Automatic control system of industrial equipment

Abstract

The invention discloses an automatic control system of industrial equipment, and relates to the technical field of control systems. The calculation layer adopts an edge calculation technology and is used for performing real-time calculation on the operation data, acquired by the data acquisition layer, of the industrial equipment; the control layer is used for further processing and analyzing the calculation result of the calculation layer, identifying the operation state of the industrial equipment through a data twinning technology, and generating a control instruction according to the operation state of the industrial equipment; an execution layer; a human-computer interface layer; a communication network; a unique data transmission mechanism is adopted between the edge computing node and the control layer, and different transmission modes and data storage modes are implemented according to network conditions, so that network bandwidth is effectively utilized, long-time accumulation of data at a local node is avoided, and meanwhile, when network delay is relatively high, the data transmission efficiency is improved. The condition of data transmission loss or error is effectively prevented, and the validity and timeliness of data transmission are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of control systems, and particularly to an automated control system for industrial equipment. Background Art

[0002] In the industrial field, with the rapid development of automation technology, the automated control system of industrial equipment plays a crucial role in improving production efficiency and reducing labor costs. However, the existing automated control systems of industrial equipment still face many challenges in practical applications, which limit the further improvement of their performance and wide application.

[0003] In terms of data transmission, in traditional automated control systems, the data transmission mechanism between edge computing nodes and the control layer is relatively single and often cannot be flexibly adjusted according to the real-time network conditions. When the network condition is good, the network bandwidth resources are not fully utilized, resulting in low data transmission efficiency; while when the network condition is poor, such as high network latency, data is likely to accumulate at local nodes for a long time, and even data transmission loss or error may occur, seriously affecting the effectiveness and timeliness of data transmission, and then affecting the response speed and stability of the entire control system.

[0004] The generation of control instructions is also a weak link in the existing system. Currently, fuzzy control and neural network optimization are two commonly used control strategies. Fuzzy control has a certain degree of flexibility and can handle uncertain and fuzzy information, but there are certain deficiencies in accuracy; neural network optimization has strong accuracy, but in dealing with complex and changing actual situations, its flexibility is relatively lacking. Most of the existing control systems separately adopt fuzzy control or neural network optimization to generate control instructions, and fail to give full play to the advantages of both, resulting in it being difficult to achieve an ideal balance between the accuracy and flexibility of control instructions, and restricting the improvement of the overall performance of the control system.

[0005] In view of this, the present application is specifically proposed. Summary of the Invention

[0006] The purpose of the present invention is to provide an automated control system for industrial equipment to solve the problems raised in the above background art.

[0007] To solve the above technical problems, the automated control system for industrial equipment provided by the present invention includes:

[0008] A data acquisition layer for real-time acquisition of the operation data of industrial equipment through intelligent sensing devices and intelligent data acquisition devices; with the help of intelligent sensing devices and intelligent data acquisition devices, the operation data of industrial equipment can be obtained in real time and accurately, providing a reliable basis for subsequent analysis and control, helping to timely detect subtle changes in equipment operation and predict potential problems in advance;

[0009] The computing layer adopts edge computing technology to perform real-time computing on the operation data of industrial equipment collected by the data acquisition layer. By using edge computing technology to perform real-time computing near the source of data generation, the data processing time is greatly shortened, the response speed of the system is improved, enabling the system to quickly respond to changes in equipment status and effectively avoiding control lags caused by data transmission delays.

[0010] The control layer is used to further process and analyze the calculation results of the computing layer, identify the operation status of industrial equipment through digital twin technology, and generate control instructions based on the operation status of industrial equipment. By using digital twin technology to deeply process and analyze the calculation results of the computing layer, the operation status of industrial equipment can be accurately identified. Digital twin technology constructs a virtual model of the equipment, conducts real-time interaction and comparison with the actual equipment operation data, can comprehensively and intuitively reflect the operation status of the equipment, and provides more accurate information for subsequent control decisions.

[0011] The execution layer is used to execute the control instructions generated by the control layer and adjust the actions of industrial equipment according to the control instructions. The execution layer can accurately execute the control instructions generated by the control layer and precisely adjust the actions of industrial equipment according to the instructions, ensuring that the equipment operates in the expected manner, improving the consistency and reliability of equipment operation, and meeting the requirements of industrial production for equipment accuracy and stability.

[0012] The human-machine interface layer is used to provide a visual monitoring interface for users and support functions such as parameter setting and alarm management. The human-machine interface layer provides a visual monitoring interface for users, enabling users to intuitively understand information such as the operation status, various parameters, and historical data of industrial equipment, facilitating users' real-time monitoring and management, and improving users' control ability over equipment operation.

[0013] The communication network is used to achieve data interaction within the automation control system of industrial equipment and between automation control systems of industrial equipment through protocols including industrial Ethernet. By using protocols such as industrial Ethernet to achieve data interaction within and between industrial equipment automation control systems, the efficiency and stability of data transmission are ensured. Industrial Ethernet has advantages such as high bandwidth, low latency, and strong anti-interference ability, can meet the strict requirements of industrial automation systems for data transmission, realize collaborative work between devices, and improve the automation level and production efficiency of the entire industrial production.

