Cloud-edge collaborative forging equipment cluster state perception linkage control method and system

Abstract

This invention relates to the field of linkage control technology for forging and pressing equipment, and discloses a cloud-edge collaborative method and system for state-aware linkage control of forging and pressing equipment clusters. The method involves real-time collection and processing of operational status data by edge computing nodes deployed on each forging and pressing equipment, generating equipment status assessment results, and uploading them to the cloud control center. Based on full cluster data, historical operational data, and production plans, the cloud utilizes a multi-objective optimization algorithm to generate a collaborative control strategy, which is then distributed to edge nodes for linkage adjustments, forming a closed-loop control. This invention effectively improves the production efficiency of forging and pressing equipment clusters and reduces energy consumption.

Description

Technical Field

[0001] This invention relates to the field of linkage control technology for forging and pressing equipment, specifically to a cloud-edge collaborative method and system for state perception and linkage control of forging and pressing equipment clusters. Background Technology

[0002] In traditional forging equipment cluster control, each piece of equipment operates independently, lacking status awareness and linkage mechanisms, resulting in limited overall production efficiency and high energy consumption. Existing technologies mostly focus on single-equipment control, with limited research on collaborative control at the cluster level. Especially in complex processes, the mutual influence between equipment is significant, and single-point control cannot meet the global optimization requirements. With the development of Industry 4.0, cloud-edge collaborative technology provides a new approach to cluster control, but how to achieve efficient and low-energy cluster linkage control remains a technical challenge. Summary of the Invention

[0003] To address the aforementioned technical shortcomings, the purpose of this invention is to provide a cloud-edge collaborative state perception and linkage control method and system for forging equipment clusters, solving the problems of low control efficiency, high energy consumption, and lack of linkage in existing forging equipment clusters.

[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: In a first aspect, the present invention provides a cloud-edge collaborative state-aware linkage control method for a forging equipment cluster, the method comprising: By collecting real-time operating status data from edge computing nodes deployed on each forging and pressing equipment, and performing data cleaning and feature extraction at the edge side, equipment status assessment results are generated. Upload the device status assessment results and real-time device control parameters generated by each edge computing node to the cloud control center; Based on the received status assessment results of all cluster devices and real-time control parameters, the cloud control center integrates historical operating data and production plans, and uses collaborative control algorithms to perform global analysis and generate collaborative control strategies. The collaborative control strategy is sent to the corresponding edge computing nodes. The edge computing nodes adjust the control parameters of their own devices and adjacent devices with process connections with them in accordance with the received strategy instructions, so as to realize the linkage control at the cluster level. Edge computing nodes feed back the execution results of control commands to the cloud, forming a closed-loop control circuit.

[0005] Preferably, in one possible implementation of the first aspect, the data cleaning includes outlier detection and missing value imputation of the real-time collected operating status data; The outlier detection employs a statistical method based on dynamic thresholds, calculating the mean within a sliding window for the collected data sequence. and standard deviation If the current data point satisfy If the value is an outlier, it will be removed. For time indexing, To adjust the sliding window size, The sensitivity coefficient is adjustable. The missing value filling method uses linear interpolation of adjacent valid values.

[0006] Preferably, in one possible implementation of the first aspect, the feature extraction includes time-domain and frequency-domain feature extraction of the cleaned running status data; The extracted time-domain features include peak value and root mean square value; the peak factor is calculated. ,in , The peak value of the operating status signal. The root mean square value of the operating status signal; Frequency domain features are obtained by performing a Fast Fourier Transform on the signal to obtain the power spectrum, and then extracting the power spectral entropy of the dominant frequency components. Its calculation formula is ,in For the first The percentage of power of each frequency component in the total power. The number of dominant frequency components; The result of the feature extraction is to generate a peak factor. and the power spectral entropy The eigenvectors formed.

[0007] Preferably, in one possible implementation of the first aspect, the generation device status assessment includes: The feature vectors are input into a trained deep belief network model to calculate the device health status index. ; The deep belief network is composed of multiple layers of restricted Boltzmann machines stacked together, and the calculation formula for its output layer is as follows: ,in It is the Sigmoid activation function. The weight matrix of the output layer. This is the output of the last hidden layer. This is the bias vector for the output layer. For network layer indexing; Equipment condition assessment results based on The values ​​are categorized into three levels: normal, attention, and abnormal.

