scale of the printing press group is continuously expanding, posing higher requirements for its safe and efficient scheduling management. Traditional printing press scheduling methods mainly rely on manual experience and rules, and it is difficult to adapt to complex and changeable production environments. With the development of Internet of Things and artificial intelligence technologies, data-driven intelligent scheduling methods have gradually emerged, providing new ideas for the optimal scheduling of printing press groups.

[0003] Difficult to comprehensively perceive the equipment status: Traditional scheduling methods usually only rely on a few key parameters for monitoring, unable to comprehensively reflect the real-time operation status of the printing press group, and prone to safety hazards and efficiency losses.

[0004] Lack of consideration for the correlation between equipment: There are complex coupling relationships between printing presses. For example, a fault in one printing press may affect the normal operation of other printing presses. Traditional scheduling methods usually ignore this correlation and it is difficult to achieve global optimal scheduling.

[0005] Poor adaptability of the scheduling scheme: Changes in the production environment, such as fluctuations in order demand and degradation of equipment performance, will affect the effectiveness of the scheduling scheme. Traditional scheduling methods are difficult to dynamically adjust the scheduling strategy according to environmental changes, resulting in poor adaptability of the scheduling scheme. Summary of the Invention

[0006] Embodiments of the present invention provide a safety production control method and system for printing press equipment, which can solve the problems in the prior art.

[0007] In the first aspect of the embodiments of the present invention A safety production control method for printing press equipment is provided, including: Real-time collecting operation data through a multi-modal sensor network deployed in the printing press group, and using a temporal attention mechanism to extract features from the standardized data stream corresponding to the operation data to obtain a dynamic feature vector of the printing press group; constructing a spatio-temporal correlation matrix of the printing press group based on the dynamic feature vector; Based on the dynamic feature vector and the spatio-temporal correlation matrix, constructing a hierarchical deep learning model, inputting historical labeled data into the hierarchical deep learning model for training, and using the trained deep learning model to analyze the dynamic feature vector to generate an equipment health status evaluation result; calculating an overall safety risk index and an equipment identification result of the printing press group based on the equipment health status evaluation result and the spatio-temporal correlation matrix; Based on the overall security risk index and the device identification result, establish the intelligent scheduling constraint conditions for the printing press group; construct a scheduling decision model based on the deep reinforcement learning algorithm, and use the security threshold constraint, resource allocation constraint, and production efficiency constraint as the evaluation indicators of the reward function; use the scheduling decision model to generate an optimal scheduling plan, where the optimal scheduling plan includes the device start-stop sequence and the load allocation strategy; send the optimal scheduling plan to each printing press control system for execution.

[0008] Based on the dynamic feature vector and the spatio-temporal correlation matrix, construct a hierarchical deep learning model. Training by inputting historical annotation data into the hierarchical deep learning model includes: The hierarchical deep learning model includes a feature fusion layer, a time series prediction layer, and a risk assessment layer; in the feature fusion layer, transform the dynamic feature vector into a query matrix, a key matrix, and a value matrix through the self-attention mechanism, and calculate the correlation weights between features through scaled dot-product attention. Input the spatio-temporal correlation matrix into the graph attention network to extract the spatial dependence features between printing presses, and adaptively fuse the correlation weights between features and the spatial dependence features through a gating mechanism to generate a first comprehensive feature representation; In the time series prediction layer, input the first comprehensive feature representation into a bidirectional long short-term memory network and a multi-scale parallel convolutional neural network respectively. The bidirectional long short-term memory network extracts time series features, and the multi-scale parallel convolutional neural network extracts local features. Fuse the output features of the two through a residual connection structure to generate a second comprehensive feature representation; Input the second comprehensive feature representation into the risk assessment layer to generate a second prediction result; Input historical annotation data into the hierarchical deep learning model for training. Calculate the first loss value of the first prediction result using the mean squared error loss function, and calculate the second loss value of the second prediction result using the cross-entropy loss function; adaptively balance the first loss value and the second loss value through learnable weights to obtain the total loss value, and optimize the parameters of the hierarchical deep learning model based on the total loss value to obtain the trained deep learning model.

[0009] Use the trained deep learning model to analyze the dynamic feature vector and generate an evaluation result of the device health status; based on the evaluation result of the device health status and the spatio-temporal correlation matrix, calculate the overall security risk index and the device identification result of the printing press group, including: Use the trained deep learning model to analyze the dynamic feature vector, and fuse multiple prediction results through a weighted voting method to generate an evaluation result of the device health status; Construct a weighted directed graph based on the device health status evaluation result and the spatio-temporal correlation matrix. The nodes of the weighted directed graph represent printing press devices, and the edge weights represent the risk propagation intensity between devices. Use the PageRank algorithm to calculate the node importance scores. Combine the node importance scores with the local risk accumulation function. The local risk accumulation function introduces a decay factor in the time dimension and graph convolution features in the space dimension to calculate the overall safety risk index of the printing press group. Conduct multi-criteria decision analysis based on the weighted directed graph, and identify devices through the fuzzy comprehensive evaluation method to generate device identification results.

[0010] According to the overall safety risk index and the device identification results, establish the intelligent scheduling constraint conditions for the printing press group; construct a scheduling decision model based on the deep reinforcement learning algorithm, and use the safety threshold constraint, resource allocation constraint, and production efficiency constraint as the evaluation indicators of the reward function, including: Set a risk threshold vector according to the overall safety risk index. The risk threshold vector is used to divide the risk area above the preset risk threshold and the risk area below the preset risk threshold, and set corresponding scheduling restriction strategies for different risk areas. Construct a resource allocation constraint based on the device identification results. The resource allocation constraint includes a device importance scoring matrix and a spare device deployment strategy. The device importance scoring matrix is used to quantify the importance of devices in the production system; combine the scheduling restriction strategy and the resource allocation constraint to form the intelligent scheduling constraint conditions for the printing press group. Construct a scheduling decision model based on the deep reinforcement learning algorithm, map the intelligent scheduling constraint conditions to the state space and action space. The state space includes the real-time operating status information of the printing press group, and the action space defines the set of executable scheduling operations; construct a composite reward function, and use the safety threshold constraint, resource allocation constraint, and production efficiency constraint as the evaluation indicators of the composite reward function.