[0014] Furthermore, when the data acquisition layer performs real-time acquisition of the operation data of industrial equipment through intelligent sensing devices and intelligent data acquisition devices, the intelligent sensing devices and intelligent data acquisition devices adopted include:

[0015] A11: Multimodal sensors, including vibration sensors, temperature sensors, pressure sensors, rotational speed and torque sensors, and pressure sensors, are used to collect the operation data of industrial equipment; covering various types of sensors such as vibration, temperature, pressure, rotational speed and torque, they can collect the operation data of industrial equipment in an all-round and multi-dimensional manner. Different sensors reflect the equipment status from different physical quantity perspectives. For example, vibration sensors can detect abnormal vibrations of the equipment, temperature sensors can monitor the temperature changes of the equipment in real time, and pressure and rotational speed and torque sensors can monitor the internal pressure and operation power parameters of the equipment. These data complement each other, providing rich basis for accurately evaluating the equipment operation status, helping to detect potential equipment failures in advance and reducing the risk of equipment damage;

[0016] A12: High-definition image acquisition devices are used to collect images of the processed products of industrial equipment and the industrial equipment itself; collecting images of the processed products of industrial equipment and the equipment itself can intuitively display the appearance status of the equipment and the processing quality of the products. Operators and managers can quickly discover abnormal situations such as surface wear and deformation of the equipment, as well as problems such as product appearance defects and dimensional deviations through the images. At the same time, combined with digital twin technology, defects in the processing technology can be targeted and the processing technology can be optimized using intelligent algorithms, and measures can be taken in a timely manner for adjustment and improvement to ensure product quality and production efficiency;

[0017] A13: Environmental data acquisition devices are used to collect environmental data of the environment where industrial equipment is located. Among them, environmental data includes pollution gas concentration, dust concentration, air temperature, air humidity, and radiation intensity; analyzing the combination of environmental data and equipment operation data can understand the impact of environmental factors on equipment operation and product quality. For example, by analyzing the changes in equipment operation efficiency and product quality under different temperature and humidity conditions, production process parameters can be optimized, the equipment operation mode can be adjusted, the equipment performance and product quality stability can be improved, and energy conservation, consumption reduction, and green production can be achieved.

[0018] Furthermore, when the computing layer performs real-time computing on the operation data of industrial equipment collected by the data acquisition layer using edge computing technology, it specifically includes:

[0019] S11: Localized data caching, temporarily storing the high-frequency sampling data collected by the data acquisition layer within the edge computing node; temporarily storing high-frequency sampling data within the edge computing node can effectively address the problem of data transmission interruption caused by network fluctuations or failures. Even if the network experiences a brief anomaly, the data cached locally will not be lost, ensuring the integrity of the operation data of industrial equipment and providing a reliable data basis for subsequent computing and analysis;

[0020] S12: Data preprocessing. The high-frequency sampling data collected by the data acquisition layer is preprocessed. The preprocessing methods adopted include missing value processing, outlier processing, and noise reduction processing. Performing preprocessing operations such as missing value processing, outlier processing, and noise reduction processing on the high-frequency sampling data can remove noise and error information in the data, improving the quality and accuracy of the data. Missing value processing ensures the integrity of the data. Outlier processing eliminates the interference of abnormal data on the calculation results. Noise reduction processing makes the data smoother and more stable, providing a more reliable data input for subsequent real-time calculations;

[0021] S13: Real-time calculation task. A preset algorithm is executed to perform anomaly detection and feature extraction on the preprocessed data. Executing the preset algorithm to perform anomaly detection on the preprocessed data can timely detect abnormal situations during the operation of industrial equipment. By real-time monitoring the operation status of the equipment, once an anomaly is detected, the system can quickly issue an alarm to notify relevant personnel to take measures, avoiding the further expansion of equipment failures and reducing production losses;

[0022] S14: Result compression and transmission. A preset data transmission mechanism is used to transmit the calculation results. Using the preset data transmission mechanism to compress and transmit the calculation results can effectively reduce the amount of data transmitted and lower the demand for network bandwidth. In the case of limited network bandwidth, compressed transmission can ensure that the calculation results can be transmitted to the target node in a timely and accurate manner, avoiding data transmission delays and losses caused by network congestion.

[0023] Furthermore, when using the preset data transmission mechanism to transmit the calculation results, the data transmission mechanism adopted is:

[0024] S21: Resource statistics. The communication network is detected to detect the network latency and usage ratio of the communication network, and the high-frequency sampling data temporarily stored in the edge computing node is screened to filter out useless data and count the size of the remaining data. The network latency of the communication network is obtained by calculating the average network latency in the previous three minutes. By detecting the communication network and calculating the average network latency in the previous three minutes, the actual latency status of the current network can be accurately evaluated, providing a reliable decision-making basis for subsequent task selection. At the same time, screening and sizing the high-frequency sampling data temporarily stored in the edge computing node, removing useless data, effectively reduces the amount of data to be transmitted, improves the pertinence and efficiency of data transmission, and counting the size of the remaining data lays a foundation for selecting an appropriate transmission strategy according to the network conditions. This pre-resource statistics operation makes the entire data transmission process more orderly and efficient, avoiding resource waste and network congestion caused by blind transmission;

[0025] S22: Task selection. Tasks are selected based on the network latency of the communication network. Specifically, when the network latency of the communication network is greater than a set threshold, the calculation results are temporarily cached and transmitted to the control layer via the communication network. When the network latency of the communication network is less than the set threshold, the calculation results are directly sent; tasks are selected according to the network latency of the communication network. When the network latency is greater than the set threshold, the calculation results are temporarily cached and transmitted to the control layer via the network, avoiding problems such as data loss or transmission failure that may occur during direct transmission when the network latency is high. When the network latency is less than the set threshold, the calculation results are directly sent, making full use of the good network state and improving the timeliness of data transmission;

[0026] S23: Local cache data processing. When the network latency of the communication network is less than the set threshold, the formula:

[0027] is used to judge the usage ratio of the communication network. When the formula holds, the filtered high-frequency sampling data is transmitted to a specified database for storage and deleted after the transmission is completed. When the formula does not hold, the filtered high-frequency sampling data is compressed and stored, waiting for subsequent transmission;

[0028] where N is the size of the filtered high-frequency sampling data, M is the set maximum allowable transmission volume, τ is the current usage ratio of the communication network, and where α is a set parameter, α = 0.75, is the set maximum allowable network latency, RTT is the real-time network latency, is the maximum allowable transmission volume under ideal conditions; a specific formula is used to judge the usage ratio of the communication network. When the formula holds, the filtered high-frequency sampling data is transmitted to a specified database for storage and deleted after the transmission is completed, effectively utilizing the network bandwidth and avoiding the long-term accumulation of data at the local node. When the formula does not hold, the data is compressed and stored, waiting for subsequent transmission. This strategy can reasonably allocate network resources and avoid interfering with other network applications when the network usage ratio is too high.

[0029] Furthermore, when the control layer further processes and analyzes the calculation results of the calculation layer, identifies the operating state of industrial equipment through data twin technology, and generates control instructions based on the operating state of industrial equipment, it specifically includes:

[0030] S31: Data twin modeling. Based on the types, signals, and parameters of industrial equipment, retrieve the high-precision digital twins of industrial equipment, synchronize real-time data with the virtual model, and continuously update the model parameters through a two-way communication channel. Retrieving the high-precision digital twins based on the types, signals, and parameters of industrial equipment can accurately construct a virtual equipment model, comprehensively reflect the equipment characteristics, and provide a reliable basis for subsequent analysis. Synchronizing real-time data with the virtual model and continuously updating the model parameters through a two-way communication channel ensure a high degree of consistency between the virtual model and the actual equipment status, realizing real-time and accurate mapping of the equipment status;

[0031] S32: Status recognition. Compare the real-time data with the ideal parameters to identify deviations, and analyze historical data through a machine learning model to establish a fault feature library. Then, perform pattern matching between the real-time data and the data in the fault feature library for fault prediction. Comparing the real-time data with the ideal parameters to identify deviations can quickly detect abnormal equipment operation. Analyzing historical data through a machine learning model to establish a fault feature library and performing pattern matching between the real-time data and the data in the fault feature library for fault prediction can discover potential fault hazards in advance, realize the transformation from passive maintenance to proactive prevention, effectively reduce the equipment failure rate, and reduce production losses;

[0032] S33: Control algorithm invocation. Dynamically adjust the control strategy using a neural network algorithm. Dynamically adjusting the control strategy using a neural network algorithm can automatically adjust the control parameters according to the real-time operating status of the equipment and changes in the external environment, realizing real-time optimization of the control strategy. This dynamic adjustment ability enables the control system to better adapt to various uncertain factors during the equipment operation process, improving the stability and robustness of the control system;

[0033] S34: Instruction generation. Generate specific control instructions based on the control strategy generated by the neural network algorithm and send them to the execution layer through a communication network. Generating specific control instructions based on the control strategy generated by the neural network algorithm and sending them to the execution layer through a communication network ensure that the control instructions can be accurately transmitted to the actuator, realizing precise control of the equipment. This instruction generation method avoids errors caused by human intervention, improving the accuracy and reliability of the control instructions.

[0034] Further, when dynamically adjusting the control strategy using a neural network algorithm, it specifically includes:

[0035] S41: Input data fuzzification. Define fuzzy sets for input variables, design Gaussian membership functions, and based on expert experience and historical data, establish fuzzy rules to form an initial control strategy. Defining fuzzy sets for input variables and designing Gaussian membership functions enable the control strategy to better handle uncertain and fuzzy input information. In an actual industrial environment, the operating state data of equipment often contains noise and uncertainty. Through fuzzification, the control strategy can have stronger adaptability to such data, avoiding the failure of the control strategy due to data fluctuations;

[0036] S42: Fuzzy inference and instruction generation. Through the Mamdani inference method, map the input fuzzy variables to fuzzy outputs, and then through the maximum membership degree method, convert the fuzzy outputs into precise control instructions. Mapping the input fuzzy variables to fuzzy outputs through the Mamdani inference method and then converting the fuzzy outputs into precise control instructions through the maximum membership degree method. This way of fuzzy inference and instruction generation realizes the conversion from fuzzy input to precise output, making the control strategy flexible and adaptable. It can generate appropriate control instructions according to different input situations, avoiding the overly rigid control methods in traditional control, and better adapting to the complex operating environment of industrial equipment;

[0037] S43: Neural network dynamic adjustment and optimization. Process the fuzzy outputs and precise control instructions through the improved neural network to output optimized control instructions. The improved neural network can consider the complex relationships between multiple input variables and output variables to achieve global optimization. Compared with traditional local optimization methods, it can adjust the control strategy as a whole, avoid falling into local optimal solutions, and improve the overall performance of the control strategy;

[0038] S44: Closed-loop verification. Input the optimized control instructions into the established high-precision digital twin for simulation verification to verify the response speed and stability of the control strategy. Input the optimized control instructions into the established high-precision digital twin for simulation verification to verify the response speed and stability of the control strategy. The digital twin can simulate the actual operation process of the equipment. Through simulation verification, the control strategy can be comprehensively evaluated before actual application, and problems in the strategy can be discovered and adjusted in a timely manner. This ensures the effectiveness and reliability of the control strategy in actual application and reduces the risks of equipment failures and production losses caused by unreasonable strategies;

[0039] S45: Iterative optimization, dynamically update the neural network weights and fuzzy rule base according to the actual operation data; the actual operation data includes the operation information of the device under various working conditions. By learning and analyzing these data, the neural network and fuzzy rule base can be continuously improved, and the adaptability and robustness of the control strategy to different working conditions can be enhanced. So that the control strategy can still maintain stable control performance when facing uncertainties such as equipment aging and environmental interference.