[0008] Preferably, in one possible implementation of the first aspect, the collaborative control algorithm adopts a multi-objective optimization decision-making method, comprehensively considering equipment status assessment results, real-time control parameters, historical operating data and production plans, with the goal of maximizing the overall efficiency of the cluster and minimizing energy consumption, generating non-dominated solutions through Pareto optimal solution set search, and selecting solutions based on expert system rules.

[0009] Preferably, in one possible implementation of the first aspect, the multi-objective optimization decision method employs an improved NSGA-II algorithm, introducing adaptive crossover probability; The adaptive crossover probability The calculation formula is:

[0010] in This represents the highest fitness value in the current population. This represents the current average fitness value of the population. The larger fitness value among the two individuals at the crossover. , This is a preset constant.

[0011] Preferably, in one possible implementation of the first aspect, the process of generating the collaborative control strategy includes: The Pareto optimal solution set obtained by multi-objective optimization is mapped as the antecedent of fuzzy rules, and the adjustment amount of equipment control parameters is used as the consequent to construct a Takagi-Sugeno type fuzzy inference system. For each optimal solution, the corresponding fuzzy rule form is: if the device state is And the production efficiency is Then the control parameter adjustment amount is ; The final policy output is obtained by a weighted average of all activated rules:

[0012] in To control the adjustment amount of the parameters, The number of rules activated. For the first The activation strength of the rule, This refers to the adjustment amount of the control parameters output by this rule.

[0013] Preferably, in one possible implementation of the first aspect, the linkage adjustment includes: After receiving the collaborative control strategy, the edge computing node parses the strategy instructions and obtains the target control parameters of its own device and the adjustment amount of the associated parameters of adjacent devices. Based on the process correlation model between equipment, calculate the adjustment sequence of the control parameters of this equipment, and simultaneously send adjustment instructions to adjacent equipment.

[0014] Preferably, in one possible implementation of the first aspect, the process association model is represented by a directed graph, where nodes represent equipment and edges represent process flows; For each device node The final adjustment amount of its control parameters for:

[0015] in It is sent from the cloud to the device. The original strategy adjustment amount, It is equipment The set of precursor equipment in the process correlation diagram and respectively equipment and its precursor equipment Health status index It is equipment For equipment The influence weight of the process, This is a correction factor based on process stability.

[0016] Secondly, the present invention provides a cloud-edge collaborative forging equipment cluster state perception and linkage control system, the system being used to implement the cloud-edge collaborative forging equipment cluster state perception and linkage control method as described in the first aspect, including: The data acquisition module is used to collect real-time operating status data through edge computing nodes deployed on each forging and pressing equipment, and to perform data cleaning and feature extraction at the edge side to generate equipment status assessment results. The data upload module is used to upload the device status assessment results and real-time control parameters of the devices generated by each edge computing node to the cloud control center. The cloud-based analytics module is used to perform global analysis based on the received status assessment results of all cluster devices and real-time control parameters of the devices, integrating historical operating data and production plans, and using collaborative control algorithms to generate collaborative control strategies. The linkage control module is used to send the collaborative control strategy to the corresponding edge computing nodes, and control the edge computing nodes to adjust the control parameters of the device and adjacent devices with process association with it in accordance with the received strategy instructions. The feedback module is used to send the execution effect of control commands back to the cloud, forming a closed-loop control circuit.

[0017] The beneficial effects of this invention are as follows: Through a cloud-edge collaborative architecture, this invention realizes real-time status perception and linkage control of forging equipment clusters, significantly improving production efficiency and reducing energy consumption.

[0018] By utilizing edge computing nodes for data cleaning and feature extraction, the burden on the cloud can be reduced and the data processing speed can be improved.

[0019] The cloud-based system uses a multi-objective optimization algorithm to generate a collaborative control strategy, taking into account equipment status, production efficiency, and energy consumption to achieve global optimization.

[0020] By continuously optimizing the control strategy through a closed-loop control circuit, the system's adaptability and robustness can be improved.

[0021] This invention effectively solves the efficiency and energy consumption problems in traditional cluster control and promotes the intelligent upgrading of the forging industry. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This application provides a flowchart of a cloud-edge collaborative forging equipment cluster state perception and linkage control method.