[0011] Use the scheduling decision model to generate an optimal scheduling plan. The optimal scheduling plan includes the device start-stop sequence and the load distribution strategy; send the optimal scheduling plan to each printing press control system for execution, including: Generate an optimal scheduling plan based on the scheduling decision model, construct a device start-stop state transition matrix through the dynamic programming algorithm; optimize the load distribution strategy in combination with the mixed integer programming method. The load distribution strategy is based on the remaining production capacity of the device and the urgency of the task, establish a task allocation weight coefficient to achieve dynamic balance of the device group load; combine the device start-stop state transition matrix and the load distribution strategy to form the optimal scheduling plan. Build a hierarchical scheduling instruction distribution system, which parses the optimal scheduling scheme into device-level control instructions; establish a real-time communication mechanism between the central scheduler and the local control system of the printing press, and send the device-level control instructions to each printing press control system.

[0012] Generate an optimal scheduling scheme based on the scheduling decision model, and construct a device start-stop state transition matrix through the dynamic programming algorithm; optimize the load distribution strategy by combining the mixed integer programming method. The load distribution strategy is based on the remaining production capacity of the device and the urgency of the task, and establish a task allocation weight coefficient to achieve dynamic balance of the load of the device group; the combination of the device start-stop state transition matrix and the load distribution strategy to form an optimal scheduling scheme includes: The device start-stop state transition matrix includes the state transition law, state transition cost and preheating time constraint of the printing press; use the value iteration algorithm to calculate the state value function and state-action value function, and determine the optimal start-stop strategy based on the state value function and the state-action value function; Calculate the remaining production capacity of the device based on the operating state of the printing press, and evaluate the urgency of the task according to the delivery time of the production task; establish a task allocation weight coefficient using the remaining production capacity of the device and the urgency of the task, and the task allocation weight coefficient is used for priority sorting and resource allocation of the task; Construct a load balancing constraint condition based on the task allocation weight coefficient. The load balancing constraint condition includes device production capacity limit, task time sequence relationship and process parameter requirements; obtain a task allocation scheme that meets the load balancing constraint condition by solving the mixed integer programming model, and achieve dynamic balance of the load of the printing press group.

[0013] In the second aspect of the embodiments of the present invention, Provide a safety production control system for printing press equipment, including: The first unit is used to collect operation data in real time through a multi-modal sensor network deployed in the printing press group, extract features from the standardized data stream corresponding to the operation data by using the temporal attention mechanism, and obtain the dynamic feature vector of the printing press group; construct a spatio-temporal correlation matrix of the printing press group based on the dynamic feature vector; The second unit is used to construct a hierarchical deep learning model based on the dynamic feature vector and the spatio-temporal correlation matrix, input historical labeled data into the hierarchical deep learning model for training, analyze the dynamic feature vector by using the trained deep learning model, and generate an evaluation result of the device health state; calculate the overall safety risk index and device identification result of the printing press group based on the device health state evaluation result and the spatio-temporal correlation matrix; A third unit is configured to establish intelligent scheduling constraint conditions for the printing press group according to the overall safety risk index and the device identification result; construct a scheduling decision model based on a deep reinforcement learning algorithm, and use safety threshold constraints, resource allocation constraints, and production efficiency constraints as evaluation indicators of the reward function; generate an optimal scheduling plan by using the scheduling decision model, where the optimal scheduling plan includes the device start-stop sequence and the load allocation strategy; and send the optimal scheduling plan to each printing press control system for execution.

[0014] In a third aspect of the embodiments of the present invention, a kind of electronic device is provided, including: a processor; a memory for storing instructions executable by the processor; wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0015] In a fourth aspect of the embodiments of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0016] The beneficial effects of this application are as follows: 1. Improve the safety of the printing press group: By real-time monitoring and analyzing the operating status of the printing press group, potential safety risks can be effectively identified, and corresponding scheduling measures can be taken in a timely manner to avoid accidents and ensure production safety.

[0017] 2. Optimize resource allocation and improve production efficiency: The intelligent scheduling method based on deep reinforcement learning can dynamically adjust the load allocation and start-stop sequence of the printing press group according to the health status of the devices and production requirements, thereby optimizing resource allocation and improving production efficiency.

[0018] 3. Realize the autonomous learning and continuous optimization of the printing press group: By feeding the execution effect data back into the model training process, the performance of the deep learning model and the scheduling decision model can be continuously optimized, making the scheduling strategy of the printing press group more intelligent and adaptive. Description of the Drawings

[0019] Figure 1 is a schematic flow chart of a method for controlling the safe production of a printing press device according to an embodiment of the present invention; Figure 2 is a schematic structural diagram of a control system for the safe production of a printing press device according to an embodiment of the present invention. Detailed Embodiments

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0021] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0022] Figure 1 The following is a schematic flowchart of a method for controlling the safe production of a printing press device according to an embodiment of the present invention. As Figure 1 shown, the method includes: Collecting operation data in real time through a multi-modal sensor network deployed in a printing press group, and using a temporal attention mechanism to extract features from the standardized data stream corresponding to the operation data to obtain a dynamic feature vector of the printing press group; constructing a spatio-temporal correlation matrix of the printing press group based on the dynamic feature vector; Based on the dynamic feature vector and the spatio-temporal correlation matrix, constructing a hierarchical deep learning model, inputting historical labeled data into the hierarchical deep learning model for training, and using the trained deep learning model to analyze the dynamic feature vector to generate an evaluation result of the device health state; calculating the overall safety risk index and device identification result of the printing press group based on the device health state evaluation result and the spatio-temporal correlation matrix; According to the overall safety risk index and the device identification result, establishing intelligent scheduling constraint conditions for the printing press group; constructing a scheduling decision model based on a deep reinforcement learning algorithm, and using safety threshold constraints, resource allocation constraints, and production efficiency constraints as evaluation indicators of the reward function; generating an optimal scheduling plan using the scheduling decision model, where the optimal scheduling plan includes the device start-stop sequence and load distribution strategy; and sending the optimal scheduling plan to each printing press control system for execution.

