A welding process quality traceability and playback system and method based on digital twinning

Abstract

This application discloses a welding process quality traceability and playback system and method based on digital twins, including: a physical welding module, a quality inspection module, a digital twin modeling module, a data management module, a virtual-real fusion playback module, and a statistical analysis module. The physical welding module generates multi-source real-time data; the quality inspection module outputs quality result data; the digital twin modeling module constructs a three-dimensional virtual model; the data management module assigns unique identifiers to the workpiece to be welded and the weld seam, using these identifiers as a joint index to form data records based on individual weld seams; the virtual-real fusion playback module loads the virtual model and data records, drives the virtual model to reproduce the welding process, and overlays and displays the waveforms of process parameter execution, sensor characteristic curves, and quality results; the statistical analysis module retrieves and statistically analyzes weld seam records, outputting a correlation analysis between process parameter data and quality result data. This achieves unified association and storage of multi-source data, refined quality traceability, and virtual-real playback.

Description

Technical Field

[0001] This application relates to the fields of welding automation and digital twin technology, and in particular to a welding process quality traceability and playback system and method based on digital twin. Background Technology

[0002] With the continuous improvement of industrial automation, robotic welding technology has been widely used in fields such as engineering machinery, rail transportation, shipbuilding, and energy equipment. The stability of welding quality directly affects the safety and service life of structural components, thus placing increasingly higher demands on real-time monitoring and quality traceability of the welding process. Current mainstream robotic welding systems typically consist of an industrial robot body and controller, a digital welding power supply, weld seam tracking sensors, and various detection devices, enabling a certain degree of automated control and basic data recording of the welding process. However, these subsystems often operate independently. The robot controller records the motion trajectory and program number, the welding power supply stores process parameters such as current and voltage, the sensor system collects weld seam geometric information, and the quality inspection equipment outputs defect reports. Various types of data are stored in different devices, lacking a unified data model and time reference. When welding defects occur, users find it difficult to quickly correlate the quality problem of a specific weld seam with the robot's posture, process parameter fluctuations, and sensor characteristics at the time. Quality traceability heavily relies on personal experience, resulting in low efficiency and difficulty in forming a systematic knowledge base.

[0003] Meanwhile, the application of digital twin technology in the welding field is gradually emerging. Currently, it mainly involves building 3D models of robots and tooling using offline programming software for accessibility analysis, collision detection, and process simulation verification. These applications essentially remain at the design and simulation level, with weak coupling between the virtual model and the actual welding process, lacking spatiotemporal correlation between multi-source real-time data and the digital twin model. Even some high-end systems with process monitoring functions often record data in units of the entire welding program or a long time window, resulting in coarse data granularity and making it impossible to perform refined process reproduction and comparative analysis for individual welds or single workpieces. The lack of coordinate system calibration information between the weld path in virtual space and the actual robot trajectory leads to spatial discrepancies when comparing virtual and real data, making it difficult to support accurate defect location and cause determination. Summary of the Invention

[0004] This application provides a welding process quality traceability and playback system and method based on digital twins, which solves the technical problems of scattered multi-source data, disconnect between digital twin models and actual processes, and difficulty in achieving fine-grained quality traceability at the workpiece and weld levels in existing welding processes.

[0005] Firstly, this application provides a welding process quality traceability and playback system based on digital twins, including:

[0006] The module includes a physical welding module, a quality inspection module, a digital twin modeling module, a data management module, a virtual-real fusion playback module, and a statistical analysis module.

[0007] The physical welding module includes an industrial robot and its corresponding robot controller, a welding power supply and its corresponding power control unit, and a process sensing unit; it is used for welding operations of the industrial robot and generates multi-source real-time data reflecting the welding process; the multi-source real-time data includes robot motion data, process parameter data, and sensor feature data; the process parameter data includes process parameter set values ​​and process parameter executed values.

[0008] The quality inspection module is connected to the physical welding module in the process. It is used to perform quality inspection on the target workpiece after welding and output quality result data including defect type, defect location and defect level.