[0024] Figure 2 This application provides a structural diagram of a cloud-edge collaborative forging equipment cluster state perception and linkage control system.

[0025] Attached reference numerals: 1-Data acquisition module, 2-Data upload module, 3-Cloud analysis module, 4-Linkage control module, 5-Feedback module. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Example 1: As Figure 1 As shown, this invention provides a cloud-edge collaborative state-aware linkage control method for forging equipment clusters, including: By collecting real-time operating status data from edge computing nodes deployed on each forging and pressing equipment, and performing data cleaning and feature extraction at the edge, equipment status assessment results are generated.

[0028] In this embodiment, the edge computing node is connected to sensors on the forging equipment, including vibration sensors, temperature sensors, and pressure sensors, to continuously collect raw data sequences during equipment operation at a high-frequency sampling rate.

[0029] The data cleaning process begins with outlier detection, employing a statistical method based on dynamic thresholds. For the collected data sequences, the mean within a sliding window is calculated. and standard deviation , For time indexing, To adjust the sliding window size, For adjustable sensitivity coefficients, if the current data point satisfy If the data point is an outlier, it will be removed. Sliding window size The sensitivity coefficient is set according to the equipment's operating cycle; in this embodiment, it is determined by the sensitivity coefficient. A value of 3 is chosen to balance the sensitivity and stability of the detection. Subsequently, for any missing values ​​in the data sequence, linear interpolation of adjacent valid values ​​is used to fill them in. Specifically, for the two valid data points before and after the missing position, an estimated value is inserted through linear calculation to ensure the continuity of the data sequence.

[0030] After data cleaning, feature extraction is performed to extract key features characterizing the equipment status from the cleaned operational status data. Feature extraction includes both time-domain and frequency-domain features. In time-domain feature extraction, the peak value of the operational status signal is calculated. and root mean square value And calculate the peak factor. Its calculation formula is The peak factor reflects the impulse characteristics of a signal and is used to identify whether an abnormal impulse has occurred in the equipment.

[0031] In frequency domain feature extraction, the power spectrum is first obtained by performing a fast Fourier transform on the signal, and then the power spectral entropy of the dominant frequency components is extracted. The formula for calculating the power spectral entropy is: ,in For the first The percentage of power of each frequency component in the total power. To determine the dominant frequency components, this embodiment selects the top 10 main frequency components from the power spectrum. Power spectral entropy reflects the complexity of the signal in the frequency domain; a higher value indicates greater device instability. Ultimately, the feature extraction results in generating a peak factor... and power spectral entropy The eigenvectors formed.

[0032] The process of generating equipment condition assessment involves inputting the aforementioned feature vectors into a trained deep belief network model to calculate the equipment health status index H. A deep belief network is a deep learning model whose structure includes an input layer, at least one hidden layer, and an output layer. The input layer receives a feature vector consisting of two features: peak factor and power spectral entropy. Each hidden layer consists of a restricted Boltzmann machine. The output layer is calculated using the following formula: ,in It is the Sigmoid activation function. This is the weight matrix of the output layer. This is the output of the last hidden layer. This is the bias vector for the output layer. For network layer indexes.

[0033] This deep belief network model was pre-trained using historical data, including feature vectors from various device states such as normal and abnormal. Through unsupervised pre-training and supervised fine-tuning, the network parameters were optimized, enabling the model to accurately map feature vectors to a health status index. The device status assessment results are based on... The value is divided into three levels: when When the value is greater than or equal to 0.8, the status is normal; when... When the value is between 0.5 and 0.8, the status is "Attention"; when... When the value is less than 0.5, the status is abnormal.

[0034] The device status assessment results and real-time control parameters generated by each edge computing node are uploaded to the cloud control center.

[0035] In this embodiment, the edge computing node uploads the device status assessment results and real-time device control parameters to the cloud control center via its built-in network communication module. The device status assessment results include a health status index calculated by a deep belief network model. and based on The values ​​are categorized into status levels: normal, warning, or abnormal. Real-time control parameters cover the operational variables of the forging equipment during operation, including pressure setpoints, temperature monitoring values, and motion speed parameters.