[0040] Furthermore, when processing the fuzzy output and precise control instructions through the improved neural network and outputting the optimized control instructions, the architecture of the improved neural network model is as follows:

[0041] A21: Input layer, used to input the precise control instructions generated by the maximum membership degree method, the real-time data collected by the data acquisition layer, and normalize the data, and then filter the noise through a sliding window. Among them, the number of nodes in the input layer is consistent with the input feature dimension; normalizing the precise control instructions generated by the maximum membership degree method and the real-time data collected by the data acquisition layer can unify the data with different dimensions and ranges into a specific interval, avoiding affecting the model performance due to data scale differences. At the same time, filtering the noise through a sliding window can effectively remove the random interference in the data, making the input data cleaner and more accurate, and providing a high-quality data basis for subsequent processing;

[0042] A22: Hidden layer, adopting the RBF network structure, determining the center position and expansion constant of the radial basis function through the subtractive clustering algorithm, based on the sensitivity variance test, deleting redundant nodes or adding new nodes to enhance the expression ability, and dynamically adjusting the weights from the hidden layer to the output layer using the linear least squares method, with the mean square error as the objective function; determining the center position and expansion constant of the radial basis function through the subtractive clustering algorithm. Compared with traditional methods, the subtractive clustering algorithm can more effectively extract the clustering centers from the data, making the center position and expansion constant of the RBF network more in line with the data distribution characteristics, and improving the fitting ability and generalization performance of the network;

[0043] A23: Output layer, used to fuse the fuzzy control logic and the neural network optimization results through a multi-mode fusion and dynamic parameter adjustment mechanism to obtain the final control instruction; fuse the fuzzy control logic and the neural network optimization results through a multi-mode fusion and dynamic parameter adjustment mechanism. The fuzzy control logic can handle uncertainties and fuzzy information, while the neural network optimization results have strong learning and adaptive capabilities. The fusion of the two gives full play to their respective advantages, making the final obtained control instruction have both the flexibility of fuzzy control and the precision of neural network optimization, and improving the performance and stability of the control system.

[0044] Furthermore, when the output layer fuses the fuzzy control logic and the neural network optimization results through the multi-mode fusion and dynamic parameter adjustment mechanism to obtain the final control instruction, the multi-mode fusion and dynamic parameter adjustment mechanism specifically includes:

[0045] A31: PID parameter dynamic adjustment mode, the output layer generates the adjustment amount of the proportional coefficient, integral coefficient and differential coefficient, and fits the error and parameter relationship through nonlinear functions; the adjustment amount of the proportional coefficient, integral coefficient and differential coefficient generated by the output layer can accurately adjust the parameters of the PID controller according to the real-time operating status and error of the system. This helps the PID controller to better adapt to the dynamic characteristics of the system, improve control accuracy, and reduce the steady-state error of the system; at the same time, dynamic adjustment of PID parameters can enable the controller to maintain good control performance under different working conditions. Whether the system is in a stable operating state or is subject to external interference, the stability of the system can be maintained by adjusting the parameters, and the system's adaptability to complex working conditions can be enhanced;

[0046] A32: Directly generate control instructions, and perform weighted fusion of fuzzy control instructions and neural network optimization results; weighted fusion of fuzzy control instructions and neural network optimization results fully utilizes the respective advantages of fuzzy control and neural network optimization. Fuzzy control can handle uncertainty and fuzzy information, and has good robustness and adaptability; neural network optimization has strong learning and adaptive capabilities, and can learn and optimize based on a large amount of data. The fusion of the two makes the final control instruction have both the flexibility of fuzzy control and the accuracy of neural network optimization, improving the overall performance of the control system.

[0047] Furthermore, when the execution layer executes the control instructions generated by the control layer and adjusts the action of the industrial equipment according to the control instructions, it specifically includes:

[0048] S51: Instruction parsing: receiving and parsing the control instructions generated by the control layer; the parsing process can extract key information in the control instructions, such as target position, action speed, execution time, etc., providing accurate data support for subsequent priority sorting and action execution;

[0049] S52: Priority sorting, dynamically scheduling the execution order according to the priority level and urgency of the control instructions; Dynamically scheduling the execution order according to the priority level and urgency of the control instructions can reasonably allocate the resources of the execution layer to ensure that high-priority and urgent instructions are executed first. This avoids waste and conflict of resources and improves the overall efficiency of the execution layer. Dynamic scheduling of the execution order can reduce the waiting time of low-priority instructions for high-priority instructions, allowing industrial equipment to respond to critical tasks faster, improving the real-time performance and response speed of the system;

[0050] S53: Action execution and feedback, driving the actuator to perform actions, completing control execution, and real-time feedback of the action information of the actuator through the encoder; driving the actuator to perform actions, and being able to precisely control the actions of industrial equipment according to the parsed instructions and the priority sorting results. This ensures that the industrial equipment can move according to the predetermined trajectory and parameters, improving product quality and production efficiency.

