Multi-source heterogeneous servo welding data acquisition and cooperative control method and system and readable storage medium

Abstract

The invention relates to a multi-source heterogeneous servo welding data acquisition and cooperative control method and system and a readable storage medium. The system comprises a multi-source heterogeneous data acquisition module, an associated storage module, an interactive decision module and an edge cooperative control module. The method comprises the following steps: S1, data acquisition and fusion; s2, data slicing and modeling; s3, collaborative decision and control; and S4, executing feedback. According to the method, nondestructive collection and deep mining of data of the whole welding process are achieved, and the stability of the welding quality and the efficiency of process optimization are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of welding process technology, and in particular relates to a method, system and readable storage medium for multi-source heterogeneous servo welding data acquisition and collaborative control. Background Technology

[0002] In modern manufacturing, such as automotive body shops, resistance spot welding is a primary joining process, and its quality directly affects product structural safety. Servo welding technology, due to its high precision and flexibility, has become a standard process in modern automated production lines, especially in automobile manufacturing. Currently, the mainstream control method involves welding using preset fixed parameters, supplemented by post-weld quality inspection (such as ultrasonic testing) or limited process monitoring (such as monitoring current or pressure limits). Welding quality is highly dependent on the matching and stability of welding parameters (pressure, current, time). During the welding process, the massive amounts of multi-source process data generated by the equipment, mainly including pressure and displacement from the servo system, and feedback information such as transformer secondary current, secondary voltage, and welding time from the welding controller, are the most direct and crucial evidence for evaluating weld quality. However, the inventors discovered that the existing technology has at least the following technical problems: 1. Open-loop control, unable to adapt to disturbances: Preset fixed parameters cannot cope with various disturbances in production, such as electrode cap wear, changes in plate gaps, and power grid voltage fluctuations, leading to fluctuations in solder joint quality. 2. Control lag: Even if abnormal quality trends are detected through the data acquisition system, parameter adjustments require manual intervention by engineers, with response times measured in minutes or hours, which cannot meet the demands of high-speed production. 3. Insufficient coordination: Parameters such as pressure, current, and time are interdependent and jointly affect solder joint quality. Existing systems lack the ability to jointly analyze and coordinately optimize these parameters, often adjusting them independently with limited effectiveness. 4. Lack of closed-loop data value: The large amount of process data collected is only used for monitoring and traceability, failing to be effectively transformed into real-time control commands to form a closed loop of "data-decision-control".

[0003] 5. High data heterogeneity: Servo pressure and displacement are usually high-frequency analog signals (sampling rate ≥ 1kHz), while welding current, voltage, and status signals are mostly low-frequency digital signals obtained through PLC communication (sampling rate usually ≤ 100Hz). The sampling frequency, accuracy, physical meaning, and communication protocols of different types of data vary greatly, forming "data silos".

[0004] 6. Difficulties in Fusion and Synchronization: Traditional data acquisition systems (such as single SCADA systems) lack effective means to accurately synchronize high- and low-frequency data in time. Due to the lack of a unified and precise time scale, analysts find it difficult to determine the pressure value at the current peak or whether the precise moment when the displacement begins to change is synchronized with the pressure, resulting in an inability to accurately reconstruct the dynamic picture of the welding process. Furthermore, existing technologies struggle to effectively fuse multi-source heterogeneous servo welding data (such as servo pressure, displacement, welding secondary current / voltage, time, etc.) from servo welding equipment, which vary significantly in sampling frequency, accuracy, and format. 7. Insufficient Information Integration and Utilization: Due to the aforementioned difficulties, a large amount of valuable information is missed or cannot be effectively integrated. Current quality assessments rely heavily on threshold monitoring of individual parameters (such as current and time) or post-construction destructive testing, failing to achieve online, accurate, and comprehensive assessment of the quality of each solder joint.

