Method and system for collecting and monitoring control of plc data of automation industry equipment

Abstract

The application relates to the field of automatic industrial control technology and discloses a kind of automatic industrial equipment plc data acquisition monitoring control method and system.The method comprises: receiving multi-source PLC data stream, generating equipment health degree evaluation matrix by processing through multimodal data fusion model;Accordingly, a dynamic control strategy generation model is constructed, and a multi-objective optimization instruction set is output;The interactive influence of equipment group is simulated by using a process chain collaborative constraint model, and the real-time control parameters of the optimization instruction set are optimized;Through a distributed edge computing framework, the parameters are adjusted in parallel, and the equipment collaborative control signal is output to the industrial monitoring center.The system includes a distributed data acquisition module, a multimodal fusion calculation module, a dynamic strategy generation module, an edge collaborative control module and an industrial communication gateway.The application can efficiently process multi-source data, accurately evaluate equipment health, optimize control strategy and realize equipment collaboration, thereby improving industrial production efficiency and energy utilization rate.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automation industrial control technology, in particular to a PLC data acquisition monitoring and control method and system for automation industrial equipment. BACKGROUND

[0002] In modern industrial production, the degree of automation is constantly improving, and the operation status of industrial equipment directly affects production efficiency, product quality and economic benefits of enterprises. PLC, as the core control equipment of industrial automation, is widely used in various industrial scenes, and the data collected by PLC contains key information of equipment operation. However, the current PLC data processing and monitoring and control technology of industrial equipment faces many challenges.

[0003] From the perspective of data acquisition, the multi-source PLC data generated by industrial equipment is complex and diverse, including device operation status signals, energy consumption time series data and environmental sensor feedback information. The sampling frequency, precision and format of different types of data differ greatly, making data fusion difficult. Traditional data acquisition methods cannot comprehensively, accurately and timely acquire these multi-source heterogeneous data, resulting in a large amount of valuable information being missed or unable to be effectively integrated and utilized, affecting accurate judgment of the overall operation status of the equipment.

[0004] In terms of data processing and analysis, the existing technology lacks effective multi-modal data fusion means. The noise interference in multi-source PLC data is serious, which reduces the data quality and affects the accuracy of subsequent analysis. Moreover, the feature extraction between data is not accurate enough, which cannot deeply mine the potential rules and health status related features of equipment operation. Therefore, it is difficult to build an accurate equipment health degree evaluation system, which cannot timely discover early fault hidden dangers of equipment, leading to equipment sudden failure, production interruption, increased maintenance cost and other problems.

[0005] In terms of control strategy formulation, the previous control methods are mostly based on fixed rules or experience, lacking dynamic adjustment ability. With the increasing complexity of industrial production processes, the coupling relationship between devices is close, and the running status change of a single device will trigger a chain reaction. Traditional control strategies cannot generate effective multi-objective optimization instruction set according to real-time health status of equipment and process chain coordination requirements, making it difficult to realize precise control of start-stop scheduling and parameter adjustment of equipment, reducing production efficiency and resource utilization.

[0006] In addition, there are also deficiencies in the simulation and optimization of process chain coordination constraints. The interaction between industrial equipment groups is complex, and the existing technology cannot fully consider the energy efficiency transmission, delay response and real-time working condition disturbance between devices. In actual production, there are often cases of poor coordination between devices, serious energy waste, which cannot guarantee the stable and efficient operation of the entire process chain.

[0007] Meanwhile, the application of edge computing in industrial control is not mature enough. The distributed edge computing framework fails to fully exert its advantages, and when multiple nodes are cooperatively optimized, the target is single, and the energy consumption and production efficiency cannot be considered simultaneously. Moreover, the cooperation efficiency between nodes is low, the weight distribution is unreasonable, and it is difficult to quickly converge to the optimal solution, so it is difficult to output the device cooperative control signal that meets the actual needs of industrial production. SUMMARY

[0008] The purpose of the present application is to provide an automatic industrial equipment PLC data acquisition monitoring and control method and system to solve the problems raised in the background art.