[0036] The upload process uses Industrial Ethernet as the communication medium, and reliable data packet transmission is achieved based on Transmission Control Protocol (TCP) and Internet Protocol (IP). Data messages are encapsulated in a structured format, including a unique device identifier, timestamp, health status index value, status level code, and specific values ​​of real-time control parameters. The upload frequency is dynamically adjusted according to the device's health status: once every 5 seconds under normal conditions, once every 1 second under monitoring conditions, and once every 200 milliseconds under abnormal conditions.

[0037] To cope with network outages or fluctuations, edge computing nodes integrate a data caching mechanism to temporarily store data that was not successfully uploaded and automatically retransmit it after the network is restored, with a maximum of three retries. For security, a transport layer security protocol is applied to encrypt the communication channel end-to-end to prevent data leakage or tampering. The cloud control center is equipped with a data receiving interface that parses incoming message content in real time and persistently stores it in a distributed time-series database.

[0038] Based on the received status assessment results of all cluster devices and real-time control parameters, the cloud control center integrates historical operating data and production plans, and uses collaborative control algorithms to perform global analysis to generate collaborative control strategies.

[0039] In this embodiment, after receiving the device status assessment results and real-time control parameters uploaded from all edge computing nodes in the cluster, the cloud control center first performs data fusion processing. The device status assessment results include the health status index of each forging press. And the corresponding status level, namely normal, warning or abnormal, while the real-time control parameters of the equipment include pressure setpoint, temperature monitoring value and motion speed parameters, etc.

[0040] The cloud-based control center integrates this real-time data with historical operational data stored in a distributed database and current production plan information. The historical operational data covers long-term equipment operating status trends, fault records, and maintenance history, while the production plan includes order requirements, delivery timelines, and capacity allocation targets. The integration process employs data association technology, aligning multi-source data through unique equipment identifiers and timestamps to form a complete cluster status view.

[0041] The collaborative control algorithm employs a multi-objective optimization decision-making method, with the core optimization objectives of maximizing overall cluster efficiency and minimizing energy consumption, while comprehensively considering equipment status assessment results, real-time control parameters, historical operating data, and production plan constraints. The efficiency maximization objective function is modeled based on equipment output rate and running time, while the energy consumption minimization objective function is constructed using power consumption data. The multi-objective optimization problem is transformed into finding the Pareto optimal solution set, i.e., the non-dominated solution set, where improvement in any objective does not lead to deterioration of other objectives. The optimization process uses a modified NSGA-II algorithm, which introduces an adaptive crossover probability mechanism to improve search efficiency. The calculation formula is:

[0042] in This represents the highest fitness value in the current population. This represents the current average fitness value of the population. The larger fitness value among the two individuals at the crossover. , As a preset constant, in this embodiment The value is 1.0. The value is set to 0.5. During algorithm initialization, a population is randomly generated, with each individual representing a set of possible control parameter configurations. The population evolves iteratively through crossover, mutation, and selection operations. The crossover operation uses simulated binary crossover, the mutation operation uses polynomial mutation, and the selection operation is based on non-dominated sorting and crowding calculation to ensure population diversity and convergence.

[0043] After generating the Pareto optimal solution set, solutions are selected based on expert system rules. The expert system has a built-in domain knowledge base, which includes equipment operation specifications, process constraints, and priority rules. For example, when the equipment health status is abnormal, solutions that reduce load are preferred; when the production plan is urgent, the focus is on maximizing efficiency. The selected Pareto optimal solution set is then mapped to form the basis for generating the collaborative control strategy.

[0044] Specifically, each optimal solution is mapped to a fuzzy rule in a Takagi-Sugeno type fuzzy inference system. The antecedent of the rule corresponds to input variables such as equipment state and production efficiency, while the consequent corresponds to control parameter adjustments. Equipment state variables are defined as fuzzy sets such as normal, attention, and abnormal; production efficiency variables are defined as fuzzy sets such as low, medium, and high. Both are fuzzified using triangular membership functions. For each optimal solution, the corresponding fuzzy rule takes the form: if the equipment state is... And the production efficiency is Then the control parameter adjustment amount is The fuzzy inference system aggregates the outputs of all activated rules using a weighted average method, controlling the adjustment of parameters. The calculation formula is as follows:

[0045] Where is the adjustment amount of the control parameter, is the number of activated rules, is the activation strength of the -th rule, which is calculated by the minimum operator from the membership degree values of the antecedent variables, is the adjustment amount of the control parameter output by this rule. In this embodiment, the adjustment amount of the control parameter includes the adjustment ranges of pressure, temperature, and speed. The output of the fuzzy inference system generates a specific cooperative control strategy after defuzzification.