[0005] 8. Limited visualization capabilities: Existing monitoring interfaces typically only display real-time trend charts for a single parameter or simple numerical reports. They lack interactive visualization methods that can correlate, scale, and compare multiple parameters on a unified time axis, which is not conducive to process engineers quickly locating the root cause of problems.

[0006] Therefore, there is an urgent need for a method that can overcome the above-mentioned defects, realize comprehensive, accurate, real-time acquisition, fusion and in-depth visualization analysis of multi-source heterogeneous servo welding data, and break down the barriers between data acquisition, analysis decision-making and execution control to achieve adaptive, forward-looking and collaborative control. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide a multi-source heterogeneous servo welding data acquisition and collaborative control method, system and readable storage medium to realize a closed loop of "perception-analysis-decision-control" and improve the adaptive control capability of welding quality.

[0008] The technical problem solved by this invention is achieved through the following technical solution: A method for multi-source heterogeneous servo welding data acquisition and collaborative control, the method comprising the following steps: S1. Data Acquisition and Fusion: Through multi-channel asynchronous acquisition and unified time stamping technology, servo pressure, servo displacement, welding secondary current, welding secondary voltage and equipment status information during the servo welding process are collected synchronously to form a multi-source data stream with a unified timestamp. S2. Data Slicing and Modeling: Based on the welding triggering event, data of the complete welding cycle is extracted from the multi-source data stream to form a data slice that uniquely corresponds to a single weld point, and the key quality feature values ​​of the weld point are extracted. S3. Collaborative Decision-Making and Control: The feature values ​​or real-time data stream of the current weld point data slice are input into a pre-trained edge control model. The edge control model is based on multivariate coupling analysis and outputs a collaborative control signal for adjusting the welding parameters of the next weld point. The collaborative control signal includes at least the adjustment amount of the servo pressure setpoint, welding current setpoint, or welding time setpoint. Collaborative decision-making and control are performed on computing devices deployed at the edge of the production line network, with a computation delay of less than one welding cycle time, to achieve online real-time control. S4. Execution Feedback: The coordinated control signal is sent to the welding controller and servo driver to execute the welding operation and collect new weld point data to complete the closed-loop control.

[0009] Moreover, S2 specifically refers to: Listen for the "welding start" event signal from the PLC and use it as the trigger point for the data slice. Extract all data from the multi-source data stream with the unified time scale, from the pre-pressure stage before "welding start" to the holding pressure stage after "welding end". Associate the data of all channels in the time window with a unique weld point number to form a data slice of a complete welding cycle, and store the data slice as a complete data object in the time series database.

[0010] Furthermore, when storing the data slice, the key characteristic values ​​of the weld joint are calculated and stored simultaneously, including: pressure peak value, displacement springback amount, effective current value, current peak value, energizing time, and pressure rise slope.

[0011] Moreover, the edge control model is based on a large number of historical welding data slices and is trained by machine learning algorithms to obtain a weld quality prediction model and a multi-objective parameter optimization model.

[0012] Moreover, the machine learning algorithm is a random forest, gradient boosting decision tree, or neural network; the model uses the feature values ​​of the current and several preceding weld points as input to predict the risk probability of quality defects in the next weld point, and solves in reverse the optimal combination of welding parameter adjustments.

[0013] A multi-source heterogeneous servo welding data collaborative control system for implementing the method includes: Multi-source heterogeneous data acquisition module: includes a unified data acquisition module and a preprocessing unit, used to implement the data perception and fusion steps; The associated storage module is used to implement the data slicing and modeling steps and store the data slices in the time series database. Interactive decision-making module: Provides a visual human-computer interface to display welding process curves, quality characteristics and analysis results, and allows engineers to configure control strategies and model parameters; Edge collaborative control module: As the core of the system, it has a built-in asynchronous parallel computing framework and the edge control model, which is used to execute the collaborative decision and control steps and generate collaborative control signals.

