Industrial control system and method based on industrial internet

Abstract

The application discloses an industrial control system and method based on an industrial internet, and relates to the technical field of industrial control.The application comprises a perception layer, a network layer, a platform layer and an application layer, and each layer cooperatively realizes intelligent control of an automobile welding process.The application adopts optical fiber transmission to replace traditional cables in view of strong electromagnetic interference of a welding scene, and improves data transmission anti-interference capability.Sensor sampling frequency is dynamically adapted to a welding process to balance precision and energy consumption.The application automatically adapts specific operation parameters of a welding robot to a process of reverse optimization of a spatter amount to reduce the spatter amount, thereby reducing spatter risk in the welding process.The network layer adopts TSN+industrial Ethernet through a backbone network to allocate a highest priority time slot for a motion instruction of the welding robot, ensures short-time deterministic transmission, avoids collision risk of mobile devices and mechanical arms at the same time, and shortens protocol conversion delay.

Description

Technical Field

[0001] This invention relates to the field of industrial control technology, and more specifically to an industrial control system and method based on the Industrial Internet. Background Technology

[0002] Among the four major processes in automobile manufacturing—stamping, welding, painting, and final assembly—the welding workshop is a core link connecting the vehicle body structure and precision, and its quality directly determines the body's strength and safety. Traditional welding workshops rely on manual inspection and fixed program control, which suffers from problems such as missed defects such as spatter, low efficiency of robot collaboration, and redundant energy consumption. Therefore, it is extremely necessary to conduct inspection and control of the welding workshop.

[0003] Existing technology, such as the invention application patent with publication number CN118606864B, discloses an intelligent control system and method for welding equipment based on multi-source information. This system collects multi-source information in real time during the welding process, extracts features from this information to generate multi-source feature information, and determines whether the fused multi-source feature information indicates normal or abnormal welding. It then provides early warnings for abnormal situations during the welding process, achieving intelligent early warning and preventing such abnormalities. This ultimately leads to optimal welding results, improving welding quality, efficiency, and stability.

[0004] Based on the above solutions, it can be found that there are still some shortcomings in the intelligent control of the automotive welding process in the existing technology, which are specifically reflected in the following aspects: strong electromagnetic interference often occurs during welding. The existing technology uses traditional cables, which have weak anti-interference ability for data transmission. Moreover, the sensor sampling frequency is not dynamically adapted to the welding process. At the same time, there is a lack of reverse optimization process for spatter amount, making it difficult to automatically adapt to the specific operating parameters of the welding robot to achieve the purpose of reducing spatter amount, thereby increasing the risk in the welding process. Summary of the Invention

[0005] The purpose of this invention is to provide an industrial control system and method based on the Industrial Internet, which solves the problems existing in the background technology.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides an industrial control system based on the Industrial Internet, including a perception layer, a network layer, a platform layer and an application layer that are connected in sequence, and each layer works together to realize intelligent control of the automotive welding process.

[0007] The sensing layer includes a high-frequency current sensor integrated at the end effector of the welding robot, an infrared thermal imager deployed around the welding torch, a six-axis force sensor at the joints of the robotic arm, a laser displacement sensor deployed around the workpiece, and workshop environment sensors. These sensors are used to capture arc stability, record the temperature field distribution of the molten pool, monitor welding pressure fluctuations, monitor workpiece deformation data, and collect workshop dust concentration and humidity. The sensing layer uses fiber optic data transmission and can dynamically adapt the sensor sampling frequency according to the welding process.

[0008] The platform layer includes a welding quality digital twin engine, a process parameter knowledge base, and an energy consumption optimization model. The welding quality digital twin engine reconstructs a 3D model of the welding area based on 3D point clouds and maps the molten pool morphology and heat-affected zone range in real time. The process parameter knowledge base integrates historical welding cases and generates an optimal parameter recommendation model. The energy consumption optimization model correlates welding-related parameters to calculate a benchmark value for energy consumption per unit weld point.

[0009] The application layer includes a spatter closed-loop control module and a multi-robot collaborative scheduling module. When the spatter amount exceeds a threshold, the spatter closed-loop control module uses the platform-layer optimal parameter recommendation model to deduce the optimal welding parameters and sends them to the robots. The multi-robot collaborative scheduling module simulates path conflicts using digital twins and dynamically optimizes the welding sequence.

[0010] A second aspect of the present invention provides a method for executing the industrial control system based on the Industrial Internet of Things as described in the present invention, comprising: an initialization phase: after the system is started, the sensors in the perception layer complete self-testing and calibration, the network layer establishes a communication connection, the platform layer loads a digital twin model and a process parameter knowledge base, and the application layer receives the initial welding task.

[0011] Welding execution phase: The perception layer collects welding process data in real time and transmits it to the platform layer via the network layer. The platform layer processes the data, generates quality assessments, parameter suggestions, and early warning information through model calculations, and transmits them to the application layer. The application layer executes welding control, parameter adjustments, or anomaly handling based on the platform layer's output and feeds back the execution results to the platform layer.

[0012] Closed-loop optimization phase: The platform layer updates the digital twin model and process parameter knowledge base based on feedback from the application layer, optimizing the model algorithm and parameters. The perception layer adjusts the sensor acquisition strategy according to the optimization instructions from the platform layer.

[0013] End Phase: After the welding task is completed, the system generates a welding process report, which includes quality data, energy consumption data and robot operation data for each weld point. The report is stored in the historical database at the platform level to provide data support for subsequent production analysis. At the same time, the system enters a low-power standby state, waiting for the next round of welding tasks to start.

[0014] The beneficial effects of this invention are as follows: (1) This invention addresses the strong electromagnetic interference in welding scenarios by using optical fiber transmission instead of traditional cables, thereby improving the anti-interference capability of data transmission. The sensor sampling frequency is dynamically adapted to the welding process, balancing accuracy and energy consumption.

[0015] (2) The present invention optimizes the spatter amount in reverse by automatically adapting the specific operating parameters of the welding robot to reduce the spatter amount, thereby reducing the spatter risk in the welding process.