[0051] Furthermore, the communication network includes:

[0052] A41: Short-distance communication network, adopting multiple of UWB communication technology, Bluetooth communication technology, Wi-Fi communication technology, and infrared data transmission technology, for realizing data interaction between the data acquisition layer and the edge computing nodes in the computing layer; adopting multiple of UWB (Ultra Wideband) communication technology, Bluetooth communication technology, Wi-Fi communication technology, and infrared data transmission technology, being able to realize fast and efficient data interaction between the data acquisition layer and the edge computing nodes in the computing layer. The edge computing nodes can timely obtain various information collected by the data acquisition layer, and perform real-time processing and analysis, providing strong support for subsequent control decisions;

[0053] A42: Main data transmission network, used to select ModbusRTU and Profinet according to the scenario, construct a star-shaped redundant network, and realize data transmission between the computing layer and the control layer; constructing a star-shaped redundant network, when a certain node or link in the network fails, the redundant node or link can quickly take over the communication task to ensure the normal operation of the network. This network structure improves the reliability and stability of the main data transmission network, reducing production interruptions caused by network failures;

[0054] A43: Control network, used to construct a direct transmission network between the control layer and the execution layer to achieve low-latency issuance of control instructions. Constructing a direct transmission network between the control layer and the execution layer avoids interference and latency in the intermediate links, can achieve low-latency issuance of control instructions, and fewer intermediate devices also means fewer potential failure points, improving the reliability of the entire system.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] 1. Optimize data transmission: A unique data transmission mechanism is adopted between the edge computing nodes and the control layer, implementing different transmission methods and data storage methods according to the network conditions, effectively utilizing the network bandwidth, avoiding long-term accumulation of data at local nodes, and at the same time, effectively preventing data transmission loss or error when the network latency is high, ensuring the effectiveness and timeliness of data transmission;

[0057] 2. Precise control instructions: By weighted fusion of fuzzy control instructions and neural network optimization results, the respective advantages of fuzzy control and neural network optimization are fully utilized. The fusion of the two makes the final control instructions have both the flexibility of fuzzy control and the precision of neural network optimization, improving the overall performance of the control system.

[0058] 3. Increase the reliability of the network architecture: Short-range communication technology is adopted between the acquisition device and the edge computing node, which can achieve fast and efficient data interaction between the data acquisition layer and the edge computing node in the computing layer. The edge computing node can timely obtain various information collected by the data acquisition layer and perform real-time processing and analysis. A direct transmission network is constructed between the control layer and the execution layer, avoiding interference and delay in the intermediate links, enabling low-latency issuance of control instructions, and fewer intermediate devices also mean fewer potential fault points, improving the reliability of the entire system. Brief Description of the Drawings

[0059] Figure 1 It is a schematic structural diagram of an automated control system for an industrial device. Detailed Embodiments

[0060] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0061] Please refer to Figure 1 , the present invention provides a technical solution: an automated control system for an industrial device:

[0062] Embodiment 1

[0063] Suppose in a mobile phone assembly factory, the automated control system of this industrial device is assembled in the intelligent assembly line. The system includes a data acquisition layer, a computing layer, a control layer, an execution layer, a human-machine interface layer, and a communication network. Next, in conjunction with the specific application of the intelligent assembly line, the functions and implementation methods of each layer will be described in detail.

[0064] (1) Data acquisition layer

[0065] In the intelligent assembly line, the data acquisition layer uses intelligent sensing devices and intelligent data acquisition devices to collect the operation data of industrial devices in real time, including:

[0066] Intelligent sensing devices:

[0067] Multimodal sensors: Vibration sensors, temperature sensors, pressure sensors, rotational speed and torque sensors, etc. are adopted and installed at key positions on the intelligent assembly line, such as assembly robotic arms, conveyor belts, assembly tools, etc., for collecting operation data such as vibration, temperature, pressure, rotational speed and torque of the equipment.

[0068] High-definition image acquisition equipment: High-definition cameras are installed at key process positions on the assembly line to collect images of the assembled products and the equipment itself for quality inspection and equipment status monitoring.

[0069] Environmental data acquisition equipment: Environmental data acquisition equipment is installed in the working environment of the assembly line to collect environmental data such as pollution gas concentration, dust concentration, air temperature, air humidity and radiation intensity to ensure that the working environment meets the production requirements.

[0070] Intelligent data acquisition equipment:

[0071] An intelligent data acquisition module with data preprocessing function is adopted to preliminarily process the data collected by the multimodal sensors and high-definition image acquisition equipment, such as data format conversion, data compression, etc., and then transmit it to the computing layer through the communication network.

[0072] (2) Computing layer

[0073] The computing layer adopts edge computing technology to perform real-time computing on the operation data of industrial equipment collected by the data acquisition layer. The specific functions include:

[0074] 1. Localized data caching: Temporarily store the high-frequency sampled data collected by the data acquisition layer within the edge computing node, such as the motion trajectory data of the assembly robotic arm, the pressure data of the assembly tool, etc.

[0075] 2. Data preprocessing: Perform preprocessing on the high-frequency sampled data, including missing value processing, outlier processing and noise reduction processing. For example, for the motion trajectory data of the assembly robotic arm, the interpolation method is used to process missing values, the statistical-based method is used to identify and process outliers, and the wavelet transform method is used for noise reduction processing.

[0076] 3. Real-time computing tasks: Execute the preset algorithms to perform anomaly detection and feature extraction on the preprocessed data. For example, the machine learning-based anomaly detection algorithm is used to identify anomalies in the assembly process, such as assembly deviation, tool wear, etc.; the feature extraction algorithm is used to extract key features in the assembly process, such as assembly force, assembly accuracy, etc.

[0077] 4. Result Compression and Transmission: The calculated results are transmitted using a preset data transmission mechanism. Based on the network latency and usage ratio of the communication network, an appropriate transmission method is selected. When the network latency is large, the calculated results are temporarily cached and transmitted after the network conditions improve; when the network latency is small, it is decided whether to directly transmit the data or transmit it after compression according to the network usage ratio.