[0009] To achieve the above purpose, the present application provides the following technical solution: an automatic industrial equipment PLC data acquisition monitoring and control method and system, the method comprising:

[0010] Receiving the multi-source PLC data stream of the industrial equipment, the data stream comprising the device running state signal, the energy consumption time series data and the environmental sensor feedback information;

[0011] Based on the preset multi-modal data fusion model, the multi-source PLC data is denoised, aligned and feature extracted to generate a device health degree evaluation matrix;

[0012] According to the health degree evaluation matrix, a dynamic control strategy generation model is constructed, and a multi-objective optimization instruction set is output, the instruction set comprising a device start-stop scheduling sequence and a parameter adjustment gradient scheme;

[0013] Based on the preset process chain cooperative constraint model, the interaction between the device groups in the future preset period is simulated, and the real-time control parameters in the instruction set are optimized;

[0014] The real-time control parameters are adjusted in parallel through a distributed edge computing framework, and a device cooperative control signal is output to an industrial monitoring center.

[0015] Preferably, the construction step of the multi-modal data fusion model comprises:

[0016] Collecting historical device running logs and fault case libraries, and constructing a multi-dimensional training set containing signal fluctuation patterns, device degradation characteristics and environmental interference factors;

[0017] The multi-dimensional training set is subjected to noise suppression and feature enhancement by a generative adversarial network, and the steady-state running features and transient abnormal components are separated;

[0018] Combined with the device physical constraint equation, a state space constraint condition of the time sequence dependence relationship between features is constructed;

[0019] The constraint condition is embedded into a time sequence convolutional neural network to generate the multi-modal data fusion model supporting online updating.

[0020] Preferably, the dynamic control strategy generation model comprises:

[0021] According to the entropy value change of the health degree evaluation matrix, the device performance degradation level is dynamically divided;

[0022] Based on the device topology connection relationship and the process coupling strength, the system cascade failure risk score is calculated;

[0023] The risk score and the degradation level are dynamically weighted and fused by a segmented linear function to generate a device-specific control threshold;

[0024] According to the threshold, the execution condition of the multi-level process adjustment protocol is triggered.

[0025] Preferably, the construction step of the process chain synergy constraint model comprises:

[0026] Collect historical interaction data between device groups to construct a process causal graph data set;

[0027] Infer the energy efficiency transmission coefficient and the delay response parameter between devices through a Bayesian network;

[0028] Combine digital twin technology to construct a process chain virtual simulation environment to quantify the energy flow balance constraint between nodes;

[0029] Input the constraint and real-time working condition disturbance factor into a spatio-temporal graph neural network to generate the process chain synergy constraint model.

[0030] Preferably, the method further comprises:

[0031] According to the simulation result of the process chain synergy constraint model, a key energy flow bottleneck node is identified;

[0032] In the instruction set, a dynamic energy efficiency compensation strategy is configured for the node;

[0033] Based on the compensation strategy, a cross-device collaborative optimization scheme is automatically generated, including load balancing allocation instructions and power factor correction parameters.

[0034] Preferably, the calculation of the system cascade failure risk score comprises:

[0035] Obtain a real-time device communication delay matrix and a spare parts inventory distribution map to construct a system resilience evaluation tensor;

[0036] Extract implicit correlation factors between multi-dimensional features through a tensor decomposition algorithm;

[0037] Perform Hadamard product operation on the correlation factors and the evaluation tensor to obtain a comprehensive risk score.

[0038] Preferably, the embedding of the state space constraints comprises:

[0039] Perform Lie group symmetry analysis on the feature enhancement results to screen time sequence evolution patterns consistent with physical laws;

[0040] Generate state transition trajectories satisfying device dynamics constraints through Markov chain Monte Carlo sampling;

[0041] Sparse training of the convolution kernel of the time sequence convolution network using trajectory data to ensure that the model output conforms to the process chain conservation law.

[0042] Preferably, the execution of the distributed edge computing framework comprises:

[0043] Define a multi-node collaborative optimization loss function containing dual objectives of minimum energy consumption and maximum production efficiency;

[0044] Train the local control strategy of each edge node through an asynchronous parallel gradient descent algorithm;

[0045] In each iteration, dynamically adjust the weight distribution ratio between nodes according to the target deviation;

[0046] Output the device collaborative control signal satisfying the Pareto optimality condition;

[0047] Wherein, the loss function is:

[0048]

[0049] In the formula, L represents the loss value, E represents the total energy consumption score of the system, P represents the production efficiency score, is a dynamic adjustment factor.

[0050] Preferably, the method further comprises:

[0051] After configuring the dynamic energy efficiency compensation strategy, real-time monitoring of the energy efficiency fluctuation entropy value of the energy flow bottleneck node;

[0052] If the entropy value exceeds the preset safety boundary, trigger the chaotic optimization algorithm to re-plan the load distribution topology of the device group.