[0023] In an alternative embodiment, constructing a hierarchical deep learning model based on the dynamic feature vector and the spatio-temporal correlation matrix, and inputting historical labeled data into the hierarchical deep learning model for training includes: The hierarchical deep learning model includes a feature fusion layer, a time series prediction layer, and a risk assessment layer; in the feature fusion layer, the dynamic feature vector is converted into a query matrix, a key matrix, and a value matrix through a self-attention mechanism, and the feature correlation weights are calculated through scaled dot-product attention. The spatio-temporal correlation matrix is input into a graph attention network to extract the spatial dependence features between printing presses, and the feature correlation weights and the spatial dependence features are adaptively fused through a gating mechanism to generate a first comprehensive feature representation; In the time series prediction layer, the first comprehensive feature representation is respectively input into a bidirectional long short-term memory network and a multi-scale parallel convolutional neural network. The bidirectional long short-term memory network extracts time series features, and the multi-scale parallel convolutional neural network extracts local features. The output features of the two are fused through a residual connection structure to generate a second comprehensive feature representation; The second comprehensive feature representation is input into the risk assessment layer to generate a second prediction result; The historical labeled data is input into the hierarchical deep learning model for training. The mean squared error loss function is used to calculate the first loss value of the first prediction result, and the cross-entropy loss function is used to calculate the second loss value of the second prediction result; the first loss value and the second loss value are adaptively balanced through learnable weights to obtain a total loss value, and the parameters of the hierarchical deep learning model are optimized based on the total loss value to obtain a trained deep learning model.

[0024] Obtain the dynamic feature vector and spatio-temporal correlation matrix of the device. The dynamic feature vector can include the operating data, environmental data, etc. of the device, such as temperature, pressure, vibration frequency, etc. The spatio-temporal correlation matrix describes the spatial position relationship and mutual influence degree between different devices. For example, the element values in the matrix can represent the distance, connection relationship, or information transmission intensity between devices. Suppose there are 3 devices, and their dynamic feature vectors are [1, 2, 3], [4, 5, 6], and [7, 8, 9] respectively, and the spatio-temporal correlation matrix is [[1, 0.5, 0.2], [0.5, 1, 0.8], [0.2, 0.8, 1]].

[0025] Construct a hierarchical deep learning model, which consists of a feature fusion layer, a temporal prediction layer, and a risk assessment layer. In the feature fusion layer, first use the self-attention mechanism to convert the dynamic feature vector into a query matrix, a key matrix, and a value matrix. Then, calculate the correlation weights between features through scaled dot-product attention to capture the mutual relationships between different feature dimensions. At the same time, input the spatio-temporal correlation matrix into the graph attention network to extract the spatial dependence features between devices. Finally, adaptively fuse the correlation weights between features and the spatial dependence features through a gating mechanism to generate the first comprehensive feature representation. For example, the correlation weights between features of device 1 are [0.2, 0.3, 0.5], the spatial dependence features are [0.6, 0.3, 0.1], and the fused first comprehensive feature representation is [0.4, 0.35, 0.3].

[0026] In the temporal prediction layer, input the first comprehensive feature representation into a bidirectional long short-term memory network and a multi-scale parallel convolutional neural network respectively. The bidirectional long short-term memory network can capture the long-term dependence relationships in time series data and extract temporal features. The multi-scale parallel convolutional neural network can extract local features at different scales. Then, fuse the output features of the two through a residual connection structure to generate the second comprehensive feature representation. For example, the temporal features of device 1 are [0.1, 0.2, 0.7], the local features are [0.3, 0.5, 0.2], and the fused second comprehensive feature representation is [0.2, 0.35, 0.45].

[0027] Input the second comprehensive feature representation into the risk assessment layer. This layer constructs a main branch and a secondary branch based on the multi-task learning framework. The main branch uses a multi-layer perceptron to predict the device health indicator and generates the first prediction result. The secondary branch uses a conditional random field model to model the risk propagation mechanism between devices and generates the second prediction result. For example, the predicted value of the health indicator of device 1 is 0.8, and the risk propagation probability is [0.1, 0.2, 0.7].

[0028] Use historical labeled data to train the model. Use the mean squared error loss function to calculate the first loss value of the first prediction result, and use the cross-entropy loss function to calculate the second loss value of the second prediction result. Adaptively balance the first loss value and the second loss value through learnable weights to obtain the total loss value. Optimize the parameters of the model based on the total loss value to obtain the trained deep learning model. For example, if the true value of the health indicator of device 1 is 0.9, then the first loss value is 0.01. Suppose the true risk propagation probability of device 1 is [0.2, 0.1, 0.7], then the second loss value is 0.1. Suppose the learnable weight is 0.5, then the total loss value is 0.055.

[0029] The solution of this application can: The prediction accuracy is improved. By fusing the dynamic feature vectors and spatio-temporal correlation matrices, and combining multiple deep learning models, it is able to capture the device status information more comprehensively, thus improving the prediction accuracy. The robustness of the model is enhanced. The multi-task learning framework and residual connection structure can effectively alleviate the overfitting problem, enhancing the generalization ability and robustness of the model. The prediction efficiency is improved. The hierarchical structure and parallel computing mechanism can effectively reduce the computational complexity of the model and improve the prediction efficiency.

[0030] In an optional implementation, the trained deep learning model is used to analyze the dynamic feature vectors to generate an evaluation result of the device health status; based on the evaluation result of the device health status and the spatio-temporal correlation matrix, calculating the overall safety risk index and device identification result of the printing press group includes: The trained deep learning model is used to analyze the dynamic feature vectors, and multiple prediction results are fused by a weighted voting method to generate an evaluation result of the device health status; A weighted directed graph is constructed based on the evaluation result of the device health status and the spatio-temporal correlation matrix. The nodes of the weighted directed graph represent printing press devices, and the edge weights represent the risk propagation intensity between devices. The PageRank algorithm is used to calculate the node importance scores; The node importance scores are combined with a local risk accumulation function. The local risk accumulation function introduces a decay factor in the time dimension and graph convolution features in the space dimension to calculate the overall safety risk index of the printing press group; Based on the weighted directed graph, multi-criteria decision analysis is performed, and devices are identified through a fuzzy comprehensive evaluation method to generate device identification results.