[0016] (3) The network layer of the present invention adopts TSN+ industrial Ethernet through the backbone network to allocate the highest priority time slot for the motion command of the welding robot, ensuring short-time deterministic transmission and solving the command delay problem of traditional Ethernet when multiple robots are in concurrency. The mobile device adopts 5G-URLLC to avoid the collision risk when the mobile device and the robotic arm work together. The edge gateway has a built-in welding protocol parsing module to connect to the mainstream welding robot protocol and shorten the protocol conversion delay. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the system structure connection of the present invention.

[0019] Figure 2 This is a schematic diagram of the implementation steps of the method of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 As shown, the present invention provides an industrial control system based on the Industrial Internet, comprising a perception layer, a network layer, a platform layer and an application layer that are connected in sequence, and the layers work together to achieve intelligent control of the automotive welding process.

[0022] The sensing layer includes a high-frequency current sensor integrated at the end effector of the welding robot, an infrared thermal imager deployed around the welding torch, a six-axis force sensor at the joints of the robotic arm, a laser displacement sensor deployed around the workpiece, and workshop environment sensors. These sensors are used to capture arc stability, record the temperature field distribution of the molten pool, monitor welding pressure fluctuations, monitor workpiece deformation data, and collect workshop dust concentration and humidity. The sensing layer uses fiber optic data transmission and can dynamically adapt the sensor sampling frequency according to the welding process.

[0023] In a specific embodiment of the present invention, the specific implementation process of the sensing layer includes a sensor collaborative acquisition step: the high-frequency current sensor, infrared thermal imager, six-axis force sensor, laser displacement sensor, and workshop environment sensor start acquisition according to a preset time synchronization mechanism. Among them, the high-frequency current sensor acquires the current signal of the welding arc in real time, the infrared thermal imager synchronously captures the temperature field image of the molten pool in the corresponding arc action area, the six-axis force sensor records the pressure data of the welding torch in contact with the workpiece, the laser displacement sensor acquires the workpiece deformation data, and the workshop environment sensor acquires the environmental parameters within the preset welding station perimeter.

[0024] Data collected by each sensor is transmitted to a local data aggregation node via optical fiber. The data aggregation node performs timestamp alignment processing on the received heterogeneous data, and associates current signals, temperature field images, pressure data, workpiece deformation data and environmental parameters with the same welding moment to form a multi-dimensional data set of the welding process.

[0025] In a specific embodiment of the present invention, the specific implementation process of dynamically adapting the sensor sampling frequency according to the welding process in the perception layer includes: a preset welding process type identification module, which identifies whether the current welding process is spot welding or arc welding by receiving the working instructions of the welding robot.

[0026] When the process is identified as spot welding, a sampling frequency increase command is sent to the high-frequency current sensor and the six-axis force sensor. The increase command specifically increases the sampling frequency from the reference value to the high-frequency value. For example, the sampling frequency of the high-frequency current sensor is increased from the reference value of 1MHz to 2MHz, and the sampling frequency of the six-axis force sensor is simultaneously increased to the corresponding high frequency to capture the sudden change in current and pressure peak at the moment of spot welding.

[0027] When the process is identified as arc welding, a sampling frequency reduction command is sent to each sensor. For example, the high-frequency current sensor is reduced to the reference sampling frequency of 500kHz, and the sampling frequency of the six-axis force sensor is reduced synchronously to balance the amount of data collected and the energy consumption.

[0028] During the switching of welding processes, the sensing layer adjusts the sampling frequency through a smooth transition algorithm to avoid data loss caused by sudden frequency changes. For example, the smooth transition algorithm is implemented by gradually adjusting the sampling frequency in 10 steps within 50ms.

[0029] The platform layer includes a welding quality digital twin engine, a process parameter knowledge base, and an energy consumption optimization model. The welding quality digital twin engine reconstructs a 3D model of the welding area based on 3D point clouds and maps the molten pool morphology and heat-affected zone range in real time. The process parameter knowledge base integrates historical welding cases and generates an optimal parameter recommendation model. The energy consumption optimization model correlates welding-related parameters to calculate a benchmark value for energy consumption per unit weld point.

[0030] In a specific embodiment of the present invention, the specific implementation process of the welding quality digital twin engine in the platform layer includes: 3D model construction stage: 3D scanning of the welding workpiece and welding robot is performed by a laser scanner deployed in the welding workshop to obtain point cloud data.

[0031] An initial 3D model of the physical entity is constructed based on point cloud data. The model includes the geometric dimensions of the workpiece, the structural parameters of the welding robot, and the layout information of the welding station.

[0032] Real-time mapping core function: Acquire real-time data of the welding process through the sensing layer, including molten pool temperature field data collected by infrared thermal imager, arc data collected by high-frequency current sensor, and workpiece deformation data collected by laser displacement sensor.

[0033] Real-time data is input into the digital twin engine, which calculates the molten pool morphology changes through thermodynamic simulation algorithms, calculates the range of the heat-affected zone based on the finite element analysis model, and updates the calculation results to the three-dimensional model in real time, so that the molten pool position, temperature distribution, and workpiece deformation state of the virtual model are synchronized with the physical world.

[0034] Extended functionality: Based on the molten pool morphology data generated by the real-time mapping core function, the motion trajectory and distribution density of splash particles are derived by combining the splash formation mechanism model, providing a data foundation for the simulation verification of splash quantity.

[0035] Deviation correction mechanism: The actual dimensional deviation of the welded workpiece is obtained periodically through manual inspection or high-precision measuring equipment, and the deviation data is fed back to the digital twin engine. The engine adjusts the simulation parameters through iterative optimization algorithms.

[0036] It should be noted that the specific implementation process of the thermodynamic simulation algorithm in the welding quality digital twin engine includes: constructing a thermodynamic model of the molten pool, which includes the heat conduction equation, convection equation and energy conservation equation of the molten pool, taking into account factors such as electric arc heating, workpiece heat dissipation and molten pool flow.

[0037] The molten pool temperature data acquired by the infrared thermal imager is used as the boundary condition of the model and input into the model's temperature field calculation module.

[0038] The heat conduction equation was solved using the finite volume method to calculate the temperature distribution inside the molten pool. The time step was set to 1 ms and the spatial mesh generation accuracy was set to 0.1 mm.

[0039] Based on the temperature distribution, the solidification rate and morphological changes of the molten pool are calculated. Combined with the material's melting point, specific heat capacity and other physical parameters, and linked to the workpiece deformation state, the size of the molten pool, including its length, width, depth and the shape of the weld after solidification, is predicted.