[0014] Furthermore, the multi-source heterogeneous data acquisition module includes: Data Acquisition Module: The module communicates with the servo driver and the PLC controller of the welding controller via the industrial Ethernet protocol, and obtains real-time servo pressure and servo displacement through servo encoder feedback; it also acquires transformer secondary voltage and secondary current through the analog input module; the module runs a multi-threaded acquisition engine, employing an asynchronous acquisition strategy for different data sources. High-frequency data stream channel: Real-time acquisition of analog signals of servo pressure and servo displacement at the first sampling frequency (≥1kHz); Low-frequency status data channel: acquires digital signals of welding secondary current, secondary voltage, welding status flag, and alarm code from the PLC controller at a second sampling frequency (≤100Hz); Event Marking Channel: Real-time monitoring of welding start, welding end, and alarm trigger event signals issued by the PLC; marking all acquired data packets with a unified microsecond-level timestamp from the module's high-precision clock, and encapsulating the data to form a multi-source data stream with a unified timestamp.

[0015] Furthermore, the interactive decision module is used to implement a visual interface and interactive functions, providing a control mode switching function, allowing users to switch between "manual monitoring mode," "automatic control mode," and "parameter self-learning mode." In the parameter self-learning mode, the system automatically incorporates newly generated qualified solder joint data into the training set for online updating of the edge control model; including: Main curve view: Within a coordinate plane, using a shared time axis, pressure-time curves, displacement-time curves, current-time curves, voltage-time curves, and dynamic resistance-time curves are simultaneously plotted in a single solder joint data slice; Interactive controller: Provides global timeline zoom and pan controls, allowing users to zoom in on any time interval of the main curve view to view details, with all curves moving in sync. Eigenvalue Parallel Coordinate Graph: In another view, the key eigenvalues ​​of multiple solder joints are displayed in the form of parallel coordinate axes. Each solder joint is represented as a broken line that crosses each coordinate axis, which is used to visually identify solder joint quality clusters and abnormal points. When a user selects one or more abnormal solder joints in the feature value parallel coordinate graph, the main curve view is automatically updated to display the dynamic curve of the entire process corresponding to these abnormal solder joints. Layer management: Users can select to show or hide any one or more curves in the main curve view, and it supports overlaying the standard curve of qualified solder joints as a reference layer.

[0016] Furthermore, the edge collaboration control module includes: Feature extraction unit: used to parse the input multi-source data stream in real time, calculate key feature values, and determine whether the welding process is in a normal state; Model Inference Unit: Loads and runs the edge control model, receives the output of the state feature extraction unit, and performs forward inference; Control signal generation unit: Converts the model inference results into an instruction protocol that can be recognized by the welding controller and servo driver, and sends it asynchronously.

[0017] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method.

[0018] The positive effects that this invention can produce are: 1. Achieved precise data fusion: By using a unified timescale and multi-threaded asynchronous acquisition, the problem of synchronizing high- and low-frequency data was solved, laying the foundation for accurate correlation analysis.

[0019] 2. Improved accuracy and comprehensiveness of quality judgment: By visualizing the shape and key features of the associated pressure, displacement and current curves, defects such as "splashing", "cold soldering" and "poor contact" can be identified more accurately, realizing process-based intelligent quality judgment.

[0020] 3. Significantly improves process optimization efficiency: Process engineers can intuitively see the impact of changes in process parameters on the entire process curve through interactive analysis, thereby quickly finding the optimal welding parameter window.

[0021] 4. Provide data support for predictive maintenance: By monitoring changes in pressure curve shape and displacement repeatability over a long period of time, the wear status of the servo mechanism or electrode head can be predicted.

[0022] 5. Adaptive control is achieved: The system can dynamically adjust parameters according to actual working conditions, automatically compensate for disturbances such as electrode wear and changes in plate gap, and ensure the consistency of weld quality.

[0023] 6. Improved control timeliness: By delegating control decisions to the edge, millisecond-level real-time response is achieved, fundamentally solving the control lag problem.