[0046] The cloud control center encodes the generated cooperative control strategy into a structured message and sends it to the corresponding edge computing node through a secure communication protocol. The strategy generation process is carried out in real time. At the same time, the cloud monitors the execution effect of the strategy, combines the feedback data, and continuously optimizes the algorithm parameters to ensure the self - adaptability and robustness of the control strategy.

[0047] Send the cooperative control strategy to the corresponding edge computing node. The edge computing node联动 adjusts the control parameters of this device and its adjacent devices with process associations according to the received strategy instructions to achieve联动 control at the cluster level.

[0048] In this embodiment, the cloud control center sends the generated cooperative control strategy to the corresponding edge computing node through a secure communication protocol. After receiving the strategy, the edge computing node first performs parsing and processing of the strategy instructions. The parsing process includes decoding the structured message, extracting the target control parameters of this device and the associated parameter adjustment amounts of adjacent devices. The target control parameters involve the key operating variables of the forging equipment, including the pressure setting value, temperature monitoring value, and motion speed parameter, while the associated parameter adjustment amount indicates the required change amplitude of adjacent devices. After parsing, the edge computing node calculates the adjustment sequence of the control parameters of this device according to the process association model between devices and synchronously sends adjustment instructions to adjacent devices to achieve联动 control at the cluster level.

[0049] The process association model is represented by a directed graph, where the nodes represent forging equipment and the edges represent the process flow direction. The structure of the directed graph reflects the dependency relationship between devices. For example, in a continuous production line, the output of the predecessor device is directly used as the input of the successor device. Each device node has a clear set of predecessor devices in the graph, and this set includes all the superior devices that directly affect the operation of device . The process influence weight is used to quantify the influence of device on device The degree of influence is predetermined based on historical process data and production experience, and is determined by analyzing the material flow, energy flow, and information flow between equipment. Weighting The value ranges from 0 to 1, with a larger value indicating a more significant impact. This is a correction coefficient based on process stability. Adjustments are made dynamically based on the overall system operation status, such as during production line startup or shutdown. Increased stability, while in steady-state operation, The value is reduced to improve response speed.

[0050] For each device node The final adjustment amount of its control parameters Calculated using the following formula:

[0051] in It is the original policy adjustment amount sent from the cloud to device i, which comes directly from the output of the collaborative control policy. and respectively equipment and its precursor equipment The health status index, calculated by a deep belief network model, ranges from 0 to 1, with a higher value indicating better health. The higher the value, the better. (In the formula...) This reflects the corrective effect of the preceding equipment's status on the adjustment amount of this equipment. When Greater than When the condition of the preceding equipment is better than that of the current equipment, the correction term is positive, thus increasing the adjustment amount to utilize the superior condition of the preceding equipment; conversely, when the condition of the preceding equipment is worse, the correction term is positive. Less than When the correction term is negative, the adjustment amount is reduced to compensate for the potential risks of the preceding equipment.

[0052] Before calculating the adjustment amount, the edge computing node verifies the completeness and rationality of the received policy instructions. The verification process includes checking whether the parameter range is within the device's safety limits and whether the adjustment instructions are compatible with the current operating mode. If an anomaly is detected, the edge computing node triggers local safety logic, suspends the adjustment, and immediately sends an alert to the cloud. The calculation of the adjustment sequence considers timing factors to ensure synchronization of actions across multiple devices. For example, for process chains with a strict sequence, the adjustment sequence is decomposed into multiple time steps, each step corresponding to a control cycle. The edge computing node assigns a timestamp to each adjustment instruction based on device response characteristics and communication latency, thereby generating an ordered adjustment sequence.

[0053] When synchronously sending adjustment commands to adjacent devices, edge computing nodes employ a reliable communication mechanism. The command message includes the target device identifier, adjustment parameter type, adjustment value, and timestamp, and is transmitted via industrial Ethernet. To cope with network jitter, edge computing nodes implement a retransmission strategy to ensure command delivery. Simultaneously, the adjustment of the device's control parameters is based on calculated... The value is executed directly and acts on the equipment controller, such as adjusting the opening of hydraulic valves or the speed of motors.