[0031] The operating data of the printing press group, such as vibration signals, temperature, current, pressure, etc., are collected in real time through a distributed sensor network. These data are processed through data cleaning and feature engineering and transformed into high-dimensional dynamic feature vectors to characterize the real-time operating state of the devices. For example, after the vibration signal of a printing press undergoes a fast Fourier transform, the vibration amplitudes in multiple frequency bands can be obtained, and these amplitudes together with data such as temperature and current constitute the dynamic feature vector of the printing press.

[0032] To improve the robustness and generalization ability of the model, the dynamic feature vectors are preprocessed. The sliding window mechanism is adopted to divide the dynamic feature vector sequence into multiple overlapping feature windows, and each window contains the feature data within a certain time length. For example, with a window length of 10 seconds and a sliding step of 5 seconds, the continuous dynamic feature vector sequence is divided into multiple overlapping 10 - second windows. Then, wavelet transform denoising is performed on each feature window to remove high - frequency noise interference. Finally, the denoised feature sequence is normalized to map the data to a unified numerical range and eliminate the influence of different feature dimensions. For example, all feature data are normalized to the range between 0 and 1.

[0033] The preprocessed feature sequence is input into multiple heterogeneous sub - models for health state assessment. These sub - models adopt different deep - learning network structures and parameter configurations, such as convolutional neural networks, recurrent neural networks, long short - term memory networks, etc. Each sub - model is trained with a large amount of historical data and can capture the subtle changes in the device operation state from different perspectives. To make full use of the advantages of different sub - models, the evaluation weights of each sub - model are dynamically adjusted according to historical performance. For example, if a certain sub - model has a higher prediction accuracy for device failures in the past period, a higher weight is assigned to it. The prediction results of multiple sub - models are fused through weighted voting to generate the final device health state assessment result. For example, if the health states of a printing press predicted by three sub - models are 0.8, 0.9, and 0.7 respectively, and the corresponding weights are 0.3, 0.5, and 0.2 respectively, then the final health state assessment result is 0.8×0.3 + 0.9×0.5 + 0.7×0.2 = 0.83.

[0034] Based on the device health state assessment result and the pre - established spatio - temporal correlation matrix, a weighted directed graph is constructed. The nodes in the graph represent the printing press devices, and the edge weights represent the risk propagation intensity between devices, which is determined by factors such as the physical distance between devices and the correlation degree of the process flow. For example, between two printing presses that are adjacent in location and closely related in the process flow, the risk propagation intensity is higher, and the corresponding edge weight is also larger. The PageRank algorithm is used to calculate the importance score of each node, and this score reflects the influence of the device in the printing press group. For example, a printing press in a core position and connected to multiple other devices has a higher importance score. The node importance score is combined with the local risk accumulation function to calculate the overall safety risk index of the printing press group. The local risk accumulation function introduces a decay factor in the time dimension and graph convolution features in the space dimension, comprehensively considering the influence of the device's own risk and the risks of its surrounding devices. The decay factor in the time dimension makes the risks that occurred recently have a greater impact on the overall risk index, while the graph convolution features in the space dimension consider the propagation effect of risks between devices.

[0035] Multi-criteria decision analysis is carried out based on the constructed weighted directed graph, comprehensively considering the intrinsic importance indicators of equipment (such as the value and performance of equipment), topological importance indicators (such as the centrality of equipment in the network), and dynamic importance indicators (such as the real-time health status of equipment). The key equipment that has the greatest impact on the overall safety risk of the printing press group is identified through the fuzzy comprehensive evaluation method. For example, a printing press with high intrinsic importance, a critical topological position, and a poor health status will be identified as key equipment. The calculated overall safety risk index and equipment identification results are input into the intelligent scheduling decision-making module to optimize the production scheduling plan, such as adjusting the production plan and arranging equipment maintenance, etc., to reduce the overall safety risk.

[0036] The solution of this application can: Improve the accuracy and reliability of equipment health status assessment: Through the multi-model fusion and dynamic weight adjustment mechanism, the limitations of a single model can be effectively reduced, and the accuracy and reliability of the assessment results can be improved. Precisely quantify the overall safety risk of the printing press group: By constructing a weighted directed graph and introducing a spatio-temporal correlation matrix, the risk propagation effect between equipment can be considered more comprehensively, so as to more precisely quantify the overall safety risk of the printing press group. Achieve intelligent scheduling decision-making and risk pre-control: By inputting the overall safety risk index and equipment identification results into the intelligent scheduling decision-making module, the production scheduling plan can be optimized, preventive measures can be taken in advance, the safety risk can be reduced, and major accidents can be avoided.

[0037] In an optional implementation manner, according to the overall safety risk index and the equipment identification results, the intelligent scheduling constraint conditions of the printing press group are established; a scheduling decision model is constructed based on the deep reinforcement learning algorithm, and the safety threshold constraint, resource allocation constraint, and production efficiency constraint are used as evaluation indicators of the reward function, including: Set a risk threshold vector according to the overall safety risk index, and the risk threshold vector is used to divide the risk area higher than the preset risk threshold and the risk area lower than the preset risk threshold, and corresponding scheduling restriction strategies are set for different risk areas; Construct a resource allocation constraint based on the equipment identification results, and the resource allocation constraint includes an equipment importance scoring matrix and a spare equipment deployment strategy, and the equipment importance scoring matrix is used to quantify the importance of equipment in the production system; combine the scheduling restriction strategy and the resource allocation constraint to form the intelligent scheduling constraint conditions of the printing press group; Construct a scheduling decision model based on the deep reinforcement learning algorithm, map the intelligent scheduling constraint conditions to the state space and action space, the state space includes the real-time operation state information of the printing press group, and the action space defines the set of executable scheduling operations; construct a composite reward function, and use the safety threshold constraint, resource allocation constraint, and production efficiency constraint as evaluation indicators of the composite reward function.

[0038] Conduct safety risk assessment and equipment identification. Collect real-time operation data of the printing press group through a sensor network, such as temperature, pressure, vibration, etc., as well as information on the operation status and fault records of the equipment. Using these data, combined with historical data and expert knowledge, calculate the overall safety risk index for each printing press. At the same time, identify each device and record information such as its type, performance parameters, maintenance history, etc. For example, the safety risk index of one printing press is 0.8, and that of another is 0.2. The equipment identification results show that the former is an old device and the latter is a new device.