[0009] The digital twin modeling module is used to build and store a three-dimensional virtual model of the welding scene. The three-dimensional virtual model includes the linkage kinematic model of the industrial robot, the geometric model of the workpiece to be welded, and the weld space path corresponding to the weld seam on the workpiece to be welded, which is generated offline.

[0010] The data management module communicates with the physical welding module and the quality inspection module respectively. It is used to configure a unique workpiece identifier and weld identifier for each workpiece to be welded and each corresponding weld seam. Using the workpiece identifier and weld seam identifier as a joint index, it associates and stores multi-source real-time data, quality result data, and corresponding welding program version information and coordinate system calibration information to construct a data record with a single weld seam as the smallest data organization unit.

[0011] The virtual-real fusion playback module is connected to the digital twin modeling module and the data management module, respectively. It is used to respond to the playback command of the target weld, load the corresponding three-dimensional virtual model from the digital twin modeling module, and read the data record associated with the identifier of the target weld from the data management module. Based on the time series data in the data record, it drives the three-dimensional virtual model to reproduce the welding process, and displays the process parameter execution waveform generated by the process parameter execution value, the quality result data, and the sensor feature curve generated by the sensor feature data on the display interface, and perform spatiotemporally synchronized visualization and overlay display with the motion process of the three-dimensional virtual model.

[0012] The statistical analysis module, connected to the data management module, is used to receive statistical conditions, retrieve and statistically analyze data records of multiple target welds, and output the correlation analysis results between process parameter settings, process parameter execution values ​​and quality result data under different working conditions.

[0013] Secondly, this application provides a method for quality traceability and playback of welding processes based on digital twins, including:

[0014] A data model for the welding process is established. The data model uses workpiece identification and weld identification as a joint index, and writes welding program version information and coordinate system calibration information together with workpiece identification and weld identification into the database. The data model specifies the associated storage format of robot motion data, process parameter data, sensor feature data and quality result data.

[0015] During the welding process of the weld seam to be welded, multi-source real-time data is continuously collected with a unified time reference, and a data timestamp, corresponding workpiece identifier and weld seam identifier are added to each collected multi-source real-time data.

[0016] After the target workpiece is welded, the corresponding quality result data is obtained. The quality result data includes the defect type, defect location, and defect level. The quality result data is then correlated with the collected multi-source real-time data according to the corresponding workpiece identifier and weld identifier.

[0017] The multi-source real-time data and quality result data are organized according to the data format defined by the data model, with a single weld seam to be welded as the smallest data organization unit, to construct data records containing complete welding life cycle information and store them in the database;

[0018] In response to the playback command of the target weld, a three-dimensional virtual model corresponding to the target weld is loaded; the three-dimensional virtual model includes the geometric model of the workpiece to be welded and the weld space path as a spatial comparison reference; and according to the coordinate system calibration information, the virtual coordinate system in the three-dimensional virtual model is spatially aligned with the physical coordinate system of the welding scene.

[0019] The robot motion data is read from the data record in chronological order to drive the virtual robot in the 3D virtual model to reproduce the pose changes of each joint of the robot and the motion trajectory of the tool center point during the welding process; and the corresponding process parameter data and sensor feature data are read from the data record and displayed on the display interface in the form of process parameter execution waveforms and sensor feature curves in sync with the motion process of the virtual robot.

[0020] During playback, in response to the triggering operation of the quality inspection mark extracted from the quality result data on the display interface, the welding time period corresponding to the quality inspection mark is calculated according to the mapping relationship between the defect location and the arc length of the weld path, and the playback time axis is automatically positioned to the start time of the welding time period to display the robot motion data, process parameter data and sensor feature data within the welding time period.

[0021] In response to statistical analysis commands, cross-sample statistics are performed on data records of multiple target welds in the database to establish a quantitative relationship between quality result data and process parameter data, including process parameter set values ​​and process parameter execution values, under different welding conditions, and to determine the optimized process based on the quantitative relationship.

[0022] Thirdly, this application provides a readable medium including executable instructions, which, when executed by a processor of an electronic device, cause the electronic device to perform any of the methods described in the second aspect.