[0040] The predicted molten pool morphology is compared with the actual molten pool morphology obtained by laser scanning. Deviation values ​​such as length deviation and depth deviation are calculated. Parameters such as thermal conductivity coefficient and convection coefficient in the model are adjusted through feedback iteration to reduce prediction deviation and ensure that the digital twin engine accurately maps the molten pool morphology.

[0041] For example, in a real-world vehicle body side panel welding scenario, the application process is as follows:

[0042] In a car factory, traditional methods for welding aluminum alloy components for the side panels of car bodies often fail to monitor the molten pool in real time, frequently resulting in component deformation due to an excessively large heat-affected zone. After applying this engine, a laser scanner is first used to perform a 3D scan of the side panel components and the welding robot to acquire point cloud data and construct an initial 3D model. The model clearly shows the curved surface structure of the side panel, the location of the weld points, and the robot's motion trajectory range.

[0043] During welding, an infrared thermal imager collects molten pool temperature field data every 1 ms, while a high-frequency current sensor synchronously records changes in arc current. This data is transmitted to the digital twin engine in real time. The engine's thermodynamic simulation algorithm calculates the temperature distribution based on the molten pool's heat conduction equation. Combining this with the aluminum alloy's melting point of 660℃ and specific heat capacity of 900 J / (kg·K), the virtual model maps in real time the molten pool's length of approximately 8 mm, width of approximately 5 mm, and depth of approximately 2 mm, as well as the heat-affected zone's diameter of approximately 15 mm.

[0044] When the welding current suddenly increases from 200A to 220A, the engine predicts that the molten pool will expand to a length of 10mm and a width of 7mm within 3 seconds, and the diameter of the heat-affected zone will increase to 18mm, exceeding the upper limit allowed by the process. If the diameter of the heat-affected zone is ≤16mm, the engine, through extended functions, deduces, based on the molten pool morphology data and the spatter formation mechanism model, that the trajectory of the spatter particles will cover the sealing groove area of ​​the side wall. If spatter is not allowed in this area, the digital twin engine immediately feeds this prediction result back to the application layer. The spatter closed-loop control module of the application layer initiates parameter adjustment, pulling the welding current back to 205A. After adjustment, the engine updates the virtual model in real time, showing that the molten pool and heat-affected zone have returned to the normal range, and the spatter trajectory avoids the sealing groove, preventing component deformation and quality defects. The spatter formation mechanism model specifically states that the amount of spatter is positively correlated with the temperature of the molten pool; for example, for every 50°C increase in temperature, the amount of spatter increases by approximately 20%.

[0045] In a specific embodiment of the present invention, the specific implementation process of the process parameter knowledge base in the platform layer includes: historical welding case acquisition module: importing welding case data from the enterprise MES system and quality traceability system through the interface, wherein the welding case data includes the material of the welded workpiece, plate thickness, welding current, welding voltage, welding speed and corresponding welding quality inspection results.

[0046] The welding quality inspection results include whether there are defects such as porosity, cracks, or lack of fusion.

[0047] The materials include galvanized steel sheets, aluminum alloys, etc.

[0048] Case data preprocessing module: Cleans the imported case data, removes outliers, fills in missing values, and performs standardization.

[0049] For example, the process of removing outliers, such as invalid data where the current exceeds the device's range, filling in missing values, such as filling in missing fields like plate thickness using the average of adjacent cases, and standardization processing, such as converting parameters in different units into dimensionless data.

[0050] A process parameter knowledge base is constructed based on the preprocessed case data, and the knowledge base adopts a distributed storage architecture.

[0051] The optimal parameter recommendation model training process is as follows: The random forest algorithm is used as the basic model, and the material, plate thickness, and spatter amount per unit area are used as input features, welding current, voltage, and speed are used as parameters to be optimized, and welding quality detection results are used as target labels.

[0052] The model was trained using a 5-fold cross-validation method, and the model parameter combination with the highest prediction accuracy was selected as the optimal model.

[0053] Model update mechanism: Every time a preset number of new welding cases are accumulated, the incremental training process of the model is triggered. The new case data is input into the existing model for parameter fine-tuning, so that the model can adapt to the new welding process scenario. The incremental training process does not interrupt the normal query service of the knowledge base.

[0054] It should be noted that the core of the optimal parameter recommendation model is to establish a mapping relationship between "input features → optimal parameters". This relationship can be represented by a combination of mathematical expressions and logical rules. The following is a specific explanation using a welding scenario:

[0055] 1. Quantization of input features: Let the model input feature vector be X, containing the following quantifiable parameters: X1: Welding material, quantized through encoding, such as galvanized steel plate = 1, ordinary steel plate = 2, aluminum alloy = 3.

[0056] X2: Plate thickness, unit: mm, such as 1.5mm is represented as 1.5.

[0057] X3: Current splash amount per unit area, unit: splashes / square centimeter, such as 3 splashes / square centimeter is represented as 3.

[0058] X4: Quality defect label, binary quantification, no defects = 0, defects such as incomplete fusion or porosity = 1.

[0059] The input feature vector is X = [X1, X2, X3, X4].

[0060] 2. The model output consists of two core welding parameters: optimal welding current I (unit: A) and optimal welding speed V (unit: m / min).

[0061] 3. Based on historical data fitting, the relationship between the optimal parameters and input features can be expressed as: Optimal welding current I = I0 - k1 * f(X1,X2) * X3 - k2 * X4 * X2. Where I0 is the current welding current, k1 is the spatter influence coefficient, f(X1,X2) is the material-plate thickness correction function, and k2 is the defect influence coefficient, such as a fixed value of 5.0, which is triggered when a defect exists.

[0062] Physical meaning: The greater the amount of splashing, the greater the current reduction. The presence of defects and the thickness of the board result in a smaller decrease in current.

[0063] Optimal welding speed: V = V0 * [1 + 0.1 * g(X1) * min(1, X3 / T)], where V0 is the current welding speed, g(X1) is the material speed correction coefficient, for example, 1.0 for galvanized steel plate and 0.9 for ordinary steel plate, T is the spatter threshold, such as T = 3 spatters per square centimeter for galvanized steel plate, min(1, X3 / T) ensures that the speed increase does not exceed 10%, and takes 1 when X3 > T.