[0039] Establish intelligent scheduling constraints. Set a risk threshold vector according to the overall safety risk index. For example, set the risk threshold to 0.5. The area above 0.5 is the high-risk area, and the area below 0.5 is the low-risk area. Set corresponding scheduling restriction strategies for different risk areas. For example, for the printing presses in the high-risk area, restrict their processing of high-load tasks or increase the inspection frequency. For the printing presses in the low-risk area, the restrictions can be relaxed. Construct resource allocation constraints based on the equipment identification results. According to information such as the importance and performance parameters of the equipment, construct an equipment importance scoring matrix. For example, the score of a new device is 10, and the score of an old device is 5. At the same time, formulate a spare equipment allocation strategy. For example, when an important device fails, how to quickly allocate spare equipment to ensure the continuity of production. Combine the scheduling restriction strategies and resource allocation constraints to form the intelligent scheduling constraints for the printing press group. For example, a high-risk old device will be restricted from processing high-load tasks and will be given priority in arranging spare equipment to reduce risks.

[0040] Construct a scheduling decision model. Based on the deep reinforcement learning algorithm, map the intelligent scheduling constraints to the state space and action space. The state space contains real-time operation status information of the printing press group, such as the safety risk index, load condition, operation status, etc. of each printing press. The action space defines the set of executable scheduling operations, such as which printing press to assign a certain task to, or start / stop a certain printing press, etc. Construct a composite reward function, taking the safety threshold constraint, resource allocation constraint, and production efficiency constraint as evaluation indicators. For example, if the scheduling plan meets the safety threshold constraint, a positive reward is given; if it violates the resource allocation constraint, a negative reward is given; if the production efficiency is improved, a positive reward is given. Continuously optimize the scheduling strategy through the deep reinforcement learning algorithm to maximize the composite reward function. For example, through training, learn to preferentially assign high-load tasks to low-risk new devices, thereby improving production efficiency while ensuring safety.

[0041] Apply the trained scheduling decision model to actual production. According to the real-time operating status of the printing press fleet, the model will automatically generate the optimal scheduling plan and send it to the execution layer.

[0042] The solution of this application can: Improve safety: By conducting real-time risk assessment on the printing press fleet and formulating scheduling strategies in combination with equipment identification results, the safety risks in the production process can be effectively reduced and accidents can be avoided. For example, avoid assigning high-load tasks to high-risk old equipment, thereby reducing the probability of equipment failure. Optimize resource allocation: By constructing an equipment importance scoring matrix and a spare equipment allocation strategy, resource allocation can be optimized, equipment utilization rate can be improved, and resource waste can be avoided. For example, prioritize tasks to be assigned to equipment with better performance and higher reliability, and promptly allocate spare equipment to handle emergencies. Enhance production efficiency: By continuously optimizing the scheduling strategy through the deep reinforcement learning algorithm, production efficiency can be maximized under the premise of meeting safety constraints and resource constraints. For example, through learning, find the best task allocation plan, thereby reducing production time and increasing output.

[0043] In an alternative embodiment, use the scheduling decision model to generate an optimal scheduling plan, where the optimal scheduling plan includes the equipment start-stop sequence and the load distribution strategy; sending the optimal scheduling plan to each printing press control system for execution includes: Generate an optimal scheduling plan based on the scheduling decision model, and construct an equipment start-stop state transition matrix through the dynamic programming algorithm; optimize the load distribution strategy in combination with the mixed integer programming method. The load distribution strategy is based on the remaining production capacity of the equipment and the urgency of the tasks, establish a task allocation weight coefficient to achieve dynamic balance of the load of the equipment group; combine the equipment start-stop state transition matrix and the load distribution strategy to form the optimal scheduling plan; Construct a hierarchical scheduling instruction distribution system, where the hierarchical scheduling instruction distribution system parses the optimal scheduling plan into equipment-level control instructions; establish a real-time communication mechanism between the central scheduler and the local control system of the printing press, and send the equipment-level control instructions to each printing press control system.

[0044] In the data collection link, collect the historical operation data of the printing press fleet, including equipment operation status (such as: on-off time, operating speed, energy consumption, etc.), task information (such as: order type, quantity, delivery date, etc.), environmental parameters (such as: temperature, humidity, etc.), and quality data in the production process (such as: printing accuracy, color deviation, etc.). For example, collect the operation data of a certain printing press in the past week, including the on-off time, printing speed, energy consumption, the quantity and type of orders completed every day, and the corresponding quality inspection results, etc.

[0045] Train a hierarchical deep learning model using the collected historical data. This model adopts a hierarchical structure and extracts data features layer by layer. The bottom layer network learns the underlying operating features of the devices, such as the energy consumption patterns of individual devices; the middle layer network learns task features, such as the resource demand patterns of different order types; the top layer network fuses the features extracted by the bottom and middle layer networks and learns the complex correlation relationships between devices and tasks. For example, through the deep learning model, the production time and resource consumption of different order types on different printing presses can be predicted. Suppose an order requires printing 10,000 color brochures. The model predicts that it takes 8 hours and consumes 100 degrees of electricity on printing press A, and it takes 10 hours and consumes 120 degrees of electricity on printing press B.

[0046] Build a scheduling decision model. This model constructs a device start-stop state transition matrix based on the dynamic programming algorithm. For example, according to the historical data and prediction results, it can be calculated that it takes 30 minutes for printing press A to start from the shutdown state to the normal operating state, while it takes 45 minutes for printing press B. Combine the mixed integer programming method to optimize the load distribution strategy. This strategy establishes a task allocation weight coefficient based on the remaining production capacity of the devices and the urgency of the tasks, and realizes the dynamic balance of the load of the device group. For example, if an urgent order needs to be completed as soon as possible, a higher weight coefficient is assigned to it and production is arranged preferentially. Suppose the remaining production capacity of printing press A is 20 hours and that of printing press B is 15 hours. Then, according to the weight coefficient and the predicted production time, this urgent order is assigned to printing press A. Combine the device start-stop state transition matrix and the load distribution strategy to form an optimal scheduling plan, which includes the device start-stop sequence and the load distribution strategy.