[0023] This application provides a welding process quality traceability and playback system and method based on digital twins, including: a physical welding module, a quality inspection module, a digital twin modeling module, a data management module, a virtual-real fusion playback module, and a statistical analysis module. The physical welding module includes a robot, a welding power source, and process sensing units, generating multi-source real-time data such as robot motion data, process parameter setpoints and execution values, and sensor feature data. The quality inspection module outputs quality result data including defect type, location, and level. The digital twin modeling module constructs a three-dimensional virtual model, including a robot kinematic model, a workpiece geometric model, and a weld space path. The data management module assigns a unique identifier to the workpiece to be welded and the weld, using the identifier as a joint index to associate and store multi-source real-time data, quality result data, program version, and coordinate system calibration information, forming data records based on a single weld. The virtual-real fusion playback module loads the virtual model and data records, drives the virtual model to reproduce the welding process, and overlays and displays the process parameter execution waveform, sensor feature curves, and quality results. The statistical analysis module retrieves and statistically analyzes weld records, outputting a correlation analysis of process parameter setpoints, execution values, and quality results. It enables unified association and storage of multi-source data, as well as refined quality traceability and virtual-real playback.

[0024] The further effects of the aforementioned non-conventional preferred method will be explained below in conjunction with specific embodiments. Attached Figure Description

[0025] To more clearly illustrate the embodiments of this application or the existing technical solutions, 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 recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A schematic diagram of a welding process quality traceability and playback system based on digital twin provided in an embodiment of this application;

[0027] Figure 2A schematic diagram of the data management module in a welding process quality traceability and playback system based on digital twins, provided as an embodiment of this application;

[0028] Figure 3 A schematic diagram of the structure of a virtual-real fusion playback module in a welding process quality traceability and playback system based on digital twins provided in an embodiment of this application;

[0029] Figure 4 This is a flowchart illustrating a welding process quality traceability and playback method based on digital twins, provided in one embodiment of this application. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] With the continuous improvement of industrial automation, robotic welding technology has been widely used in fields such as engineering machinery, rail transportation, shipbuilding, and energy equipment. The stability of welding quality directly affects the safety and service life of structural components, thus placing increasingly higher demands on real-time monitoring and quality traceability of the welding process. Current mainstream robotic welding systems typically consist of an industrial robot body and controller, a digital welding power supply, weld seam tracking sensors, and various detection devices, enabling a certain degree of automated control and basic data recording of the welding process. However, these subsystems often operate independently. The robot controller records the motion trajectory and program number, the welding power supply stores process parameters such as current and voltage, the sensor system collects weld seam geometric information, and the quality inspection equipment outputs defect reports. Various types of data are stored in different devices, lacking a unified data model and time reference. When welding defects occur, users find it difficult to quickly correlate the quality problem of a specific weld seam with the robot's posture, process parameter fluctuations, and sensor characteristics at the time. Quality traceability heavily relies on personal experience, resulting in low efficiency and difficulty in forming a systematic knowledge base.

[0032] Meanwhile, the application of digital twin technology in the welding field is gradually emerging. Currently, it mainly involves building 3D models of robots and tooling using offline programming software for accessibility analysis, collision detection, and process simulation verification. These applications essentially remain at the design and simulation level, with weak coupling between the virtual model and the actual welding process, lacking spatiotemporal correlation between multi-source real-time data and the digital twin model. Even some high-end systems with process monitoring functions often record data in units of the entire welding program or a long time window, resulting in coarse data granularity and making it impossible to perform refined process reproduction and comparative analysis for individual welds or single workpieces. The lack of coordinate system calibration information between the weld path in virtual space and the actual robot trajectory leads to spatial discrepancies when comparing virtual and real data, making it difficult to support accurate defect location and cause determination.

[0033] To address this issue, this application proposes a welding process quality traceability and playback system based on digital twins. This system aims to solve the technical problems of fragmented multi-source data, disconnect between the digital twin model and the actual process, and difficulty in achieving refined quality traceability at the workpiece and weld levels in existing welding processes. See also... Figure 1 The image shows a specific embodiment of a welding process quality traceability and playback system based on digital twins provided by the present invention. In this embodiment, the welding process quality traceability and playback system based on digital twins includes: a physical welding module 101, a quality inspection module 102, a digital twin modeling module 103, a data management module 104, a virtual-real fusion playback module 105, and a statistical analysis module 106.