[0064] Physical meaning: When the spatter amount exceeds the threshold, the speed is increased proportionally, up to 10%, and the increase amplitude is different for different materials.

[0065] The above formula also needs to satisfy logical constraints, specifically the current safety constraint, the speed upper limit constraint, and the defect priority rule.

[0066] That is, 1. Current safety constraint: if I < I_min(X1, X2) THEN I = I_min(X1, X2), (I_min is the minimum current corresponding to the material - plate thickness, such as for 1.5mm galvanized steel plate, I_min = 200A).

[0067] 2. Speed upper limit constraint: if V > V_max(X1) THEN V = V_max(X1), (V_max is the maximum speed corresponding to the material, such as for aluminum alloy, V_max = 1.5m / min).

[0068] 3. Defect priority rule: if X4 = 1 THEN ignore the influence of spatter amount, and give priority to executing I = I0 + 5, slightly increasing the current to ensure fusion.

[0069] The coefficients in the above formula are not set manually, but are the optimal solutions obtained by training historical data through the random forest algorithm.

[0070] It should be noted that the process of updating and maintaining the case data in the process parameter knowledge base includes: setting case access conditions: new welding cases need to include complete process parameters, such as material, plate thickness, current, voltage, speed, quality inspection report, and production environment parameters. Only cases that meet the conditions can enter the update process.

[0071] Case classification and storage: Classify new cases according to body parts and welding process types, and store them in the corresponding database partitions to improve the case query efficiency.

[0072] Case validity verification: Verify the validity of the cases in the knowledge base every quarter, eliminate invalid cases caused by equipment aging and process iteration, such as cases of old equipment 5 years ago, and retain valid cases to ensure the timeliness of the knowledge base.

[0073] Knowledge graph construction: Build a welding process knowledge graph based on case data. The nodes of the graph include process parameters, quality results, environmental factors, etc. The edges between nodes represent the correlation strength between parameters and results. The knowledge graph supports visual query, providing intuitive parameter optimization reference for process personnel. The said correlation strength is such as the positive correlation between high current and spatter amount.

[0074] In a specific embodiment of the present invention, the specific implementation process of the energy consumption optimization model in the platform layer includes: energy consumption benchmark value calculation module: collecting historical welding production data, including welding current, welding time, robot standby time, workshop ambient temperature and corresponding actual energy consumption data for each weld point.

[0075] The correlation between various parameters and energy consumption was analyzed using a multiple linear regression algorithm to identify the key parameters affecting energy consumption.

[0076] A calculation model for the benchmark value of energy consumption per unit weld joint is constructed based on key parameters. The model expression is: E=a×I+b×t+c×T+d, where E is the benchmark value of energy consumption per unit weld joint, I is the welding current, t is the welding time, T is the ambient temperature of the workshop, and a, b, c, and d are coefficients obtained by fitting historical data.

[0077] Real-time energy consumption optimization stage: Real-time acquisition of current welding-related parameters, including welding process current, time, and ambient temperature data, and input into the energy consumption benchmark value calculation model to obtain real-time benchmark energy consumption.

[0078] The actual energy consumption is compared with the benchmark energy consumption. When the actual energy consumption exceeds the preset ratio of the benchmark energy consumption, energy consumption optimization suggestions are generated. These suggestions include adjusting the welding current and shortening the robot's standby time.

[0079] After the optimization suggestions are reviewed by the platform layer, they are sent to the application layer, where the welding parameters are adjusted to form a closed loop for energy consumption optimization.

[0080] For example, the vehicle floor has 200 weld points made of ordinary steel plate with a thickness of 2.0 mm. In traditional welding, the robot has a long standby time and large energy consumption fluctuations. The energy consumption optimization model collected historical energy consumption data and obtained the baseline value calculation formula through multiple linear regression fitting: E=0.002×I+0.5×t+0.01×T+0.3.

[0081] In actual production, at 10:00 AM, during peak electricity consumption, with an electricity price of 0.8 yuan / kWh, the real-time data for welding were: current 240A, time 25ms, and ambient temperature 28℃. The model calculated the baseline energy consumption as E = 0.002×240 + 0.5×0.025 + 0.01×28 + 0.3 = 0.48 + 0.0125 + 0.28 + 0.3 = 1.0725 kWh / point. The total baseline energy consumption for 200 welding points was 214.5 kWh.

[0082] At this point, the system detected an actual energy consumption of 236 kWh, exceeding the baseline by 10%. Analysis revealed: ① The robot idled for 5 seconds after completing each welding point. ② The current of 240A was higher than the historical best value. The energy consumption optimization model suggested reducing the current to 220A and compressing the idle time to 2 seconds.

[0083] After the application layer performs the adjustment, the energy consumption under the new parameters is: E=0.002×220+0.5×0.025+0.01×28+0.3=0.44+0.0125+0.28+0.3=1.0325kWh / point, the total energy consumption is reduced to 206.5kWh, saving 10.5kWh. Calculated based on peak electricity price, the savings per vehicle is 8.4 yuan, and the savings for producing 500 vehicles per day is 4200 yuan.

[0084] At 18:00 in the evening, during a low electricity consumption period with an electricity price of 0.4 yuan / kWh, the model allows the current to be appropriately increased to 230A to speed up the welding process. At this time, energy consumption increases but the total production cost decreases, which reflects the model's adaptability to dynamic scenarios and is in line with the logic of "optimizing peak and valley electricity prices of the grid".

[0085] The present invention optimizes the spatter amount by automatically adapting to the specific operating parameters of the welding robot to reduce the amount of spatter, thereby reducing the risk of spatter during the welding process.

[0086] The application layer includes a spatter closed-loop control module and a multi-robot collaborative scheduling module. When the spatter amount exceeds a threshold, the spatter closed-loop control module uses the platform-layer optimal parameter recommendation model to deduce the optimal welding parameters and sends them to the robots. The multi-robot collaborative scheduling module simulates path conflicts using digital twins and dynamically optimizes the welding sequence.

[0087] In a specific embodiment of the present invention, the specific implementation process of the spatter closed-loop control module in the application layer includes: presetting a spatter threshold database, wherein the database presets different spatter thresholds according to the welding material, plate thickness and welding position. For example, the spatter threshold for welding galvanized steel plate is set to 3 spatters / cm. 2 The threshold for spatter during welding of ordinary steel plates is set at 5 spatters / cm. 2 .