[0047] Send the optimal scheduling plan to each printing press control system for execution. Build a hierarchical scheduling instruction distribution system to parse the optimal scheduling plan into device-level control instructions. For example, parse the instruction of "start printing press A" into specific control signals and send them to the control system of printing press A. Establish a real-time communication mechanism between the central scheduler and the local control system of the printing press to send the device-level control instructions to each printing press control system. Design an instruction execution confirmation mechanism and an emergency handling plan to ensure the reliable execution of the scheduling instructions. For example, if printing press A fails to start successfully after receiving the start instruction, the system will issue an alarm and start the emergency plan to reassign the task to other available printing presses.

[0048] Deploy a multi-dimensional monitoring system to collect execution effect data. This system records the operating status of devices, the progress of task execution, and product quality data in real time. For example, it monitors the running speed, temperature, energy consumption of Printer A in real time, as well as the completion progress of the current order and the quality inspection results. Use the multi-source data fusion method to preprocess the execution effect data and establish a data anomaly detection and compensation mechanism. For example, if it is detected that the printing speed of Printer A is lower than expected, the system will automatically adjust its operating parameters or issue an alarm to notify the maintenance personnel for repair. Feed the preprocessed execution effect data back to the data collection link for continuously optimizing the performance of the hierarchical deep learning model and the scheduling decision model. Update the feature extraction parameters and network structure weights of the hierarchical deep learning model through the incremental learning method; optimize the state value evaluation and policy network parameters of the scheduling decision model based on the online learning strategy to achieve continuous improvement of model performance and dynamic accumulation of knowledge.

[0049] The solution of this application can: Improve production efficiency: By optimizing the device start-stop sequence and load distribution strategy, minimize the device idle time and task waiting time, thereby improving production efficiency. For example, through intelligent scheduling, the daily printing task completion time can be shortened from 10 hours to 8 hours. Reduce production costs: By optimizing the device start-stop strategy, reduce the device energy consumption; by optimizing the load distribution strategy, improve the device utilization rate, thereby reducing production costs. For example, through intelligent scheduling, the daily energy consumption can be reduced from 1000 kWh to 800 kWh. Improve product quality: Through real-time monitoring and data analysis, promptly discover and correct abnormal situations in the production process to ensure the stability of product quality. For example, through intelligent scheduling, the product qualification rate can be increased from 95% to 98%.

[0050] In an optional implementation manner, generate an optimal scheduling plan based on the scheduling decision model, construct a device start-stop state transition matrix through the dynamic programming algorithm; optimize the load distribution strategy in combination with the mixed integer programming method, and the load distribution strategy is based on the remaining production capacity of the device and the urgency of the task, establish a task allocation weight coefficient to achieve dynamic balance of the device group load; combining the device start-stop state transition matrix and the load distribution strategy to form an optimal scheduling plan includes: The device start-stop state transition matrix includes the state transition law, state transition cost, and warm-up time constraint of the printer; use the value iteration algorithm to calculate the state value function and state action value function, and determine the optimal start-stop strategy based on the state value function and the state action value function; Calculate the remaining production capacity of the equipment based on the operating status of the printing press, and evaluate the urgency of the task according to the delivery time of the production task; establish a task allocation weight coefficient by using the remaining production capacity of the equipment and the urgency of the task, and the task allocation weight coefficient is used to sort the priorities of tasks and allocate resources; Construct a load balancing constraint condition based on the task allocation weight coefficient, and the load balancing constraint condition includes equipment production capacity limitation, task time sequence relationship and process parameter requirements; obtain a task allocation plan that meets the load balancing constraint condition by solving a mixed integer programming model, so as to realize the dynamic balance of the load of the printing press group.

[0051] Construct a state transition matrix for equipment startup and shutdown. This matrix describes the state changes of the printing press in different time periods, such as startup, shutdown, standby and preheating, etc. Each element in the matrix represents the cost of transitioning from one state to another, such as energy consumption, time cost, etc. In addition, the matrix also considers the preheating time constraint of the printing press, for example, it takes a certain preheating time to transition from the shutdown state to the startup state. For example, assume there are three printing presses, and the states are divided into shutdown, standby and running. Then a 3x3xT matrix can be constructed, where T represents the total number of time periods in the scheduling cycle. Each element in the matrix represents the cost of a certain printing press transitioning from one state to another in a specific time period. For example, the cost for the first printing press to transition from the shutdown state to the running state in the first time period is 10 (including preheating energy consumption and time cost), while the cost for transitioning from the standby state to the running state is 5 (only including time cost).

[0052] Use the value iteration algorithm to calculate the state value function and the state-action value function. The state value function represents the long-term value of the printing press being in a certain state in a certain time period. The state-action value function represents the long-term value after the printing press performs a certain action (such as startup, shutdown) in a certain time period. Through iterative calculation, the optimal value of each state and action can be obtained. For example, assume that the first printing press is in the shutdown state in the first time period, and its state value function is 50, representing the expected future income. If it is turned on, the state-action value function is 60, indicating that the expected future income after turning on is higher, so it should be turned on.

[0053] Based on the calculated state value function and state-action value function, the optimal startup and shutdown strategy can be determined. This strategy stipulates which state each printing press should be in in each time period to maximize the long-term value. For example, according to the previous calculation results, the first printing press should be turned on in the first time period.

[0054] Optimize the load distribution strategy. First, calculate the remaining production capacity of the equipment based on the operating status of the printing press. For example, when the first printing press is in the on state, it can print 1000 copies per hour. If a printing task of 500 copies has already been arranged, the remaining production capacity is 500 copies. Second, evaluate the urgency of the task according to the delivery time of the production task. For example, a printing task that needs to be completed within 2 hours is more urgent than a task that needs to be completed within 24 hours.

[0055] Establish a task allocation weight coefficient using the remaining production capacity of the equipment and the urgency of the task. This coefficient is used to prioritize tasks and allocate resources. For example, a task with high urgency and large remaining production capacity of the printing press will obtain a higher weight coefficient. Suppose a task has an urgency score of 8 (out of 10), and the remaining production capacity of the target printing press is 600 copies, then the task allocation weight coefficient can be calculated as 8 * 600 = 4800.