[0053] Preferably, the present application further comprises an automatic industrial equipment PLC data acquisition monitoring and control system, the system comprises:

[0054] A distributed data acquisition module is deployed at the edge node of the industrial field to acquire real-time multi-source PLC data streams;

[0055] A multi-modal fusion computing module is connected to the data acquisition module and internally has a time series convolutional neural network and a generative adversarial network to perform data denoising and feature extraction.

[0056] A dynamic strategy generation module integrates a process chain coordination constraint model and a digital twin simulation environment to generate a multi-objective optimization instruction set.

[0057] An edge coordination control module adjusts device control parameters through an asynchronous parallel computing framework and outputs coordination control signals.

[0058] An industrial communication gateway transmits the control signals to a PLC execution terminal and feeds back real-time working conditions to a monitoring center.

[0059] Compared with the prior art, the present application has the following advantages:

[0060] In terms of data processing, the multi-modal data fusion model is used to denoise and align the multi-source PLC data of the industrial equipment and extract features. The historical equipment operation logs and the fault case library are used to construct a multi-dimensional training set. The generative adversarial network is used to suppress noise and enhance features. The state space constraint conditions are constructed by combining the device physical constraint equation and embedding the time series convolutional neural network. This not only effectively removes data noise and ensures data accuracy, but also accurately extracts key features related to equipment health, and the generated equipment health degree evaluation matrix can accurately reflect the actual health status of the equipment. Compared with traditional data processing methods, the data quality and analysis reliability are greatly improved, which helps to detect potential equipment failures in advance, reduce equipment failure rates, reduce production downtime caused by equipment failures, and thus improve production continuity and stability.

[0061] For control strategy generation, a dynamic control strategy generation model is constructed according to the health degree evaluation matrix. The model divides the device performance degradation level according to the entropy value change, calculates the system cascade failure risk score by combining the device topology connection relationship and process coupling strength, generates the control threshold value by dynamic weighted fusion through the piecewise linear function, and triggers the multi-level process adjustment protocol. This dynamic and intelligent control strategy can adjust the equipment operation in real time according to the real-time state of the equipment and the system risk condition, avoid excessive degradation of the equipment, and reduce the system failure risk. Compared with the traditional control method based on fixed rules, it is more flexible and accurate, which can effectively improve the service life and operation safety of the equipment, and reduce maintenance costs and downtime losses.

[0062] In the process chain collaborative optimization, the process chain collaborative constraint model plays an important role. The historical interaction data between the equipment groups are collected to construct a process causal graph dataset, the Bayesian network is used to infer the parameters, the digital twin technology is used to construct a virtual simulation environment, the energy flow balance constraint is quantified, and the space-time graph neural network generation model is input. The model simulates the interaction between equipment groups, optimizes real-time control parameters, identifies key energy flow bottleneck nodes and configures dynamic energy efficiency compensation strategies, and generates a cross-device collaborative optimization scheme. This series of operations greatly improves the collaboration between devices, makes energy utilization more reasonable, reduces energy waste, improves energy utilization rate, and reduces production cost. At the same time, it ensures the stable operation of the process chain, improves product quality and production efficiency.

[0063] In the distributed edge computing framework execution, a loss function containing the dual objectives of energy consumption minimization and production efficiency maximization is defined, the local control strategy of each edge node is trained by an asynchronous parallel gradient descent algorithm, the weight distribution ratio between nodes is dynamically adjusted according to the target deviation, and the device collaborative control signal meeting the Pareto optimality condition is output. This realizes efficient collaborative optimization of multiple nodes, fully utilizes edge node computing resources, and takes into account the two key indicators of energy consumption and production efficiency, so that industrial equipment can maintain high production efficiency while saving energy, and the comprehensive competitiveness of industrial production is enhanced.

[0064] Overall, the method and system of the present application comprehensively improve the intelligent management level of industrial equipment, from data acquisition, processing and analysis to control strategy generation and execution, realize full-process optimization, bring significant economic and social benefits to industrial enterprises, and promote the development of industrial automation to a higher level. BRIEF DESCRIPTION OF DRAWINGS

[0065] Figure 1 The working principle diagram of the automatic industrial equipment PLC data acquisition monitoring and control method described in the present application;

[0066] Figure 2 The flowchart for equipment performance degradation evaluation and process adjustment;

[0067] Figure 3 The flowchart for system resilience risk assessment and early warning;

[0068] Figure 4 The flowchart for energy efficiency fluctuation entropy monitoring and load distribution optimization. DETAILED DESCRIPTION

[0069] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0070] Please refer to Figures 1-4 The present application provides a kind of automation industrial equipment PLC data acquisition monitoring control method, and its specific implementation steps are as follows:

[0071] Receive multi-source PLC data stream: in industrial field, through the sensor and data acquisition equipment deployed in each key position, the multi-source PLC data stream of industrial equipment is received in real time.These data streams cover equipment running state signals, such as whether the equipment is running normally, whether there is abnormal vibration, etc.;Energy consumption time series data, i.e. energy consumption of equipment at different time;Environmental sensor feedback information, including environmental temperature, humidity, dust concentration and other environmental factors that may affect equipment operation data.