[0088] An infrared thermal imager acquires image data of the welding area in real time and transmits it to the image recognition module at the application layer. The image recognition module is used to identify spatter particles in the image and count the number of spatter particles per unit area.

[0089] The statistical spatter amount is compared with the corresponding preset threshold. If the spatter amount does not exceed the threshold, the current welding parameters are kept unchanged.

[0090] When the spatter amount exceeds the threshold, the parameter back-reasoning process is initiated: the current spatter amount, welding material, and plate thickness data are input into the process parameter knowledge base of the platform layer, the optimal parameter recommendation model is called, and the optimal welding current and welding speed that can reduce the spatter amount are back-reasoned.

[0091] The optimal parameters obtained by reverse engineering are simulated and verified by the extended function of the digital twin engine. After the verification is successful, control commands are generated. The simulation verification is specifically to confirm whether the splash amount after simulating the parameters has dropped below the threshold.

[0092] Control commands are sent to the control system of the welding robot through the TSN network. The robot completes parameter adjustments, while the infrared thermal imager continuously monitors the changes in the amount of spatter. If the amount of spatter still exceeds the threshold after adjustment, the parameter reverse calculation process is repeated until the amount of spatter meets the standard.

[0093] In a specific embodiment of the present invention, the specific implementation process of the multi-robot collaborative scheduling module in the application layer includes: welding task decomposition stage: receiving the body welding task issued by the production planning system, wherein the task includes the number of weld points to be welded, the position coordinates and the welding sequence requirements.

[0094] The vehicle body welding task is broken down into several sub-tasks for robots. Each sub-task includes a list of welding points to be welded by the corresponding robot and the estimated welding time.

[0095] Path conflict simulation phase: The sub-tasks of each robot are input into the robot digital twin engine. The engine simulates the welding path of each robot based on the robot's kinematic model and identifies potential conflict points in the path. For example, a potential conflict point is when two robots arrive at the same spatial location at the same time.

[0096] Conflict resolution algorithm: For the identified conflict points, a priority scheduling strategy is adopted to optimize the welding sequence. The strategy includes: prioritizing the processing of welding points located in the critical structure of the vehicle body and prioritizing the scheduling of robots with lower load rates. For example, welding points in the critical structure of the vehicle body are such as the welding points connecting the columns and beams.

[0097] The optimized welding sequence generates scheduling instructions, which are then sent to each robot through the network layer. The robots then execute the welding operations according to the instructions.

[0098] Real-time monitoring phase: The actual position of the robot is tracked in real time by the position sensor of the perception layer. The actual position is compared with the simulated position in the digital twin engine. When the deviation exceeds the preset deviation, the path conflict simulation and optimization process is restarted, and the scheduling instructions are dynamically adjusted to ensure the smoothness of multi-robot collaborative welding.

[0099] It should be noted that the multi-robot collaborative scheduling module also includes an emergency response implementation process, specifically: a preset emergency scenario library, which includes emergency situations such as robot malfunction, workpiece positioning deviation, sudden abnormal welding quality, and abnormal workshop environment, as well as corresponding handling plans.

[0100] The robot's operating status, such as motor temperature, joint speed, workpiece positioning accuracy, and welding quality data, is monitored in real time by six-axis force sensors at the joints of the robotic arm in the perception layer and additional sensors.

[0101] When a robot malfunction is detected, such as motor overload, the subtask reassignment plan for the malfunctioning robot is immediately activated: the welding task of the malfunctioning robot is decomposed to other idle or lightly loaded robots, and the path is re-simulated through the digital twin engine to ensure that the welding cycle after reassignment does not exceed the safety margin, which is specifically the buffer amount of the welding cycle, such as 10 seconds.

[0102] When the workpiece positioning deviation is detected to exceed the workpiece positioning deviation threshold, the deviation compensation plan is activated: the deviation amount is calculated through the digital twin engine, the welding gun position coordinates of the relevant robot are adjusted, the positioning deviation is compensated, and welding misalignment caused by workpiece offset is avoided. The workpiece positioning deviation threshold is specifically defined by the production workshop personnel according to production needs, for example, 1mm.

[0103] When a sudden abnormality in welding quality is detected, the quality emergency plan is activated: all robot welding operations are suspended, the process parameter knowledge base of the platform layer is called to analyze the cause of the abnormality, generate temporary adjustment parameters, which are then sent to the robots after manual confirmation. After welding is resumed, the quality changes are continuously tracked until the abnormality is eliminated. The process parameter knowledge base of the platform layer stores an integrated mapping table between the cause of the abnormality and the welding quality detection results, as well as a mapping table between the cause of the abnormality and the adjustment parameters.

[0104] When abnormal fluctuations in welding pressure are detected, the current welding task should be immediately suspended to prevent weld defects from expanding due to unstable pressure, and an alarm and recording mechanism should be activated.

[0105] For example, when an abnormal dust concentration is detected in the workshop.

[0106] When the dust concentration is less than the first-level threshold: the environmental sensor triggers an early warning, the multi-robot collaborative scheduling module at the application layer adjusts the welding sequence, prioritizes completing the welding at the current workstation, then pauses and starts the workshop ventilation system.

[0107] When the dust concentration is greater than the first-level threshold but less than the second-level threshold: the platform layer issues a "reduce power production" instruction to the application layer, all robot welding current is reduced by a set ratio, and the AGV feeding vehicle suspends material transportation to avoid airflow disturbance causing dust diffusion.

[0108] When the dust concentration exceeds the secondary threshold but is less than the tertiary threshold: the application layer initiates an emergency shutdown procedure, all robots immediately stop welding, and the fire suppression system is activated to suppress dust until the concentration drops to 5 mg / m³. 3 The following steps are required to restart.

[0109] For example, a quality alert is issued when abnormal humidity is detected in the workshop:

[0110] When the humidity falls below the humidity threshold, data collected by environmental sensors triggers a quality warning at the platform layer. At this time, the process parameter knowledge base pushes an "anti-porosity parameter package" to the application layer—suggesting that the welding current be increased by a certain percentage and the gas protection duration be extended to reduce porosity defects by increasing the molten pool temperature and the protection effect.