[0034] The physical welding module 101 is responsible for executing welding operations and generating multi-source real-time data. The quality inspection module 102 performs quality inspection on the workpiece after welding. The digital twin modeling module 103 provides a three-dimensional virtual model of the welding scene. The data management module 104 uniformly collects, identifies, and stores the above data. The virtual-real fusion playback module 105 drives the virtual model to reproduce the welding process and provides a visual display when traceability is required. The statistical analysis module 106 performs cross-sample statistical analysis on historical data. The collaborative operation of these modules enables the system to achieve closed-loop management of the entire process from data collection and storage to traceability playback and statistical analysis.

[0035] The physical welding module 101 includes an industrial robot and its corresponding robot controller, a welding power supply and its corresponding power control unit, and a process sensing unit; it is used for welding operations of the industrial robot and generates multi-source real-time data reflecting the welding process; the multi-source real-time data includes robot motion data, process parameter data, and sensor feature data; the process parameter data includes process parameter set values ​​and process parameter executed values.

[0036] During the welding process, the robot controller periodically outputs robot motion data such as joint angles, tool center point position, and posture. The power control unit records in real time the set values ​​and actual execution values ​​of process parameters such as welding current, welding voltage, and wire feed speed. The set values ​​of process parameters are derived from pre-configured process instructions in the welding program, while the executed values ​​are the physical quantity measurements sampled in real time by the power control unit during the welding process. Both together reflect the execution status of the welding process. The process sensing unit performs online sensing of the weld area during the movement of the welding torch and outputs sensing feature data reflecting the geometric shape of the weld.

[0037] The process sensing unit includes a line laser profile sensor and a signal processing subunit. The signal processing subunit is used to filter and extract features from the laser cross-section profile data acquired frame by frame by the line laser profile sensor. It generates the weld center deviation curve and height deviation curve with the arc length coordinate in the weld length direction as the horizontal axis, which constitute the weld geometric deviation component in the sensing feature data. The signal processing subunit outputs the weld geometric deviation component to the data management module, and the virtual-real fusion playback module maps the weld geometric deviation component to the weld path of the three-dimensional virtual model in a three-dimensional spatial superposition manner during playback.

[0038] A line laser profile sensor is installed near the welding torch and moves synchronously with the torch along the weld path, acquiring cross-sectional profile data perpendicular to the welding direction at a fixed frame rate. The signal processing subunit sequentially performs noise filtering and edge feature extraction on each frame of profile data, identifying the weld center position and height information. Combined with the accumulated arc length of the robot tool's center point, the frame-by-frame feature values ​​are converted into continuous curves with arc length coordinates as the horizontal axis, namely the weld center deviation curve and the height deviation curve. During playback, the virtual-real fusion playback module 105 overlays the geometric deviation values ​​from these curves onto the weld path of the virtual model in the form of three-dimensional deviation annotations, according to the correspondence between arc length coordinates and the weld spatial path. This allows the user to intuitively judge the spatial deviation of the actual weld trajectory from the designed path.

[0039] The quality inspection module 102 is connected to the physical welding module 101 in the process. It is used to perform quality inspection on the target workpiece after welding and output quality result data including defect type, defect location and defect level.

[0040] After the workpiece welding is completed, the quality inspection module 102 inspects each weld seam on the target workpiece one by one using methods such as visual inspection, ultrasonic inspection, or radiographic inspection. The inspection results are recorded in the form of structured data, where the defect type describes the morphological characteristics of the defect (such as porosity, lack of fusion, undercut, etc.), the defect location is represented by weld arc length coordinates or spatial coordinates, and the defect level is graded according to relevant standards to indicate the severity of the defect. The quality result data is linked with the corresponding multi-source real-time data through workpiece identification and weld identification, and stored in the data management module 104 to provide a quality basis for subsequent virtual-real fusion playback and statistical analysis.