[0072] Multi-source PLC data processing and generating equipment health degree evaluation matrix: based on the preset multi-modal data fusion model, the above multi-source data is processed. The model first denoises and aligns the data, removes noise interference in the data, and ensures the accuracy and consistency of the data. Then feature extraction is performed to mine key features related to equipment health condition from complex data. Through these operations, an equipment health degree evaluation matrix that can accurately reflect the health degree of the equipment is generated.

[0073] Constructing a dynamic control strategy generation model and outputting instruction set: according to the generated health degree evaluation matrix, a dynamic control strategy generation model is constructed. The model will consider multiple factors and output a multi-objective optimization instruction set, which includes device start-stop scheduling sequence, i.e. according to the health condition of the device, production demand and other factors, reasonably arranging the start and stop time of the device;And parameter adjustment gradient scheme, which clearly defines the adjustment direction and amplitude of the device operating parameters.

[0074] Simulate interactive influence and optimize real-time control parameters: based on the preset process chain coordination constraint model, the interactive influence between device groups in the future preset period is simulated. For example, the running state change of a device may have a chain reaction on other devices related to it, and this influence can be predicted through the model. Then, according to the simulation results, the real-time control parameters in the instruction set are optimized, so that the control strategy is more in line with the actual production demand.

[0075] Adjusting parameters and outputting collaborative control signals: The optimized real-time control parameters are adjusted in parallel through a distributed edge computing framework. The distributed edge computing framework can make full use of the computing resources of the edge nodes in the industrial field to achieve efficient parallel computing. Finally, the adjusted parameters are converted into device collaborative control signals and output to the industrial monitoring center, where the industrial devices are monitored and controlled uniformly.

[0076] The implementation of the present application will be further described below in connection with Examples 1 to 6.

[0077] Example 1

[0078] In this embodiment, the construction process of the multi-modal data fusion model is described.

[0079] A large number of historical device operation logs and fault case libraries are collected. The historical device operation logs contain various data of the device during long-time operation, such as changes in operating parameters, performance under different working conditions, etc.; the fault case library records the specific circumstances when the device fails, including fault phenomena, occurrence time, maintenance measures, etc. Through the organization and analysis of these data, a multi-dimensional training set containing signal fluctuation patterns, device degradation characteristics and environmental interference factors is constructed. For example, the signal fluctuation pattern can reflect the change law of some key signals during the operation of the device, the device degradation characteristic can reflect the performance decline trend of the device as the use time increases, and the environmental interference factor records the influence of different environmental factors on the operation of the device.

[0080] The multi-dimensional training set is processed using a generative adversarial network. The generative adversarial network consists of a generator and a discriminator, the generator is responsible for generating false data, and the discriminator is used to judge whether the data is real or generated. In this process, the generative adversarial network suppresses noise in the multi-dimensional training set, removes noise interference in the data, and makes the data more pure; at the same time, it enhances the features, highlighting those features that are important to the health assessment of the device. In this way, the steady-state operating features and transient abnormal components are separated, and the features of the device during normal operation and the features that may appear when an anomaly occurs are distinguished.

[0081] Combined with the device physical constraint equation, the state space constraint condition of the time sequence dependence relationship between features is constructed. The device physical constraint equation is a mathematical model established based on the physical principles and operating mechanisms of the device, which describes the internal relationship between the parameters of the device. For example, for a motor device, there is a certain physical relationship between its speed, torque and power, which can be expressed through the physical constraint equation. According to these equations, the time sequence dependence relationship between features is determined, and the state space constraint condition is constructed.

[0082] Embed the above constraints into the time series convolutional neural network. The time series convolutional neural network is a neural network specially designed to process time series data, which can effectively extract the time series features in the data. By embedding the constraints, a multi-modal data fusion model supporting online updating is trained. During the model training process, a large amount of training data is used to optimize the network, so that the model can accurately denoise and align the multi-source PLC data and extract features, providing reliable data support for subsequent equipment health assessment.