[0024] 7. The synergistic effect of parameters is brought into play: Through multivariate coupling analysis, pressure, current and time are jointly optimized to achieve a control effect of "1+1>2" and avoid the limitations of single parameter adjustment.

[0025] 8. A closed loop of data value has been formed: massive amounts of process data are transformed into valuable control knowledge and ultimately fed back into production execution, truly realizing data-driven manufacturing. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall system architecture and process of the present invention; Figure 2 This is a structural diagram of the servo welding system of the present invention. Detailed Implementation

[0027] The present invention will be further described in detail below through specific embodiments. The following embodiments are merely descriptive and not limiting, and should not be used to limit the scope of protection of the present invention.

[0028] A method for multi-source heterogeneous servo welding data acquisition and collaborative control, the innovation of which lies in the following steps: S1. Data Acquisition and Fusion: Through multi-channel asynchronous acquisition and unified time stamping technology, servo pressure, servo displacement, welding secondary current, welding secondary voltage and equipment status information during the servo welding process are collected synchronously to form a multi-source data stream with a unified timestamp.

[0029] S2. Data Slicing and Modeling: Based on the welding triggering event, data of the complete welding cycle is extracted from the multi-source data stream to form a data slice that uniquely corresponds to a single weld point, and the key quality feature values ​​of the weld point are extracted. Listen for the "welding start" event signal from the PLC and use it as the trigger point for the data slice; extract all data from the multi-source data stream with the unified time scale, from the pre-pressure stage before "welding start" to the pressure holding stage after "welding end" within the entire time window; associate the data of all channels within the time window with a unique weld point number to form a data slice of a complete welding cycle, and store the data slice as a complete data object in the time series database; When storing the data slice, the key characteristic values ​​of the weld joint are calculated and stored simultaneously, including: peak pressure, displacement rebound, effective current value, peak current, energizing time, and pressure rise slope.

[0030] S3. Collaborative Decision-Making and Control: The feature values ​​or real-time data stream of the current weld point data slice are input into a pre-trained edge control model. The edge control model is based on multivariate coupling analysis and outputs a collaborative control signal for adjusting the welding parameters of the next weld point. The collaborative control signal includes at least the adjustment amount of the servo pressure setpoint, welding current setpoint, or welding time setpoint. Collaborative decision-making and control are performed on computing devices deployed at the edge of the production line network, with a computation delay of less than one welding cycle time, to achieve online real-time control. The edge control model is based on a large number of historical welding data slices and is trained by machine learning algorithms to obtain a weld quality prediction model and a multi-objective parameter optimization model.

[0031] The machine learning algorithm is a random forest, gradient boosting decision tree, or neural network; the model takes the feature values ​​of the current and several previous weld points as input, predicts the risk probability of quality defects in the next weld point, and solves in reverse the optimal combination of welding parameter adjustments.

[0032] S4. Execution Feedback: The coordinated control signal is sent to the welding controller and servo driver to execute the welding operation and collect new weld point data to complete the closed-loop control.

[0033] An innovative multi-source heterogeneous servo welding data collaborative control system for implementing the method includes: Multi-source heterogeneous data acquisition module: includes a unified data acquisition module and a preprocessing unit, used to implement the data perception and fusion steps; The associated storage module is used to implement the data slicing and modeling steps and store the data slices in the time series database. Interactive decision-making module: Provides a visual human-computer interface to display welding process curves, quality characteristics and analysis results, and allows engineers to configure control strategies and model parameters; Edge collaborative control module: As the core of the system, it has a built-in asynchronous parallel computing framework and the edge control model, which is used to execute the collaborative decision and control steps and generate collaborative control signals.