[0111] In a specific embodiment of the present invention, the network layer uses a combination of TSN and industrial Ethernet as the backbone network to allocate the highest priority time slots for the motion commands of the welding robot to ensure deterministic transmission. Mobile devices, such as AGV feeding vehicles, use 5G-URLLC for communication. The edge gateway of the network layer has a built-in welding protocol parsing module for interfacing with mainstream welding robot protocols, such as FANUC ArcLink and KUKA KRL, to compress protocol conversion latency.

[0112] The specific implementation process for interfacing with mainstream welding robot protocols includes:

[0113] The edge gateway has a built-in protocol parsing library, which contains parsing rules for mainstream welding robot protocols.

[0114] When a welding robot accesses the network layer, the edge gateway sends a protocol probe frame and identifies the protocol type used by the welding robot based on the response message returned by the robot.

[0115] Based on the identified protocol type, the corresponding parsing rules in the protocol parsing library are invoked to convert the robot's private protocol data into OPC UA unified protocol data. The conversion process includes data format mapping: mapping the robot's custom data format to standard structured data; unit conversion: converting the robot's local units to international standard units; and data cleaning: removing invalid data fields.

[0116] The converted unified protocol data is transmitted to the platform layer via the TSN network. At the same time, the control commands issued by the platform layer are reverse-converted according to the robot's private protocol format to ensure that the commands are accurately recognized and executed by the robot.

[0117] This invention addresses the strong electromagnetic interference in welding scenarios by employing fiber optic transmission instead of traditional cables, thus improving data transmission anti-interference capabilities. The sensor sampling frequency dynamically adapts to the welding process, balancing accuracy and energy consumption.

[0118] The network layer of this invention adopts TSN+ Industrial Ethernet in the backbone network to allocate the highest priority time slots for the motion commands of the welding robot, ensuring short-time deterministic transmission and solving the command delay problem of traditional Ethernet when multiple robots are running concurrently. The mobile device adopts 5G-URLLC to avoid the collision risk when the mobile device and the robotic arm are working together. The edge gateway has a built-in welding protocol parsing module to connect with mainstream welding robot protocols and shorten the protocol conversion latency.

[0119] It should be noted that the collaborative implementation process between the platform layer and the application layer includes: the platform layer receiving welding process data uploaded by the perception layer in real time, performing data processing and model calculation, and generating three types of outputs: welding quality assessment results, process parameter optimization suggestions, and anomaly warning information.

[0120] The welding quality assessment results include qualified / unqualified. For example, the presence of defects such as porosity, cracks, and lack of fusion indicates unqualified welding quality, while the absence of defects indicates qualified welding quality.

[0121] The application layer sets up a data receiving interface, for example, to obtain the above output data from the platform layer at a period of 100ms.

[0122] When the welding quality assessment result is deemed satisfactory, the application layer controls the robot to continue performing the next welding task.

[0123] When a suggestion for optimizing process parameters is received, the parameter adjustment module in the application layer verifies the feasibility of the suggestion. If the verification is successful, the parameter adjustment is executed, and the adjustment result is fed back to the platform layer. Specifically, the feasibility verification checks whether the parameters are within the robot's operating range.

[0124] Upon receiving an anomaly warning, the application layer's anomaly handling module activates the response mechanism: for equipment fault warnings, the welding operation of the relevant robots is immediately suspended, and maintenance personnel are notified to conduct repairs. For quality exceedance warnings, the rework process is initiated, instructing the robots to repair the defective welds.

[0125] The application layer feeds back the execution results to the platform layer in real time. The platform layer stores the feedback results in the process parameter knowledge base as sample data for subsequent model training, forming a closed-loop collaboration between the platform layer and the application layer. The execution results include task completion status, parameter adjustment records, and anomaly handling results.

[0126] It should be noted that the specific implementation process of the data aggregation node in the perception layer for preprocessing heterogeneous data includes: noise filtering: using wavelet transform algorithm to denoise the current signal collected by the high-frequency current sensor, eliminating high-frequency noise introduced by electromagnetic interference. Gaussian filtering is applied to the temperature field image collected by the infrared thermal imager to smooth out salt-and-pepper noise in the image.

[0127] Data compression: The temperature field image is compressed using the JPEG2000 compression algorithm with a compression ratio of 10:1 to reduce data transmission while maintaining image quality. Differential coding compression is used for current signals and pressure data to preserve data trends while reducing storage capacity.

[0128] Outlier detection: Based on the 3σ criterion, outliers in each data set are identified, such as values ​​in current signals that exceed the mean ± 3 times the standard deviation, and values ​​in pressure data that show a sudden change of more than 20%.

[0129] For detected outliers, linear interpolation is used for repair, which involves calculating the interpolation value using the normal data adjacent to the outlier and replacing the outlier.

[0130] The preprocessed data is classified and stored according to welding station number and timestamp to form a standardized welding process dataset, providing high-quality input data for model calculations at the platform layer.

[0131] Reference Figure 2 As shown, the second aspect of the present invention provides a method for executing the industrial control system based on the Industrial Internet of Things described in the present invention, comprising: an initialization phase: after the system is started, the sensors in the perception layer complete self-testing and calibration, the network layer establishes a communication connection, the platform layer loads a digital twin model and a process parameter knowledge base, and the application layer receives the initial welding task.

[0132] Welding execution phase: The perception layer collects welding process data in real time and transmits it to the platform layer via the network layer. The platform layer processes the data, generates quality assessments, parameter suggestions, and early warning information through model calculations, and transmits them to the application layer. The application layer executes welding control, parameter adjustments, or anomaly handling based on the platform layer's output and feeds back the execution results to the platform layer.

[0133] Closed-loop optimization phase: The platform layer updates the digital twin model and process parameter knowledge base based on feedback from the application layer, optimizing the model algorithm and parameters. The perception layer adjusts the sensor acquisition strategy according to the optimization instructions from the platform layer, for example, by focusing on monitoring workstations where quality problems frequently occur.

[0134] End Phase: After the welding task is completed, the system generates a welding process report, which includes quality data, energy consumption data and robot operation data for each weld point. The report is stored in the historical database at the platform level to provide data support for subsequent production analysis. At the same time, the system enters a low-power standby state, waiting for the next round of welding tasks to start.