[0054] Edge computing nodes feed back the execution results of control commands to the cloud, forming a closed-loop control circuit.

[0055] In this embodiment, after executing control commands, the edge computing node monitors the real-time changes in the operating status of the forging equipment and collects execution effect data. Feedback data includes real-time operating status data of the equipment after control parameter adjustments and a health status index. And key performance indicators. Edge computing nodes continuously collect this data through connected temperature, pressure, and vibration sensors to ensure the accuracy and timeliness of feedback information.

[0056] The collected execution performance data is encapsulated into structured messages, including a unique device identifier, timestamp, health status index value, actual adjustment amount of control parameters, and performance index value. Edge computing nodes upload the feedback data to the cloud control center via industrial Ethernet using transmission control protocols and internet protocols.

[0057] After receiving feedback data, the cloud-based control center integrates and analyzes it with historical operational data and production plans. The cloud utilizes this feedback data to evaluate the actual effectiveness of the collaborative control strategy, for example, by comparing equipment efficiency and energy consumption before and after adjustments to quantify strategy performance. Based on the evaluation results, the cloud optimizes the parameters of the collaborative control algorithm, such as adjusting the weighting coefficients in multi-objective optimization decision-making or updating the rule base of the fuzzy inference system, thereby adaptively improving the generation of subsequent control strategies.

[0058] Example 2: Figure 2 As shown, the present invention provides a cloud-edge collaborative forging equipment cluster status perception and linkage control system, comprising: Data acquisition module 1 is used to collect real-time operating status data through edge computing nodes deployed on each forging and pressing equipment, and to perform data cleaning and feature extraction at the edge side to generate equipment status assessment results.

[0059] Data upload module 2 is used to upload the device status assessment results and real-time control parameters of each edge computing node to the cloud control center.

[0060] The cloud-based analysis module 3 is used to perform global analysis based on the received status assessment results of all cluster devices and real-time control parameters of the devices, integrate historical operating data and production plans, and use collaborative control algorithms to generate collaborative control strategies.

[0061] The linkage control module 4 is used to send the collaborative control strategy to the corresponding edge computing nodes, and control the edge computing nodes to adjust the control parameters of the device and adjacent devices with process associations with it according to the received strategy instructions.

[0062] Feedback module 5 is used to send the execution effect of control commands back to the cloud to form a closed-loop control circuit.

[0063] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A cloud-edge collaborative state-sensing and linkage control method for forging equipment clusters, characterized in that, The method includes: By collecting real-time operating status data from edge computing nodes deployed on each forging and pressing equipment, and performing data cleaning and feature extraction at the edge side, equipment status assessment results are generated. Upload the device status assessment results and real-time device control parameters generated by each edge computing node to the cloud control center; Based on the received status assessment results of all cluster devices and real-time control parameters, the cloud control center integrates historical operating data and production plans, and uses collaborative control algorithms to perform global analysis and generate collaborative control strategies. The collaborative control strategy is sent to the corresponding edge computing nodes. The edge computing nodes adjust the control parameters of their own devices and adjacent devices with process connections with them in accordance with the received strategy instructions, so as to realize the linkage control at the cluster level. Edge computing nodes feed back the execution results of control commands to the cloud, forming a closed-loop control circuit.

2. The cloud-edge collaborative forging equipment cluster state perception and linkage control method as described in claim 1, characterized in that, The data cleaning includes outlier detection and missing value imputation of the real-time collected operational status data; The outlier detection employs a statistical method based on dynamic thresholds, calculating the mean within a sliding window for the collected data sequence. and standard deviation If the current data point satisfy If the value is an outlier, it will be removed. For time indexing, To adjust the sliding window size, The sensitivity coefficient is adjustable. The missing value filling method uses linear interpolation of adjacent valid values.

3. The cloud-edge collaborative forging equipment cluster state perception and linkage control method as described in claim 2, characterized in that, The feature extraction includes time-domain and frequency-domain feature extraction of the cleaned operational status data; The extracted time-domain features include peak value and root mean square value; the peak factor is calculated. ,in , The peak value of the operating status signal. The root mean square value of the operating status signal; Frequency domain features are obtained by performing a Fast Fourier Transform on the signal to obtain the power spectrum, and then extracting the power spectral entropy of the dominant frequency components. Its calculation formula is ,in For the first The percentage of power of each frequency component in the total power. The number of dominant frequency components; The result of the feature extraction is to generate a peak factor. and the power spectral entropy The eigenvectors formed.