[0034] The multi-source heterogeneous data acquisition module includes: Data Acquisition Module: The module communicates with the servo driver and the PLC controller of the welding controller via the industrial Ethernet protocol, and obtains real-time servo pressure and servo displacement through servo encoder feedback; it also acquires transformer secondary voltage and secondary current through the analog input module; the module runs a multi-threaded acquisition engine, employing an asynchronous acquisition strategy for different data sources. High-frequency data stream channel: Real-time acquisition of analog signals of servo pressure and servo displacement at the first sampling frequency (≥1kHz); Low-frequency status data channel: acquires digital signals of welding secondary current, secondary voltage, welding status flag, and alarm code from the PLC controller at a second sampling frequency (≤100Hz); Event Marking Channel: Real-time monitoring of welding start, welding end, and alarm trigger event signals issued by the PLC; marking all acquired data packets with a unified microsecond-level timestamp from the module's high-precision clock, and encapsulating the data to form a multi-source data stream with a unified timestamp.

[0035] The interactive decision module is used to implement a visual interface and interactive functions, providing a control mode switching function, allowing users to switch between "manual monitoring mode," "automatic control mode," and "parameter self-learning mode." In the parameter self-learning mode, the system automatically incorporates newly generated qualified solder joint data into the training set for online updating of the edge control model; including: Main curve view: Within a coordinate plane, using a shared time axis, pressure-time curves, displacement-time curves, current-time curves, voltage-time curves, and dynamic resistance-time curves are simultaneously plotted in a single solder joint data slice; Interactive controller: Provides global timeline zoom and pan controls, allowing users to zoom in on any time interval of the main curve view to view details, with all curves moving in sync. Eigenvalue Parallel Coordinate Graph: In another view, the key eigenvalues ​​of multiple solder joints are displayed in the form of parallel coordinate axes. Each solder joint is represented as a broken line that crosses each coordinate axis, which is used to visually identify solder joint quality clusters and abnormal points. When a user selects one or more abnormal solder joints in the feature value parallel coordinate graph, the main curve view is automatically updated to display the dynamic curve of the entire process corresponding to these abnormal solder joints. Layer management: Users can select to show or hide any one or more curves in the main curve view, and it supports overlaying the standard curve of qualified solder joints as a reference layer.

[0036] The edge collaboration control module includes: Feature extraction unit: used to parse the input multi-source data stream in real time, calculate key feature values, and determine whether the welding process is in a normal state; Model Inference Unit: Loads and runs the edge control model, receives the output of the state feature extraction unit, and performs forward inference; Control signal generation unit: Converts the model inference results into an instruction protocol that can be recognized by the welding controller and servo driver, and sends it asynchronously.

[0037] A computer-readable storage medium having a computer program stored thereon, wherein the innovation is that the program, when executed by a processor, implements the steps of the method described.

[0038] This embodiment applies the system to a spot welding station in an automotive body shop. The system is deployed at a side panel welding station of an automotive body.

[0039] 1. Data Acquisition: In the welding system, the servo driver and welding controller communicate with the Siemens S7-1500 PLC via the Profinet protocol. Communication data is periodically scanned in 1ms RT mode. A multi-threaded acquisition program is deployed in the communication module. - Thread 1 (High-frequency thread): Loops through servo pressure and displacement data fed back from the servo driver encoder at a frequency of 500Hz; - Thread 2 (Low-frequency thread): Reads secondary current and voltage from the welding controller at 200Hz; - Thread 2 (PLC communication thread): Exchanges data with the PLC at a frequency of 1000Hz.

[0040] - Thread 3 (Event Listener Thread): Continuously listens to the welding controller control bus and captures the rising edge of the welding start (from 0 to 1); - After obtaining the data, all threads immediately call the system's high-precision clock function to get the current time and put it into the memory queue along with the metadata of the data packet.

[0041] 2. Slicing and Storage: Each time welding is triggered, the system automatically generates a data slice containing all the data from this welding operation and calculates characteristic values ​​(such as peak pressure P_max, current integral I_sq, etc.), which are then stored in the database. The slice ID is associated with the vehicle body VIN code and weld point number.