[0056] Construct load balancing constraint conditions based on the task allocation weight coefficient. This constraint condition includes equipment production capacity limitations, task timing relationships, and process parameter requirements. For example, a printing press cannot be arranged with tasks exceeding its production capacity within a certain time period, there is a sequence between certain tasks, and different types of printing tasks require different process parameters.

[0057] Through the analysis of the load balancing constraint conditions, obtain a task allocation plan that meets all constraint conditions, and achieve the dynamic balance of the load of the printing press group. For example, according to the weight coefficient ranking, prioritize the allocation of tasks to the printing press with large remaining production capacity and capable of meeting the task timing relationship and process parameter requirements.

[0058] Combine and match the equipment start-stop state transition matrix and the task allocation plan to construct a timing coordination mechanism. This mechanism ensures the consistency of the equipment start-stop timing and the task allocation timing. For example, when a task is allocated to a certain printing press, the printing press must be started before the task starts and choose to standby or shut down according to the subsequent task arrangement after the task is completed. Generate an optimal scheduling plan according to the timing coordination result, which stipulates the start-stop time of each printing press, the execution time of each task, and the allocated printing press.

[0059] The solution of this application can: Improve production efficiency: By optimizing task allocation and equipment start-stop, maximize the utilization of equipment production capacity, reduce idle time, and thus improve the overall production efficiency. Reduce energy consumption: Through intelligent control of equipment start-stop, avoid unnecessary energy waste and reduce production costs. Achieve load balancing: Through the task allocation weight coefficient and load balancing constraint conditions, make the load of the printing press group more balanced, avoid the situation where some equipment is overloaded while some other equipment is idle, and extend the service life of the equipment.

[0060] Figure 2 This is a schematic structural diagram of a safety production control system for a printing press device according to an embodiment of the present invention. As Figure 2 shown, the system includes: A first unit, configured to collect operation data in real time through a multi-modal sensor network deployed in a printing press cluster, extract features from the standardized data stream corresponding to the operation data by using a temporal attention mechanism, and obtain a dynamic feature vector of the printing press cluster; construct a spatio-temporal correlation matrix of the printing press cluster based on the dynamic feature vector; A second unit, configured to construct a hierarchical deep learning model based on the dynamic feature vector and the spatio-temporal correlation matrix, input historical labeled data into the hierarchical deep learning model for training, analyze the dynamic feature vector by using the trained deep learning model, and generate an evaluation result of the device health state; calculate an overall safety risk index and a device identification result of the printing press cluster based on the device health state evaluation result and the spatio-temporal correlation matrix; A third unit, configured to establish intelligent scheduling constraint conditions for the printing press cluster according to the overall safety risk index and the device identification result; construct a scheduling decision model based on a deep reinforcement learning algorithm, and use safety threshold constraints, resource allocation constraints, and production efficiency constraints as evaluation indicators of a reward function; generate an optimal scheduling plan by using the scheduling decision model, where the optimal scheduling plan includes an equipment start-stop sequence and a load allocation strategy; send the optimal scheduling plan to each printing press control system for execution.

[0061] In a third aspect of the embodiments of the present invention, there is provided an electronic device, including: a processor; a memory for storing instructions executable by the processor; wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0062] In a fourth aspect of the embodiments of the present invention, there is provided a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0063] The present invention may be a method, an apparatus, a system, and / or a computer program product. The computer program product may include a computer-readable storage medium, on which computer-readable program instructions for executing various aspects of the present invention are loaded.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A safety production control method for a printing press device, characterized in that Including: Real-time collect operation data through a multi-modal sensor network deployed in a printing press group, and use a temporal attention mechanism to extract features from the standardized data stream corresponding to the operation data to obtain a dynamic feature vector of the printing press group; construct a spatio-temporal correlation matrix of the printing press group based on the dynamic feature vector; Based on the dynamic feature vector and the spatio-temporal correlation matrix, construct a hierarchical deep learning model, input historical labeled data into the hierarchical deep learning model for training, and use the trained deep learning model to analyze the dynamic feature vector to generate an equipment health status evaluation result; Based on the equipment health status evaluation result and the spatio-temporal correlation matrix, calculate the overall safety risk index and equipment identification result of the printing press group; According to the overall safety risk index and the equipment identification result, establish intelligent scheduling constraint conditions for the printing press group; construct a scheduling decision model based on a deep reinforcement learning algorithm, and use safety threshold constraints, resource allocation constraints, and production efficiency constraints as evaluation indicators of the reward function; use the scheduling decision model to generate an optimal scheduling plan, where the optimal scheduling plan includes equipment start-stop sequence and load allocation strategy; send the optimal scheduling plan to each printing press control system for execution.

2. The method according to claim 1, characterized in that, Based on the dynamic feature vector and the spatio-temporal correlation matrix, construct a hierarchical deep learning model. Training by inputting historical labeled data into the hierarchical deep learning model includes: The hierarchical deep learning model includes a feature fusion layer, a temporal prediction layer, and a risk assessment layer; in the feature fusion layer, convert the dynamic feature vector into a query matrix, a key matrix, and a value matrix through a self-attention mechanism, and calculate the feature inter-correlation weight through scaled dot-product attention. Input the spatio-temporal correlation matrix into a graph attention network to extract the spatial dependence features between printing presses, and adaptively fuse the feature inter-correlation weight and the spatial dependence features through a gating mechanism to generate a first comprehensive feature representation; In the temporal prediction layer, input the first comprehensive feature representation into a bidirectional long short-term memory network and a multi-scale parallel convolutional neural network respectively. The bidirectional long short-term memory network extracts temporal features, and the multi-scale parallel convolutional neural network extracts local features. Fuse the output features of the two through a residual connection structure to generate a second comprehensive feature representation; Input the second comprehensive feature representation into the risk assessment layer to generate a second prediction result; Input historical labeled data into the hierarchical deep learning model for training. Use the mean square error loss function to calculate the first loss value of the first prediction result, and use the cross-entropy loss function to calculate the second loss value of the second prediction result; adaptively balance the first loss value and the second loss value through learnable weights to obtain a total loss value, and optimize the parameters of the hierarchical deep learning model based on the total loss value to obtain the trained deep learning model.