4. The cloud-edge collaborative forging equipment cluster state perception and linkage control method as described in claim 3, characterized in that, The device status assessment includes: The feature vectors are input into a trained deep belief network model to calculate the device health status index. ; The deep belief network is composed of multiple layers of restricted Boltzmann machines stacked together, and the calculation formula for its output layer is as follows: ,in It is the Sigmoid activation function. This is the weight matrix of the output layer. This is the output of the last hidden layer. This is the bias vector for the output layer. For network layer indexing; Equipment condition assessment results based on The values ​​are categorized into three levels: normal, attention, and abnormal.

5. The cloud-edge collaborative forging equipment cluster state perception and linkage control method as described in claim 1, characterized in that, The collaborative control algorithm adopts a multi-objective optimization decision-making method, which comprehensively considers equipment status assessment results, real-time control parameters, historical operating data and production plans. With the goal of maximizing the overall efficiency of the cluster and minimizing energy consumption, it generates non-dominated solutions through Pareto optimal solution set search and selects solutions based on expert system rules.

6. The cloud-edge collaborative forging equipment cluster state perception and linkage control method as described in claim 5, characterized in that, The multi-objective optimization decision-making method adopts an improved NSGA-II algorithm and introduces adaptive crossover probability; The adaptive crossover probability The calculation formula is: in This represents the highest fitness value in the current population. This represents the current average fitness value of the population. The larger fitness value among the two individuals at the crossover. , This is a preset constant.

7. The cloud-edge collaborative forging equipment cluster state perception and linkage control method as described in claim 6, characterized in that, The process of generating the collaborative control strategy includes: The Pareto optimal solution set obtained by multi-objective optimization is mapped as the antecedent of fuzzy rules, and the adjustment amount of equipment control parameters is used as the consequent to construct a Takagi-Sugeno type fuzzy inference system. For each optimal solution, the corresponding fuzzy rule form is: if the device state is And the production efficiency is Then the control parameter adjustment amount is ; The final policy output is obtained by a weighted average of all activated rules: in To control the adjustment amount of the parameters, The number of rules activated. For the first The activation strength of the rule, This refers to the adjustment amount of the control parameters output by this rule.

8. The cloud-edge collaborative forging equipment cluster state perception and linkage control method as described in claim 1, characterized in that, The coordinated adjustment includes: After receiving the collaborative control strategy, the edge computing node parses the strategy instructions and obtains the target control parameters of its own device and the adjustment amount of the associated parameters of adjacent devices. Based on the process correlation model between equipment, calculate the adjustment sequence of the control parameters of this equipment, and simultaneously send adjustment instructions to adjacent equipment.

9. The cloud-edge collaborative forging equipment cluster state perception and linkage control method as described in claim 8, characterized in that, The process association model is represented by a directed graph, where nodes represent equipment and edges represent process flows. For each device node The final adjustment amount of its control parameters for: in It is sent from the cloud to the device. The original strategy adjustment amount, It is equipment The set of precursor equipment in the process correlation diagram and respectively equipment and its precursor equipment Health status index It is equipment For equipment The influence weight of the process, This is a correction factor based on process stability.

10. A cloud-edge collaborative forging equipment cluster status perception and linkage control system, characterized in that, The system is used to implement the cloud-edge collaborative forging equipment cluster status perception and linkage control method as described in claims 1 to 9, including: The data acquisition module is used to collect real-time operating status data through edge computing nodes deployed on each forging and pressing equipment, and to perform data cleaning and feature extraction at the edge side to generate equipment status assessment results. The data upload module is used to upload the device status assessment results and real-time control parameters of the devices generated by each edge computing node to the cloud control center. The cloud-based analytics module is used to perform global analysis based on the received status assessment results of all cluster devices and real-time control parameters of the devices, integrating historical operating data and production plans, and using collaborative control algorithms to generate collaborative control strategies. The linkage control module is used to send the collaborative control strategy to the corresponding edge computing nodes, and control the edge computing nodes to adjust the control parameters of the device and adjacent devices with process association with it in accordance with the received strategy instructions. The feedback module is used to send the execution effect of control commands back to the cloud, forming a closed-loop control circuit.

Images