[0042] 3. Decision-making and control (core): Initially, engineers trained a random forest model using data from thousands of successful solder joints via an interactive decision-making module. This model can predict the current fine-tuning required for the next solder joint to form a successful weld nugget based on features such as P_max and I_sq of the current solder joint.

[0043] Deployment: Deploy this model to the edge collaborative control module.

[0044] Operation: When welding the Nth weld point, the system calculates the feature value of the Nth weld point in real time and immediately inputs it into the edge model. The model outputs a welding current adjustment signal (e.g., +0.5 kA) for the (N+1)th weld point within 10 milliseconds.

[0045] Execution: The adjustment signal is written to the welding controller via the Profinet network before the start of the next welding cycle, thereby enabling real-time and proactive optimization of the welding process.

[0046] 4. Effect: When the electrode cap begins to wear and the contact resistance increases, the system monitors the subtle changes in the current waveform and increases the welding current in advance and slightly, which effectively compensates for the impact of electrode wear. This keeps the weld quality stable throughout the entire electrode life cycle and avoids the later quality decline or spatter problems caused by parameter solidification in traditional methods.

[0047] Although embodiments and drawings of the present invention have been disclosed for illustrative purposes, those skilled in the art will understand that various substitutions, variations and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A method for multi-source heterogeneous servo welding data acquisition and collaborative control, characterized in that: The steps of the method are as follows: S1. Data Acquisition and Fusion: Through multi-channel asynchronous acquisition and unified time stamping technology, servo pressure, servo displacement, welding secondary current, welding secondary voltage and equipment status information during the servo welding process are collected synchronously to form a multi-source data stream with a unified timestamp. S2. Data Slicing and Modeling: Based on the welding triggering event, data of the complete welding cycle is extracted from the multi-source data stream to form a data slice that uniquely corresponds to a single weld point, and the key quality feature values ​​of the weld point are extracted. S3. Collaborative Decision-Making and Control: The feature values ​​or real-time data stream of the current weld point data slice are input into a pre-trained edge control model. The edge control model is based on multivariate coupling analysis and outputs a collaborative control signal for adjusting the welding parameters of the next weld point. The collaborative control signal includes at least the adjustment amount of the servo pressure setpoint, welding current setpoint, or welding time setpoint. Collaborative decision-making and control are performed on computing devices deployed at the edge of the production line network, with a computation delay of less than one welding cycle time, to achieve online real-time control. S4. Execution Feedback: The coordinated control signal is sent to the welding controller and servo driver to execute the welding operation and collect new weld point data to complete the closed-loop control.

2. The multi-source heterogeneous servo welding data acquisition and collaborative control method according to claim 1, characterized in that: Specifically, S2 is: Listen for the "welding start" event signal from the PLC and use it as the trigger point for the data slice. Extract all data from the multi-source data stream with the unified time scale, from the pre-pressure stage before "welding start" to the pressure holding stage after "welding end". Associate the data of all channels in the time window with a unique weld point number to form a data slice of a complete welding cycle, and store the data slice as a complete data object in the time series database.

3. The multi-source heterogeneous servo welding data acquisition and collaborative control method according to claim 1, characterized in that: When storing the data slice, the key characteristic values ​​of the weld joint are calculated and stored simultaneously, including: peak pressure, displacement rebound, effective current value, peak current, energizing time, and pressure rise slope.

4. The multi-source heterogeneous servo welding data acquisition and collaborative control method according to claim 1, characterized in that: The edge control model is based on a large number of historical welding data slices and is trained by machine learning algorithms to obtain a weld quality prediction model and a multi-objective parameter optimization model.

5. The multi-source heterogeneous servo welding data acquisition and collaborative control method according to claim 4, characterized in that: The machine learning algorithm is a random forest, gradient boosting decision tree, or neural network; the model takes the feature values ​​of the current and several previous weld points as input, predicts the risk probability of quality defects in the next weld point, and solves in reverse the optimal combination of welding parameter adjustments.