[0083] Embodiment 2:

[0084] This embodiment mainly focuses on the dynamic control strategy generation model.

[0085] When constructing the dynamic control strategy generation model according to the health assessment matrix, first, according to the entropy value change of the health assessment matrix, the equipment performance degradation level is dynamically divided. Entropy is an index to measure the uncertainty of data. In the health assessment matrix, the change of entropy value can reflect the stability of equipment performance. When the entropy value is low, it means that the equipment running state is stable and the performance degradation degree is low; when the entropy value increases, it means that the equipment may have some abnormal situations and the performance degradation degree is aggravated. By setting different entropy threshold values, the equipment performance degradation level is divided into different levels, such as mild degradation, moderate degradation and severe degradation.

[0086] Based on the device topology connection relationship and process coupling strength, the system cascade failure risk score is calculated. Obtain the real-time device communication delay matrix and spare parts inventory distribution diagram, and construct the system resilience evaluation tensor. The device topology connection relationship describes the physical connection and logical relationship between industrial devices, and the process coupling strength reflects the correlation degree of different devices in the production process. The communication delay matrix records the delay of data transmission between devices, and the spare parts inventory distribution diagram shows the storage location and quantity of each device spare parts. By integrating these information, the system resilience evaluation tensor is constructed, which can comprehensively reflect the anti-interference ability and recovery ability of the system in the face of various disturbances.

[0087] Extract the implicit correlation factor between multi-dimensional features through tensor decomposition algorithm. Tensor decomposition algorithm is a technology that can decompose high-dimensional tensor into the product of multiple low-dimensional tensors. Through this algorithm, the hidden correlation between multi-dimensional features in the system resilience evaluation tensor can be extracted, and the implicit correlation factor can be obtained. Perform Hadamard product operation on the correlation factor and the evaluation tensor to obtain the comprehensive risk score. Hadamard product operation is an operation of multiplying corresponding elements. Through this operation, the implicit correlation factor can be combined with each element in the evaluation tensor, and a comprehensive score reflecting the system cascade failure risk can be obtained.

[0088] The risk score and the degradation level are dynamically weighted and fused by a piecewise linear function to generate a device-specific control threshold. The piecewise linear function is a function with different linear expressions in different intervals. According to different situations of the risk score and the degradation level, the parameters of the piecewise linear function are reasonably selected to dynamically weight and fuse the two. For example, when the device performance degradation level is high and the risk score is also high, the weight of the risk score in the fusion is appropriately increased, so that the generated control threshold is more stringent. According to the generated threshold, the execution condition of the multi-level process adjustment protocol is triggered. When the operating parameters of the device exceed the control threshold, the corresponding process adjustment protocol is triggered immediately, such as adjusting the operating speed of the device, changing the working mode of the device, etc., to ensure that the device can continue to operate safely and stably.

[0089] Embodiment 3

[0090] This embodiment details the construction steps of the process chain coordination constraint model.

[0091] Historical interaction data between device groups is collected to construct a process causal graph dataset. In industrial production processes, there are various interactions between devices, which have an impact on the production process. By collecting and organizing historical interaction data between device groups for a long time, including information such as signal transmission, energy transmission, and material flow between devices, a process causal graph dataset is constructed. This dataset can clearly show the causal relationships between devices and provide a basis for subsequent analysis.

[0092] The energy efficiency transmission coefficient and the delay response parameter between devices are inferred by Bayesian network. Bayesian network is a graphical model based on probabilistic reasoning, which can infer unknown parameters using known data and prior knowledge. In this embodiment, Bayesian network is used to analyze the process causal graph dataset to infer the energy efficiency transmission coefficient and the delay response parameter between devices. The energy efficiency transmission coefficient reflects the influence of the energy consumption of one device on other devices, and the delay response parameter represents the response delay time between devices.

[0093] A process chain virtual simulation environment is constructed in combination with digital twin technology. Digital twin technology is a technology that maps and simulates physical entities through digital means. In the process chain virtual simulation environment, a virtual model identical to the actual industrial devices and process chain is created, including the structure, performance parameters, and operating state of the devices. By simulating the operation of the virtual model, the running conditions of each node in the process chain can be monitored and analyzed in real time, and the energy flow balance constraints between nodes can be quantified. For example, in the virtual simulation environment, the operating conditions of different devices can be simulated to calculate the energy flow between nodes and determine the energy flow balance constraints.