[0078] (III) Control Layer

[0079] The control layer further processes and analyzes the calculated results of the calculation layer, identifies the operating state of industrial equipment through data twin technology, and generates control instructions based on the operating state of industrial equipment. The specific functions include:

[0080] 1. Data Twin Modeling: Based on the types, signals, and parameters of the intelligent assembly line, a high-precision digital twin is retrieved, real-time data is synchronized with the virtual model, and the model parameters are continuously updated through a two-way communication channel.

[0081] 2. State Identification: The real-time data is compared with the ideal parameters to identify deviations, and historical data is analyzed through a machine learning model to establish a fault feature library for fault prediction.

[0082] 3. Control Algorithm Invocation: The neural network algorithm is used to dynamically adjust the control strategy, and the control parameters are optimized in real time according to different assembly tasks and equipment states.

[0083] 4. Instruction Generation: Specific control instructions are generated based on the control strategy generated by the neural network algorithm and sent to the execution layer through the communication network.

[0084] (IV) Execution Layer

[0085] The execution layer executes the control instructions generated by the control layer and adjusts the actions of industrial equipment according to the control instructions. The specific functions include:

[0086] Instruction Parsing: Receive the control instructions generated by the control layer and parse them.

[0087] Priority Sorting: Dynamically schedule the execution order according to the priority level and urgency within the control instructions.

[0088] Action Execution and Feedback: Drive the actuator to perform actions, complete the control execution, and real-time feedback the action information of the actuator through the encoder.

[0089] (V) Human-Machine Interface Layer

[0090] The human-machine interface layer provides a visual monitoring interface for users and supports functions such as parameter setting and alarm management. Users can monitor the running status of the assembly line in real time through the human-machine interface layer, set parameters such as assembly speed and assembly accuracy, receive alarm information, and handle abnormal situations in a timely manner.

[0091] (6) Communication Network

[0092] The communication network realizes data interaction within the industrial equipment automation control system and between industrial equipment automation control systems through multiple protocols, including:

[0093] Short-range communication network: Adopts multiple of UWB communication technology, Bluetooth communication technology, Wi-Fi communication technology, and infrared data transmission technology to realize data interaction between the data acquisition layer and the edge computing nodes in the computing layer.

[0094] Data transmission main network: Selects ModbusRTU and Profinet according to the scenario to construct a star-shaped redundant network to realize data transmission between the computing layer and the control layer.

[0095] Control network: Constructs a direct transmission network between the control layer and the execution layer to realize low-latency issuance of control instructions.

[0096] Compared with the prior art, in the data transmission stage, a unique data transmission mechanism is adopted between the edge computing nodes and the control layer, and different transmission methods and data storage methods are implemented according to the network conditions; in the control instruction generation stage, an improved algorithm combining fuzzy control algorithm and neural network algorithm is used to generate control instructions; in terms of network architecture, a unique architecture of short-range communication network + main network + control instruction transmission network is adopted. These innovation points are not covered by the prior art and reflect the novelty of the present invention.

[0097] The present invention comprehensively uses a variety of technical means to address the high requirements for network speed in data transmission and the lack of precision in control algorithms in response to the real-time output requirements and complexity of industrial equipment control. Innovative improvements and optimizations have been made in data transmission, control programs, network architecture, etc. It is not a simple technical superposition, but through in-depth research and practice, a series of new methods and strategies have been proposed, providing new ideas and solutions for the field of data publishing.

[0098] Embodiment 2

[0099] The industrial equipment automation control system of the present invention adopts a hierarchical architecture design, and each layer realizes efficient collaborative operation through modular integration. Specifically, when implemented, the system is constructed and operates in the following manner:

[0100] I. Implementation details of the data acquisition layer

[0101] The multimodal sensor matrix configured in this system includes vibration sensors (±5% accuracy, 0-10kHz frequency response), temperature sensors (PT100, ±0.1℃ accuracy), pressure sensors (range 0-60MPa, 0.2-level accuracy), speed torque sensors (photoelectric, speed measurement range 0-30000rpm) and environmental data acquisition equipment (equipped with electrochemical gas sensors, laser dust monitors, temperature and humidity integrated sensors). The sensors adopt a distributed deployment strategy and are installed in key locations such as machine tool spindles, hydraulic stations, and transmission gearboxes. They communicate with intelligent data collectors through the ModbusRTU protocol. The high-definition image acquisition equipment uses a 20-megapixel industrial camera, with a ring-shaped LED light source, deployed at the finished product inspection station of the production line, and transmits image data through the GigEVision protocol.

[0102] 2. Edge computing layer operation mechanism

[0103] The edge computing node uses the NVIDIA Jetson AGX Xavier platform, equipped with 32GB LPDDR4X memory and a 512-core Volta architecture GPU. The localized data cache uses the Redis memory database, with a 10-minute rolling storage window. The data preprocessing module includes:

[0104] Missing value processing: linear interpolation algorithm is used to compensate for the instantaneous disconnection of the sensor data

[0105] Outlier detection: Establish an outlier model based on the isolation forest algorithm (contamination = 0.05)

[0106] Noise reduction processing: Use wavelet packet transform to perform 5-layer decomposition and reconstruction of vibration signals

[0107] The real-time computing task uses the LSTM network optimized by TensorRT to predict tool wear. The window length is set to 50 sampling cycles, and the inference delay is controlled within 8ms.