[0041] The digital twin modeling module 103 is used to construct and store a three-dimensional virtual model of the welding scene. The three-dimensional virtual model includes the linkage kinematic model of the industrial robot, the geometric model of the workpiece to be welded, and the weld space path corresponding to the weld seam on the workpiece to be welded, which is generated offline.

[0042] The linkage kinematic model of the industrial robot is established based on the robot's DH parameters. It can calculate the spatial pose of the end effector's center point in real time based on the angles of each joint, thereby driving the virtual robot to reproduce the joint movements and trajectories in the actual welding process. The geometric model of the workpiece to be welded is generated by importing CAD design files. It is used to intuitively present the workpiece structure in a 3D scene and serves as the spatial carrier of the weld path. The weld spatial path is generated by offline programming software based on the weld position in the workpiece geometric model. It records the spatial coordinates of each weld path point and the welding torch posture, serving as a benchmark reference for virtual-real comparison during playback. The 3D virtual model is constructed and persistently stored during the system initialization phase and loaded on demand by the virtual-real fusion playback module 105 during the playback phase.

[0043] The data management module 104 is communicatively connected to the physical welding module 101 and the quality inspection module 102, respectively. It is used to configure a unique workpiece identifier and weld identifier for each workpiece to be welded and each corresponding weld seam. Using the workpiece identifier and weld seam identifier as a joint index, it associates and stores multi-source real-time data, quality result data, and corresponding welding program version information and coordinate system calibration information to construct a data record with a single weld seam as the smallest data organization unit.

[0044] Before the welding task begins, the data management module 104 assigns a unique identifier to each workpiece and its weld seams based on the current production work order information, and binds the collected multi-source real-time data with the corresponding identifiers in real time during the welding process. After welding is completed, the quality result data output by the quality inspection module 102 is also written to the database using the workpiece identifier and weld seam identifier as indexes, forming a complete association with the process data. The welding program version information records the program version number executed in the current welding, and the coordinate system calibration information stores the transformation matrices of the robot base coordinate system, workpiece coordinate system, and tooling coordinate system. Together, they ensure that the welding execution environment at the time can be accurately restored during traceability and playback, achieving precise alignment between virtual and real spaces.

[0045] like Figure 2 The diagram shown is a structural schematic of the data management module 104 in a welding process quality traceability and playback system based on digital twins, as described in this embodiment. Figure 2 In the example, the data management module 104 includes: a data acquisition unit 1041, an identification unit 1042, and a storage unit 1043.

[0046] The acquisition unit 1041 is responsible for collecting data streams from various devices in real time during the welding process and establishing a unified time reference. The identification unit 1042 assigns a structured and unique identifier to the workpiece and weld. The storage unit 1043 uses the above identifier as an index to archive various types of data in a unified manner. The three work together to complete the entire process management from data acquisition to data storage.

[0047] The acquisition unit 1041 is used to acquire multi-source real-time data from the physical welding module 101 during the welding process at a unified time reference, and to add data timestamps to the multi-source real-time data.

[0048] The acquisition unit 1041 synchronously initiates the acquisition process of each data source at the start of welding, establishing independent data buffer queues for the motion data output by the robot controller, the process parameter data output by the power control unit, and the sensor feature data output by the process sensing unit. To address the issue of inconsistent sampling frequencies among different data sources, the acquisition unit 1041 uses the timestamp sequence of the data source with the lowest sampling frequency as the reference time axis. Within a set time synchronization window, it downsamples and aligns the high-frequency data sources. For data packet loss or timeouts during the acquisition process, it compensates using linear interpolation or previous value preservation methods, while also adding quality flags to mark the compensated data, ensuring that the final stored multi-source data is completely aligned in the time dimension.

[0049] The acquisition unit 1041 is connected to the robot controller and power control unit via industrial Ethernet or fieldbus, and uses a precise time synchronization protocol to calibrate the local clocks of each data source corresponding to the multi-source real-time data.