[0094] The constraints are input into the spatio-temporal graph neural network with real-time operating condition disturbance factors to generate a process chain collaborative constraint model. The spatio-temporal graph neural network is a neural network that can simultaneously process spatial and temporal information, which can effectively capture the spatial relationship between devices and time series information. The energy flow balance constraints and real-time operating condition disturbance factors are input into the spatio-temporal graph neural network, and through the learning and training of the network, a process chain collaborative constraint model that can accurately reflect the interaction between device groups is generated. This model can provide an important basis for subsequent real-time control parameter optimization.

[0095] Embodiment 4:

[0096] In this embodiment, further operations are performed based on the simulation results of the process chain collaborative constraint model.

[0097] According to the simulation results of the process chain collaborative constraint model, key energy flow bottleneck nodes are identified. In industrial production processes, some devices or nodes may become bottlenecks that limit the energy flow of the entire process chain due to insufficient energy supply, low transmission efficiency, etc. By analyzing the simulation results of the process chain collaborative constraint model, key energy flow bottleneck nodes that have a greater impact on energy flow are found. For example, by simulating the energy flow in the process chain under different operating conditions, observe which nodes have energy accumulation or insufficient supply, and determine these nodes as key energy flow bottleneck nodes.

[0098] In the instruction set, dynamic energy efficiency compensation strategies are configured for these nodes. Dynamic energy efficiency compensation strategies are a strategy that automatically adjusts energy distribution based on the real-time running conditions and energy demand of the nodes. For example, for key energy flow bottleneck nodes with insufficient energy supply, energy supply can be increased; for nodes with low energy transmission efficiency, measures such as optimizing the transmission path and improving the performance of the transmission device can be taken to improve energy transmission efficiency. By configuring dynamic energy efficiency compensation strategies, the energy utilization efficiency of the entire process chain is improved.

[0099] Based on the compensation strategy, a cross-device collaborative optimization scheme is automatically generated, including load balancing distribution instructions and power factor correction parameters. Load balancing distribution instructions are used to reasonably distribute the workloads of each device to avoid situations where some devices have excessive loads while others have light loads. Power factor correction parameters are used to improve the power factor of the device and reduce energy waste. For example, based on the actual operating power and power factor of the device, the power factor correction parameters that need to be adjusted are calculated, and the power factor is improved by adjusting the circuit parameters of the device.

[0100] After configuring the dynamic energy efficiency compensation strategy, the energy efficiency fluctuation entropy value of the energy flow bottleneck node is monitored in real time. The energy efficiency fluctuation entropy value can reflect the stability of the energy utilization efficiency of the energy flow bottleneck node. If the entropy value exceeds the preset safety boundary, the chaotic optimization algorithm is triggered to re-plan the load distribution topology of the device group. The chaotic optimization algorithm is an optimization algorithm based on chaos theory, which has the advantages of strong global search ability and fast convergence speed. When the energy efficiency fluctuation entropy value exceeds the preset safety boundary, it means that the energy utilization efficiency of the energy flow bottleneck node has a large fluctuation, which may affect the stable operation of the entire process chain. At this time, the chaotic optimization algorithm is triggered to re-plan the load distribution topology of the device group, so that the load distribution between devices is more reasonable, and the energy utilization efficiency and stability of the entire system are improved.

[0101] Embodiment 5:

[0102] This embodiment details the embedding process of the state space constraint condition. After constructing the state space constraint condition in the multi-modal data fusion model, it needs to be embedded into the time series convolutional neural network.

[0103] Perform Lie group symmetry analysis on the feature enhancement results to screen time series evolution patterns that conform to physical laws. Lie group is a mathematical structure with continuous symmetry, which has been widely used in physics and engineering. By performing Lie group symmetry analysis on the feature enhancement results, the symmetry relationship existing in the data can be found out, so as to screen out time series evolution patterns that conform to physical laws. For example, during the operation of the equipment, the change of some physical quantities may have certain symmetry, and through Lie group symmetry analysis, these rules can be found out, and abnormal patterns that do not conform to the physical laws can be excluded.

[0104] Generate state transition trajectories that meet the equipment dynamics constraints through Markov chain Monte Carlo sampling. Markov chain Monte Carlo sampling is a random sampling method based on Markov chain, which can efficiently sample in high-dimensional space. In this embodiment, Markov chain Monte Carlo sampling method is used to generate a series of state transition trajectories that meet the requirements according to the equipment dynamics constraints. The equipment dynamics constraints are established based on the physical principles and motion equations of the equipment, which describe the restriction conditions of the equipment transition between different states. By generating these state transition trajectories, more sample data can be provided for subsequent model training.