[0108] 3. Data transmission mechanism implementation

[0109] The resource statistics module performs a network probe every 30 seconds, uses the ICMP protocol to measure the end-to-end delay, and obtains the network bandwidth utilization through the SNMP protocol. In the dynamic transmission strategy, the network delay threshold is set to 50ms. When the network delay is detected to exceed the threshold, the local cache queue is enabled (the maximum cache depth is 1000 records). The parameters in the local cache data processing formula are set to: α = 0.7, β = 1.2, γ = 0.3, and the maximum allowed transmission volume M is dynamically adjusted according to the network bandwidth (the basic value is 512KB).

[0110] 4. Implementation of the core algorithm of the control layer

[0111] Data twin modeling constructs a multi-domain coupling model based on the Modelica language, including co-simulation of a mechanical system (SimMechanics), a hydraulic system (AMESim), and a control system (MATLAB / Simulink). The state recognition module uses an improved 1D-CNN network with an input window set as a 128×12-dimensional feature matrix, and is pre-trained on a public device dataset (such as the PHM Society dataset) through transfer learning. The control algorithm invocation uses the Deep Deterministic Policy Gradient (DDPG) algorithm, designs a neural network with 4 hidden layers (the number of nodes is 256-128-64-32), and the output layer generates a three-dimensional control vector (position, speed, acceleration).

[0112] V. Execution Layer Control Strategy

[0113] The instruction parsing module uses a custom protocol parser based on ANTLR, supporting binary and JSON format instructions. The priority sorting uses a multi-level feedback queue scheduling algorithm, setting 4 levels of priority (emergency stop > safety protection > process control > status monitoring). The action execution drives the servo motor through the EtherCAT bus, and the encoder feedback resolution reaches 24 bits, with a position control accuracy of ±0.001 mm.

[0114] VI. Communication Network Architecture

[0115] The short-range communication network adopts a dual-mode solution of UWB (IEEE802.15.4z) and Bluetooth 5.2 at the data acquisition layer, with an effective communication distance of 15 meters and a data transmission rate of 1 Mbps. The main data transmission network constructs a dual-ring redundant topology and uses Profinet IRT to achieve cycle communication of <1 ms. The control network uses TSN (Time-Sensitive Network) technology and configures an 802.1Qbv gating list to ensure that the end-to-end delay of control instructions is <200 μs.

[0116] VII. System Integration and Verification

[0117] During system deployment, first perform full-system simulation verification on the digital twin platform, and use OPC UA to achieve virtual-real data synchronization. After actual deployment, perform frequency-domain analysis through the Chirp-Z transform to verify that the phase margin of the control instruction is >45° and the amplitude margin is >6 dB. Long-term operation tests show that in a typical industrial noise environment (SNR = 30 dB), the system control accuracy remains ±0.5% FS, and the fault prediction accuracy rate reaches 92.3%.

[0118] In summary, through the deep integration of edge computing and digital twin, this system realizes the closed-loop optimization of device status perception, decision-making control, and execution feedback. Compared with traditional control systems, the control response speed is increased by 40%, and the advance amount of abnormal state recognition reaches 3 sampling periods, effectively improving the operation efficiency and reliability of industrial devices.

Claims

1. An automated control system for an industrial device, characterized in that: Including: A data acquisition layer for real-time acquisition of the operating data of industrial equipment through intelligent sensing devices and intelligent data acquisition devices; A computing layer that uses edge computing technology to perform real-time calculations on the operating data of industrial equipment collected by the data acquisition layer; A control layer for further processing and analysis of the calculation results of the computing layer, identifying the operating state of industrial equipment through digital twin technology, and generating control instructions based on the operating state of industrial equipment; An execution layer for executing the control instructions generated by the control layer and adjusting the actions of industrial equipment according to the control instructions; A human-machine interface layer for providing a visual monitoring interface for users and supporting parameter setting and alarm management functions; A communication network for realizing data interaction within the automated control system of industrial equipment and between the automated control systems of industrial equipment through protocols including industrial Ethernet.

2. The automated control system of an industrial device according to claim 1, characterized in that: When the data acquisition layer performs real-time acquisition of the operating data of industrial equipment through intelligent sensing devices and intelligent data acquisition devices, the intelligent sensing devices and intelligent data acquisition devices used include: A11: Multimodal sensors, including vibration sensors, temperature sensors, pressure sensors, rotational speed and torque sensors, and pressure sensors, for collecting the operating data of industrial equipment; A12: High-definition image acquisition devices for image acquisition of the processed products of industrial equipment and the industrial equipment itself; A13: Environmental data acquisition devices for collecting environmental data of the environment where industrial equipment is located, where the environmental data includes pollutant gas concentration, dust concentration, air temperature, air humidity, and radiation intensity.

3. The automated control system of an industrial device according to claim 2, wherein: When the computing layer performs real-time calculations on the operating data of industrial equipment collected by the data acquisition layer using edge computing technology, it specifically includes: S11: Localized data caching for temporarily storing the high-frequency sampling data collected by the data acquisition layer within the edge computing node; S12: Data preprocessing for preprocessing the high-frequency sampling data collected by the data acquisition layer, where the preprocessing methods used include missing value processing, outlier processing, and noise reduction processing; S13: Real-time calculation tasks for performing anomaly detection and feature extraction on the preprocessed data by executing a preset algorithm; S14: Result compression and transmission for transmitting the calculation results using a preset data transmission mechanism.