3. The method according to claim 1, wherein Use the trained deep learning model to analyze the dynamic feature vector to generate an equipment health status evaluation result; Calculating the overall safety risk index and equipment identification results of the printing press group based on the equipment health status evaluation results and the spatio-temporal association matrix includes: Analyzing the dynamic feature vectors using the trained deep learning model, and fusing multiple prediction results through weighted voting to generate equipment health status evaluation results; Constructing a weighted directed graph based on the equipment health status evaluation results and the spatio-temporal association matrix, where the nodes of the weighted directed graph represent printing press equipment, and the edge weights represent the risk propagation intensity between equipment. Using the PageRank algorithm to calculate the node importance scores; Combining the node importance scores with the local risk accumulation function, and introducing a decay factor in the time dimension and graph convolution features in the space dimension into the local risk accumulation function to calculate the overall safety risk index of the printing press group; Conducting multi-criteria decision-making analysis based on the weighted directed graph, and identifying equipment through the fuzzy comprehensive evaluation method to generate equipment identification results.

4. The method according to claim 1, characterized in that, Establishing the intelligent scheduling constraint conditions for the printing press group according to the overall safety risk index and the equipment identification results; Constructing a scheduling decision model based on the deep reinforcement learning algorithm, and taking the safety threshold constraint, resource allocation constraint, and production efficiency constraint as the evaluation indicators of the reward function includes: Setting a risk threshold vector according to the overall safety risk index, where the risk threshold vector is used to divide the risk area higher than the preset risk threshold and the risk area lower than the preset risk threshold, and setting corresponding scheduling restriction strategies for different risk areas; Constructing a resource allocation constraint based on the equipment identification results, where the resource allocation constraint includes an equipment importance scoring matrix and a spare equipment deployment strategy, and the equipment importance scoring matrix is used to quantify the importance of equipment in the production system; Combining the scheduling restriction strategy and the resource allocation constraint to form the intelligent scheduling constraint conditions for the printing press group; Constructing a scheduling decision model based on the deep reinforcement learning algorithm, mapping the intelligent scheduling constraint conditions to the state space and action space, where the state space includes the real-time operating state information of the printing press group, and the action space defines the set of executable scheduling operations; Constructing a composite reward function, and taking the safety threshold constraint, resource allocation constraint, and production efficiency constraint as the evaluation indicators of the composite reward function.

5. The method according to claim 1, characterized in that Generating an optimal scheduling plan using the scheduling decision model, where the optimal scheduling plan includes the equipment start-stop sequence and the load distribution strategy; Issuing the optimal scheduling plan to each printing press control system for execution includes: Generating an optimal scheduling plan based on the scheduling decision model, constructing an equipment start-stop state transition matrix through the dynamic programming algorithm; Optimizing the load distribution strategy by combining the mixed integer programming method, where the load distribution strategy is based on the remaining production capacity of the equipment and the urgency of the task, establishing a task allocation weight coefficient to achieve the dynamic balance of the equipment group load; Combining the equipment start-stop state transition matrix and the load distribution strategy to form the optimal scheduling plan; Build a hierarchical scheduling instruction distribution system, which parses the optimal scheduling scheme into device-level control instructions; establish a real-time communication mechanism between the central scheduler and the local control system of the printing press, and send the device-level control instructions to each printing press control system.

6. The method according to claim 5, characterized in that Generate an optimal scheduling scheme based on the scheduling decision model, and construct a device start-stop state transition matrix through the dynamic programming algorithm; optimize the load distribution strategy by combining the mixed integer programming method. The load distribution strategy is based on the remaining production capacity of the device and the urgency of the task, and establish a task allocation weight coefficient to achieve dynamic balance of the load of the device group. Combining the device start-stop state transition matrix and the load distribution strategy to form an optimal scheduling scheme includes: The device start-stop state transition matrix includes the state transition law, state transition cost, and warm-up time constraint of the printing press; use the value iteration algorithm to calculate the state value function and state-action value function, and determine the optimal start-stop strategy based on the state value function and the state-action value function. Calculate the remaining production capacity of the device based on the operating state of the printing press, and evaluate the urgency of the task according to the delivery time of the production task; establish a task allocation weight coefficient using the remaining production capacity of the device and the urgency of the task. The task allocation weight coefficient is used to sort the priorities of tasks and allocate resources. Construct a load balancing constraint condition based on the task allocation weight coefficient. The load balancing constraint condition includes device production capacity limitation, task timing relationship, and process parameter requirements; obtain a task allocation scheme that meets the load balancing constraint condition by solving the mixed integer programming model, and achieve dynamic balance of the load of the printing press group.

7. A safety production control system for a printing press device, used to implement the method described in any one of the preceding claims 1-6, characterized in that, Include: The first unit is used to collect operation data in real time through a multi-modal sensor network deployed in the printing press group, extract features from the standardized data stream corresponding to the operation data using the temporal attention mechanism, and obtain the dynamic feature vector of the printing press group; construct a spatio-temporal correlation matrix of the printing press group based on the dynamic feature vector. The second unit is used to construct a hierarchical deep learning model based on the dynamic feature vector and the spatio-temporal correlation matrix, input historical labeled data into the hierarchical deep learning model for training, and analyze the dynamic feature vector using the trained deep learning model to generate an evaluation result of the device health status. Calculate the overall safety risk index and device identification result of the printing press group based on the device health status evaluation result and the spatio-temporal correlation matrix. The third unit is used to establish the intelligent scheduling constraint conditions of the printing press group according to the overall safety risk index and the device identification result. Construct a scheduling decision model based on the deep reinforcement learning algorithm, and use safety threshold constraints, resource allocation constraints, and production efficiency constraints as evaluation indicators of the reward function; generate an optimal scheduling scheme using the scheduling decision model. The optimal scheduling scheme includes the device start-stop sequence and the load distribution strategy; send the optimal scheduling scheme to each printing press control system for execution.

8. An electronic device, characterized in that, Include: Processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, the method according to any one of claims 1 to 6 is implemented.

Images