[0105] The trajectory data is used to train the convolution kernel of the time series convolution network sparsely, ensuring that the model output conforms to the process chain conservation law. The convolution kernel of the time series convolution network is a key component in the network for extracting time series features. By sparsely training the convolution kernel, the number of model parameters can be reduced, improving the computational efficiency and generalization ability of the model. During training, the generated state transition trajectory data is used to optimize the convolution kernel, so that the model output meets the process chain conservation law. For example, in terms of energy consumption, the model output should satisfy the energy conservation law, that is, the total amount of input energy should be equal to the total amount of output energy plus the amount of energy consumed by the device. In this way, it is ensured that the multi-modal data fusion model can accurately reflect the actual operation of the device, providing reliable support for device health assessment and control strategy generation.

[0106] Embodiment 6:

[0107] When adjusting real-time control parameters in parallel using a distributed edge computing framework, first define the loss function for multi-node collaborative optimization, which includes both energy consumption minimization and production efficiency maximization. The loss function is:

[0108]

[0109] wherein, represents the loss value, which is an indicator to measure the effect of the entire optimization process; represents the total energy consumption score, which quantifies the energy consumption of the system during operation; represents the production efficiency score, reflecting the production efficiency of the system; is a dynamic adjustment factor that can dynamically adjust the weight of energy consumption and production efficiency in the loss function according to actual production needs. For example, when energy costs are high, the value of can be appropriately increased to make the optimization process focus more on energy consumption minimization; when production tasks are urgent, the value of can be reduced to focus more on production efficiency maximization.

[0110] Train the local control strategy of each edge node using the asynchronous parallel gradient descent algorithm. The asynchronous parallel gradient descent algorithm is a commonly used optimization algorithm in distributed computing environments, which allows each edge node to independently calculate and update, improving computational efficiency. On each edge node, according to the locally collected data and the defined loss function, the gradient is calculated and the local control strategy is updated. For example, for an edge node responsible for controlling a certain device, it will calculate the gradient under the current control strategy according to the real-time running data of the device and the loss function, and then adjust the parameters of the control strategy according to the gradient information.

[0111] ​In each iteration, the weight distribution ratio between nodes is dynamically adjusted according to the target deviation. The target deviation can be measured by calculating the difference between the current loss value and the ideal loss value. When the control strategy of a certain edge node contributes more to reducing the loss value, the proportion of this node in the weight distribution is appropriately increased; otherwise, the proportion is decreased. In this way of dynamically adjusting the weight distribution ratio, the distributed edge computing framework can converge to the optimal solution more efficiently.

[0112] The device cooperative control signals that meet the Pareto optimality condition are output. The Pareto optimality refers to the state that a certain objective cannot be further optimized without sacrificing other objectives. In this embodiment, by continuously adjusting the control strategies of each edge node and the weight distribution ratio, the system reaches the Pareto optimal state, and the device cooperative control signals that meet the condition are output. These signals are transmitted to the industrial monitoring center for precise control of industrial equipment, achieving the dual goals of minimizing energy consumption and maximizing production efficiency.

[0113] It should be noted that the relational terms herein, such as first and second, are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between or among the entities or operations. Moreover, the terms "comprises", "comprising", or any other variations thereof are intended to cover non-exclusive inclusions, so that a process, method, article, or apparatus that comprises a list of elements does not only include those elements, but also includes other elements not expressly listed, or other elements inherent in such a process, method, article, or apparatus.

[0114] Although the embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An automated industrial equipment PLC data acquisition monitoring control method, characterized in that, Comprise: Receive multi-source PLC data stream of industrial equipment, the data stream includes equipment running state signal, energy consumption time series data and environmental sensor feedback information; Based on the preset multi-modal data fusion model, the multi-source PLC data is denoised, aligned and feature extracted to generate a device health assessment matrix; According to the health assessment matrix, a dynamic control strategy generation model is constructed, and a multi-objective optimization instruction set is output, the instruction set includes device start-stop scheduling sequence and parameter adjustment gradient scheme; Based on the preset process chain synergy constraint model, the interaction between the equipment groups in the future preset period is simulated, and the real-time control parameters in the instruction set are optimized; Through the distributed edge computing framework, the real-time control parameters are adjusted in parallel, and the equipment coordination control signal is output to the industrial monitoring center; The dynamic control strategy generation model comprises: According to the entropy value change of the health assessment matrix, the equipment performance degradation level is dynamically divided; Based on the device topology connection relationship and the process coupling strength, the system cascade failure risk score is calculated; The risk score and the degradation level are dynamically weighted and fused through a segmented linear function to generate a device-specific control threshold; According to the threshold, the execution condition of the multi-level process adjustment protocol is triggered; The calculation of the system cascade failure risk score comprises: Obtain the real-time device communication delay matrix and the spare parts inventory distribution map to construct a system resilience evaluation tensor; Through the tensor decomposition algorithm, the implicit correlation factor between multi-dimensional features is extracted; The correlation factor and the evaluation tensor are subjected to Hadamard product operation to obtain a comprehensive risk score.