4. An automated control system for an industrial device according to claim 3, characterized in that: When using the preset data transmission mechanism to transmit the calculation results, the data transmission mechanism used is: S21: Resource statistics for detecting the communication network, detecting the network latency and usage ratio of the communication network, screening the high-frequency sampling data temporarily stored within the edge computing node, filtering out useless data, and counting the size of the remaining data, where the network latency of the communication network is obtained by calculating the average network latency in the previous three minutes; S22: Task selection for task selection based on the network latency of the communication network. Among them, when the network latency of the communication network is greater than the set threshold, the calculation results are temporarily cached and transmitted to the control layer through the communication network. When the network latency of the communication network is less than the set threshold, the calculation results are directly sent. S23: Local cache data processing. When the network latency of the communication network is less than the set threshold, the formula: is used to judge the usage ratio of the communication network. When the formula holds, the filtered high-frequency sampling data is transmitted to the specified database for storage and deleted after the transmission is completed. When the formula does not hold, the filtered high-frequency sampling data is compressed and stored, waiting for subsequent transmission; where N is the size of the high-frequency sampled data after screening, M is the set maximum allowable transmission volume, and τ is the current usage ratio of the communication network, where where α is a set parameter, α = 0.75, is the maximum allowable network latency, RTT is the real-time network latency, is the maximum allowable transmission volume under ideal conditions.

5. The automated control system for an industrial device according to claim 4, wherein: When the control layer further processes and analyzes the calculation results of the calculation layer and identifies the operating state of industrial equipment through the digital twin technology and generates control instructions according to the operating state of industrial equipment, it specifically includes: S31: Digital twin modeling. Based on the types, signals, and parameters of industrial equipment, the high-precision digital twin body of the industrial equipment is retrieved, the real-time data is synchronized with the virtual model, and the model parameters are continuously updated through the bidirectional communication channel; S32: State identification. The real-time data is compared with the ideal parameters to identify the deviation, and the historical data is analyzed through the machine learning model to establish a fault feature library, and the real-time data is pattern-matched with the data in the fault feature library for fault prediction; S33: Control algorithm call. The neural network algorithm is used to dynamically adjust the control strategy; S34: Instruction generation. According to the control strategy generated by the neural network algorithm, specific control instructions are generated and sent to the execution layer through the communication network.

6. The automated control system of an industrial device according to claim 5, wherein: When using the neural network algorithm to dynamically adjust the control strategy, it specifically includes: S41: Input data fuzzification. Define fuzzy sets for input variables, design Gaussian membership functions.,., and then establish fuzzy rules based on expert experience and historical data to form an initial control strategy; S42: Fuzzy reasoning and instruction generation. Through the Mamdani reasoning method, the input fuzzy variables are mapped to fuzzy outputs, and then through the maximum membership degree method, the fuzzy outputs are converted into precise control instructions; S43: Neural network dynamic adjustment and optimization. The improved neural network processes the fuzzy outputs and precise control instructions, and outputs the optimized control instructions; S44: Closed-loop verification. The optimized control instructions are input into the established high-precision digital twin body for simulation verification to verify the response speed and stability of the control strategy; S45: Iterative optimization. Dynamically update the neural network weights and fuzzy rule base according to the actual operation data.

7. An automated control system for an industrial device according to claim 6, characterized in that: When the improved neural network processes the fuzzy outputs and precise control instructions and outputs the optimized control instructions, the architecture of the improved neural network model is: A21: Input layer. It is used to input the precise control instructions generated by the maximum membership degree method and the real-time data collected by the data acquisition layer, normalize the data, and then filter the noise through the sliding window. Among them, the number of nodes in the input layer is consistent with the input feature dimension; A22: Hidden layer. Adopt the RBF network structure, determine the center position and expansion constant of the radial basis function through the subtractive clustering algorithm, enhance the expression ability by deleting redundant nodes or adding new nodes based on the sensitivity variance test, and dynamically adjust the weights from the hidden layer to the output layer using the linear least squares method, and the objective function is the mean square error; A23: Output layer, which is used to fuse the fuzzy control logic and the optimized results of the neural network through a multi-mode fusion and dynamic parameter adjustment mechanism to obtain the final control instruction.

8. An automated control system for an industrial device according to claim 7, characterized in that: When the output layer fuses the fuzzy control logic and the optimized results of the neural network through a multi-mode fusion and dynamic parameter adjustment mechanism to obtain the final control instruction, the multi-mode fusion and dynamic parameter adjustment mechanism specifically includes: A31: PID parameter dynamic adjustment mode. The output layer generates the adjustment amounts of the proportional coefficient, integral coefficient, and differential coefficient, and fits the relationship between the error and the parameters through a non-linear function. A32: Directly generate the control instruction, and perform weighted fusion on the fuzzy control instruction and the optimized results of the neural network.

9. An automated control system for an industrial device according to claim 8, characterized in that: When the execution layer executes the control instruction generated by the control layer and adjusts the actions of industrial equipment according to the control instruction, it specifically includes: S51: Instruction parsing, receiving the control instruction generated by the control layer and performing parsing. S52: Priority sorting, dynamically scheduling the execution order according to the priority level and urgency in the control instruction. S53: Action execution and feedback, driving the actuator to perform actions, completing the control execution, and real-time feedback of the action information of the actuator through the encoder.

10. An automated control system for an industrial device according to claim 9, characterized in that: The communication network includes: A41: Short-distance communication network, adopting multiple of UWB communication technology, Bluetooth communication technology, Wi-Fi communication technology, and infrared data transmission technology, for realizing data interaction between the data acquisition layer and the edge computing nodes in the computing layer. A42: Main data transmission network, which is used to select ModbusRTU and Profinet according to the scenario to construct a star-shaped redundant network to realize data transmission between the computing layer and the control layer. A43: Control network, which is used to construct a direct transmission network between the control layer and the execution layer to realize low-latency issuance of control instructions.

Images