6. A multi-source heterogeneous servo welding data collaborative control system for implementing the method as described in any one of claims 1 to 5, characterized in that: include: Multi-source heterogeneous data acquisition module: includes a unified data acquisition module and a preprocessing unit, used to implement the data perception and fusion steps; Used to build a unified data acquisition module, which achieves accurate alignment and encapsulation of high- and low-frequency data through multi-channel asynchronous acquisition and unified time-stamping technology; The associated storage module is used to implement the data slicing and modeling steps and store the data slices in the time series database. Interactive decision-making module: Provides a visual human-computer interface to display welding process curves, quality characteristics and analysis results, and allows engineers to configure control strategies and model parameters; Edge collaborative control module: As the core of the system, it has a built-in asynchronous parallel computing framework and the edge control model, which is used to execute the collaborative decision and control steps and generate collaborative control signals.

7. The multi-source heterogeneous servo welding data collaborative control system according to claim 6, characterized in that: The multi-source heterogeneous data acquisition module includes: Data Acquisition Module: The module communicates with the servo driver and the PLC controller of the welding controller via the industrial Ethernet protocol, and obtains real-time servo pressure and servo displacement through servo encoder feedback; it also acquires transformer secondary voltage and secondary current through the analog input module; the module runs a multi-threaded acquisition engine, employing an asynchronous acquisition strategy for different data sources. High-frequency data stream channel: Real-time acquisition of analog signals of servo pressure and servo displacement at the first sampling frequency (≥1kHz); Low-frequency status data channel: acquires digital signals of welding secondary current, secondary voltage, welding status flag, and alarm code from the PLC controller at a second sampling frequency (≤100Hz); Event Marking Channel: Real-time monitoring of welding start, welding end, and alarm trigger event signals issued by the PLC; marking all acquired data packets with a unified microsecond-level timestamp from the module's high-precision clock, and encapsulating the data to form a multi-source data stream with a unified timestamp.

8. The multi-source heterogeneous servo welding data collaborative control system according to claim 6, characterized in that: The interactive decision module is used to implement a visual interface and interactive functions, and provides a control mode switching function, allowing users to switch between "manual monitoring mode", "automatic control mode" and "parameter self-learning mode". In the parameter self-learning mode, the system automatically incorporates newly generated qualified weld point data into the training set for online updating of the edge control model; including: Main curve view: Within a coordinate plane, using a shared time axis, pressure-time curves, displacement-time curves, current-time curves, voltage-time curves, and dynamic resistance-time curves are simultaneously plotted in a single solder joint data slice; Interactive controller: Provides global timeline zoom and pan controls, allowing users to zoom in on any time interval of the main curve view to view details, with all curves moving in sync. Eigenvalue Parallel Coordinate Graph: In another view, the key eigenvalues ​​of multiple solder joints are displayed in the form of parallel coordinate axes. Each solder joint is represented as a broken line that crosses each coordinate axis, which is used to visually identify solder joint quality clusters and abnormal points. When a user selects one or more abnormal solder joints in the feature value parallel coordinate graph, the main curve view is automatically updated to display the dynamic curve of the entire process corresponding to these abnormal solder joints. Layer management: Users can select to show or hide any one or more curves in the main curve view, and it supports overlaying the standard curve of qualified solder joints as a reference layer.

9. The multi-source heterogeneous servo welding data collaborative control system according to claim 7, characterized in that: The edge collaboration control module includes: Feature extraction unit: used to parse the input multi-source data stream in real time, calculate key feature values, and determine whether the welding process is in a normal state; Model Inference Unit: Loads and runs the edge control model, receives the output of the state feature extraction unit, and performs forward inference; Control signal generation unit: Converts the model inference results into an instruction protocol that can be recognized by the welding controller and servo driver, and sends it asynchronously.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in claims 1 to 5.

Images