[0135] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications or additions should fall within the protection scope of the present invention.

Claims

1. An industrial control system based on the Industrial Internet, characterized in that, It includes a perception layer, a network layer, a platform layer, and an application layer that are connected in sequence, and each layer works together to achieve intelligent control of the automotive welding process; The sensing layer includes a high-frequency current sensor integrated at the end of the welding robot, an infrared thermal imager deployed around the welding torch, a six-axis force sensor at the joints of the robotic arm, a laser displacement sensor deployed around the workpiece, and a workshop environment sensor. These sensors are used to capture arc stability, record the temperature field distribution of the molten pool, monitor welding pressure fluctuations, monitor workpiece deformation data, and collect workshop dust concentration and humidity. The sensing layer uses optical fiber to transmit data and can dynamically adapt the sensor sampling frequency according to the welding process. The specific implementation process of dynamically adapting the sensor sampling frequency according to the welding process in the sensing layer includes: A preset welding process type identification module, which identifies whether the current welding process is spot welding or arc welding by receiving the working instructions of the welding robot; When the process is identified as spot welding, a sampling frequency increase command is sent to the high-frequency current sensor and the six-axis force sensor. Specifically, the sampling frequency is increased from the reference value to the high-frequency value, and the sampling frequency of the six-axis force sensor is simultaneously increased to the corresponding high frequency to capture the sudden change in current and pressure peak at the moment of spot welding. When the process is identified as arc welding, a sampling frequency reduction command is sent to each sensor, and the sampling frequency of the six-axis force sensor is reduced synchronously to balance the amount of data collected and the energy consumption. During the switching of welding processes, the sensing layer adjusts the sampling frequency through a smooth transition algorithm to avoid data loss caused by sudden frequency changes; The platform layer includes a welding quality digital twin engine, a process parameter knowledge base, and an energy consumption optimization model. The welding quality digital twin engine reconstructs a three-dimensional model of the welding area based on 3D point clouds and maps the molten pool morphology and heat-affected zone range in real time. The process parameter knowledge base integrates historical welding cases and generates an optimal parameter recommendation model. The energy consumption optimization model associates welding-related parameters to calculate the benchmark value of energy consumption per unit weld point. The application layer includes a spatter closed-loop control module and a multi-robot collaborative scheduling module. When the spatter closed-loop control module detects that the amount of spatter exceeds the threshold, it back-calculates the optimal welding parameters through the optimal parameter recommendation model of the platform layer and sends them to the robot. The multi-robot collaborative scheduling module simulates path conflicts through digital twins and dynamically optimizes the welding sequence.

2. The industrial control system based on the Industrial Internet according to claim 1, characterized in that, The network layer uses a combination of TSN and industrial Ethernet as the backbone network, allocating the highest priority time slots to the motion commands of the welding robot to ensure deterministic transmission. The mobile device uses 5G-URLLC for communication, and the edge gateway of the network layer has a built-in welding protocol parsing module for interfacing with mainstream welding robot protocols. The specific implementation process for interfacing with mainstream welding robot protocols includes: The edge gateway has a built-in protocol parsing library, which contains parsing rules for mainstream welding robot protocols; When the welding robot accesses the network layer, the edge gateway sends a protocol probe frame to identify the protocol type used by the welding robot based on the response message returned by the welding robot. Based on the identified protocol type, the corresponding parsing rules in the protocol parsing library are invoked to convert the robot's private protocol data into OPC UA unified protocol data. The conversion process includes data format mapping, unit conversion, and data cleaning. The converted unified protocol data is transmitted to the platform layer via the TSN network. At the same time, the control commands issued by the platform layer are reverse-converted according to the robot's private protocol format to ensure that the commands are accurately recognized and executed by the robot.

3. An industrial control system based on the Industrial Internet according to claim 1, characterized in that, The specific implementation process of the perception layer includes a sensor collaborative acquisition step: The high-frequency current sensor, infrared thermal imager, six-axis force sensor, laser displacement sensor, and workshop environment sensor are activated to collect data according to a preset time synchronization mechanism. The high-frequency current sensor collects the current signal of the welding arc in real time, the infrared thermal imager captures the temperature field image of the molten pool in the corresponding arc action area, the six-axis force sensor records the pressure data of the welding torch in contact with the workpiece, the laser displacement sensor collects the workpiece deformation data, and the workshop environment sensor collects the environmental parameters within the preset welding station's surrounding area. Data collected by each sensor is transmitted to a local data aggregation node via optical fiber. The data aggregation node performs timestamp alignment processing on the received heterogeneous data, and associates current signals, temperature field images, pressure data, workpiece deformation data and environmental parameters with the same welding moment to form a multi-dimensional data set of the welding process.

4. An industrial control system based on the Industrial Internet according to claim 1, characterized in that, The specific implementation process of the welding quality digital twin engine in the platform layer includes: 3D model construction stage: The welding workpiece and welding robot are 3D scanned by a laser scanner deployed in the welding workshop to obtain point cloud data; An initial 3D model of the physical entity is constructed based on point cloud data. The model includes the geometric dimensions of the workpiece, the structural parameters of the welding robot, and the layout information of the welding station. Real-time mapping core function: to acquire real-time data of the welding process through the sensing layer, including molten pool temperature field data collected by infrared thermal imager, arc data collected by high-frequency current sensor and workpiece deformation data collected by laser displacement sensor. Real-time data is input into the digital twin engine, which calculates the molten pool morphology changes through thermodynamic simulation algorithms, calculates the range of the heat-affected zone based on the finite element analysis model, and updates the calculation results to the three-dimensional model in real time, so that the molten pool position, temperature distribution, and workpiece deformation state of the virtual model are synchronized with the physical world. Extended functionality: Based on the molten pool morphology data generated by the real-time mapping core function, the motion trajectory and distribution density of splash particles are derived by combining the splash formation mechanism model, providing a data foundation for the simulation verification of splash quantity; Deviation correction mechanism: The actual dimensional deviation of the welded workpiece is obtained periodically through manual inspection or high-precision measuring equipment, and the deviation data is fed back to the digital twin engine. The engine adjusts the simulation parameters through iterative optimization algorithms.