2. The automated industrial equipment PLC data collection monitoring control method of claim 1, wherein, The construction steps of the multi-modal data fusion model comprise: Collect historical equipment operation logs and fault case library to construct a multi-dimensional training set containing signal fluctuation mode, equipment degradation characteristics and environmental disturbance factors; Through the adversarial generative network, the multi-dimensional training set is subjected to noise suppression and feature enhancement to separate the steady-state operation features and transient abnormal components; Combine the device physical constraint equation to construct the state space constraint condition of the time sequence dependence between features; Embed the constraint condition into the time sequence convolutional neural network to generate the multi-modal data fusion model supporting online update.

3. The automated industrial equipment PLC data collection monitoring control method of claim 1, wherein, The construction steps of the process chain synergy constraint model comprise: Collect historical interaction data between equipment groups to construct a process causal graph data set; Infer the energy efficiency transmission coefficient and delay response parameter between devices through Bayesian network; Combine digital twin technology to construct a process chain virtual simulation environment to quantify the energy flow balance constraint between nodes; Input the constraint and real-time working condition disturbance factor into the spatio-temporal graph neural network to generate the process chain synergy constraint model.

4. The automated industrial equipment PLC data collection monitoring control method of claim 3, wherein, Further comprise: According to the simulation results of the process chain synergy constraint model, identify the key energy flow bottleneck node; In the instruction set, configure a dynamic energy efficiency compensation strategy for the node; Based on the compensation strategy, automatically generate a cross-equipment collaborative optimization scheme, including load balancing allocation instruction and power factor correction parameter.

5. The automated industrial equipment PLC data collection monitoring control method of claim 2, wherein, The embedding of the state space constraint condition comprises: Perform Lie group symmetry analysis on the feature enhancement results to screen time sequence evolution modes consistent with physical laws; The state transition trajectory satisfying the equipment dynamics constraint is generated by Markov chain Monte Carlo sampling; The convolution kernel of the time series convolution network is sparsely trained using trajectory data to ensure that the model output conforms to the process chain conservation law.

6. The automated industrial equipment PLC data collection monitoring control method of claim 1, wherein, The execution of the distributed edge computing framework comprises: Defining a multi-node collaborative optimization loss function containing dual objectives of energy consumption minimization and production efficiency maximization; Training the local control strategy of each edge node through an asynchronous parallel gradient descent algorithm; In each iteration, the weight distribution ratio between nodes is dynamically adjusted according to the target deviation degree; Output the device collaborative control signal that meets the Pareto optimality condition; The loss function is: In the formula, represents a loss value, represents a total energy consumption score of the system, represents a production efficiency score, is a dynamic adjustment factor.

7. The automated industrial equipment PLC data collection monitoring control method of claim 4, wherein, Also includes: After configuring the dynamic energy efficiency compensation strategy, real-time monitoring of energy flow bottleneck node energy efficiency fluctuation entropy value; If the entropy value exceeds the preset safety boundary, trigger the chaos optimization algorithm to re-plan the load distribution topology of the equipment group.

8. An automation industrial equipment PLC data acquisition monitoring control system for implementing the automation industrial equipment PLC data acquisition monitoring control method according to any one of claims 1-7, characterized in that, Includes: Distributed data acquisition module, deployed in the industrial field edge node, used for real-time acquisition of multi-source PLC data stream; Multi-modal fusion computing module, connected to the data acquisition module, with built-in time series convolutional neural network and generative adversarial network, performing data denoising and feature extraction; Dynamic strategy generation module, integrated with process chain collaborative constraint model and digital twin simulation environment, generating multi-objective optimization instruction set; Edge collaborative control module, adjusting equipment control parameters through an asynchronous parallel computing framework, outputting collaborative control signals; Industrial communication gateway, transmitting the control signals to the PLC execution terminal and feeding back the real-time working conditions to the monitoring center.

Images