5. An industrial control system based on the Industrial Internet according to claim 1, characterized in that, The specific implementation process of the process parameter knowledge base in the platform layer includes: Historical welding case acquisition module: Imports welding case data from the enterprise MES system and quality traceability system through the interface. The welding case data includes the material of the welded workpiece, plate thickness, welding current, welding voltage, welding speed and corresponding welding quality inspection results. Case data preprocessing module: Cleans the imported case data, removes outliers, fills in missing values, and performs standardization. A process parameter knowledge base is constructed based on the preprocessed case data, and the knowledge base adopts a distributed storage architecture. The optimal parameter recommendation model training process is as follows: material, plate thickness, and spatter per unit area are used as input features, welding current, voltage, and speed are used as parameters to be optimized, and welding quality inspection results are used as target labels. The model was trained using a 5-fold cross-validation method, and the model parameter combination with the highest prediction accuracy was selected as the optimal model. Model update mechanism: Every time a preset number of new welding cases are accumulated, the incremental training process of the model is triggered. The new case data is input into the existing model for parameter fine-tuning, so that the model can adapt to the new welding process scenario. The incremental training process does not interrupt the normal query service of the knowledge base.

6. An industrial control system based on the Industrial Internet according to claim 1, characterized in that, The specific implementation process of the energy consumption optimization model in the platform layer includes: Energy consumption baseline calculation module: collects historical welding production data, including welding current, welding time, robot standby time, workshop ambient temperature and corresponding actual energy consumption data for each weld point; The correlation between various parameters and energy consumption was analyzed using a multiple linear regression algorithm to identify the key parameters affecting energy consumption. A calculation model for the benchmark value of energy consumption per unit weld joint is constructed based on key parameters. The model expression is: E=a×I+b×t+c×T+d, where E is the benchmark value of energy consumption per unit weld joint, I is the welding current, t is the welding time, T is the ambient temperature of the workshop, and a, b, c, and d are coefficients obtained by fitting historical data. Real-time energy consumption optimization stage: Real-time collection of current welding-related parameters, including welding process current, time, and ambient temperature data, and input into the energy consumption benchmark value calculation model to obtain real-time benchmark energy consumption; The actual energy consumption is compared with the benchmark energy consumption. When the actual energy consumption exceeds the preset ratio of the benchmark energy consumption, energy consumption optimization suggestions are generated. After the optimization suggestions are reviewed by the platform layer, they are sent to the application layer, where the welding parameters are adjusted to form a closed loop for energy consumption optimization.

7. An industrial control system based on the Industrial Internet according to claim 1, characterized in that, The specific implementation process of the splash closed-loop control module in the application layer includes: A preset spatter threshold database, wherein different spatter thresholds are preset according to welding material, plate thickness and welding position; An infrared thermal imager acquires image data of the welding area in real time and transmits it to the image recognition module of the application layer. The image recognition module is used to identify spatter particles in the image and count the number of spatter particles per unit area. The statistical spatter amount is compared with the corresponding preset threshold. If the spatter amount does not exceed the threshold, the current welding parameters are kept unchanged. When the spatter exceeds the threshold, the parameter back-reasoning process is initiated: the current spatter, welding material, and plate thickness data are input into the process parameter knowledge base of the platform layer, the optimal parameter recommendation model is called, and the optimal welding current and welding speed that can reduce the spatter are back-reasoned. The optimal parameters obtained by reverse engineering are verified by simulation using the extended functions of the digital twin engine. Once the verification is successful, control commands are generated. Control commands are sent to the control system of the welding robot through the TSN network. The robot completes parameter adjustments, while the infrared thermal imager continuously monitors the changes in the amount of spatter. If the amount of spatter still exceeds the threshold after adjustment, the parameter reverse calculation process is repeated until the amount of spatter meets the standard.

8. An industrial control system based on the Industrial Internet according to claim 1, characterized in that, The specific implementation process of the multi-robot collaborative scheduling module in the application layer includes: Welding task decomposition stage: Receive the body welding task issued by the production planning system. The task includes the number of weld points to be welded, their location coordinates, and the welding sequence requirements. The car body welding task is broken down into several sub-tasks for robots. Each sub-task includes a list of welding points that the corresponding robot is responsible for and the estimated welding time. Path conflict simulation phase: The sub-tasks of each robot are input into the robot digital twin engine. The engine simulates the welding path of each robot based on the robot's kinematic model and identifies potential conflict points in the path. Conflict resolution algorithm: For the identified conflict points, a priority scheduling strategy is adopted to optimize the welding sequence. The strategy includes: prioritizing the processing of weld points located in the critical structure of the vehicle body and prioritizing the scheduling of robots with lower load rates. The optimized welding sequence generates scheduling instructions, which are then sent to each robot through the network layer. The robots then execute the welding operations according to the instructions. Real-time monitoring phase: The actual position of the robot is tracked in real time by the position sensor of the perception layer. The actual position is compared with the simulated position in the digital twin engine. When the deviation exceeds the preset deviation, the path conflict simulation and optimization process is restarted, and the scheduling instructions are dynamically adjusted to ensure the smoothness of multi-robot collaborative welding.

9. A method for executing an industrial control system based on the Industrial Internet as described in any one of claims 1-8, characterized in that, include: Initialization phase: After the system starts up, the sensors in the perception layer complete self-testing and calibration, the network layer establishes communication connections, the platform layer loads the digital twin model and process parameter knowledge base, and the application layer receives the initial welding task; Welding execution phase: The perception layer collects welding process data in real time and transmits it to the platform layer via the network layer; the platform layer processes the data, generates quality assessment, parameter suggestions and early warning information through model calculation, and transmits them to the application layer; The application layer performs welding control, parameter adjustment, or anomaly handling based on the platform layer's output, and feeds back the execution results to the platform layer. Closed-loop optimization phase: The platform layer updates the digital twin model and process parameter knowledge base based on the feedback results from the application layer, and optimizes the model algorithm and parameters; the perception layer adjusts the sensor acquisition strategy according to the optimization instructions from the platform layer. End Phase: After the welding task is completed, the system generates a welding process report, which includes quality data, energy consumption data and robot operation data for each weld point. The report is stored in the historical database at the platform level to provide data support for subsequent production analysis. At the same time, the system enters a low-power standby state, waiting for the next round of welding tasks to start.

Images