[0050] The acquisition unit 1041 establishes a communication connection with the robot controller and power control unit via fieldbuses such as industrial Ethernet. It employs the IEEE 1588 Precision Time Protocol (PTP) or a similar time synchronization mechanism to periodically send clock calibration messages to each data source, keeping the deviation between the local clock of each device and the system master clock within microseconds. This ensures that multi-source real-time data from different devices has a unified time reference when timestamps are added, providing a fundamental guarantee for spatiotemporal synchronization in subsequent virtual-real fusion playback.

[0051] The identification unit 1042 is used to configure a unique workpiece identifier and weld identifier for each workpiece to be welded and the corresponding weld seam. The workpiece identifier is generated by combining the production order number, production date and serial number, and the weld seam identifier is generated according to the workpiece structural part and weld seam serial number.

[0052] The workpiece identification uses a combined coding format of "production order number - production date - serial number," which can locate a specific workpiece in a production batch, facilitating batch-based quality traceability and statistical analysis. Weld identification is based on predefined part names and weld sequence numbers in the workpiece structure drawing, for example, generated using abbreviations of workpiece structural parts plus weld sequence numbers, ensuring each weld has a unique and interpretable identifier at the workpiece level. The identification unit 1042 pre-generates and allocates identifications when the welding task is issued, and synchronously transmits the identification information to the acquisition unit 1041 and storage unit 1043, ensuring that the data remains bound to the workpiece and weld identification throughout the entire acquisition and storage process.

[0053] Storage unit 1043 is used to associate and store multi-source real-time data, quality result data, welding program version information and coordinate system calibration information with workpiece identifier and weld identifier as a joint index, and to construct a data record with a single weld seam to be welded as the smallest data organization unit.

[0054] Storage unit 1043 uses a relational database or time-series database as its underlying storage engine, and a combination of workpiece identifier and weld identifier as its primary key. It archives the time series of robot motion data, process parameter data, sensor feature data, quality result data, welding program version information, and coordinate system calibration information corresponding to the same weld into a single data record. This storage structure, with a single weld as the smallest data organization unit, allows for the retrieval of all data throughout the entire welding lifecycle of a weld through a joint query of the workpiece identifier and weld identifier during quality traceability. This eliminates the need for data aggregation across multiple independent systems, significantly improving traceability efficiency.

[0055] The virtual-real fusion playback module 105 is connected to the digital twin modeling module 103 and the data management module 104 respectively. In response to the playback command of the target weld, it loads the corresponding three-dimensional virtual model from the digital twin modeling module 103 and reads the data record associated with the identifier of the target weld from the data management module 104. Based on the time series data in the data record, it drives the three-dimensional virtual model to reproduce the welding process. On the display interface, the process parameter execution waveform generated by the process parameter execution value, the quality result data, and the sensor feature curve generated by the sensor feature data are displayed in a spatiotemporally synchronized visualization overlay with the motion process of the three-dimensional virtual model.

[0056] When quality traceability of a weld is required, a playback command can be initiated through the workpiece identifier and weld identifier. After responding to the command, the virtual-real fusion playback module 105 first loads the corresponding three-dimensional virtual model from the digital twin modeling module 103, and performs spatial alignment of each coordinate system in the virtual scene according to the stored coordinate system calibration information to ensure that the motion trajectory of the virtual robot is consistent with the spatial environment during actual welding.

[0057] Subsequently, module 105 reads the complete data record of the weld from data management module 104, drives the virtual robot to reproduce the joint movement and tool center point trajectory according to the time sequence, and simultaneously displays the process parameter execution waveform and sensor characteristic curve in the form of waveform graph and curve graph on the display interface, and superimposes the quality inspection mark on the corresponding weld path position to realize the spatiotemporal synchronous visualization playback of the welding process.

[0058] like Figure 3 The diagram shown is a schematic representation of the virtual-real fusion playback module 105 in a welding process quality traceability and playback system based on digital twins, as described in this embodiment. Figure 3 In the example, the virtual-real fusion playback module 105 includes: a model loading unit 1051, a data reading unit 1052, a synchronization driving unit 1053, and a visualization overlay unit 1054.

[0059] The model loading unit 1051 completes the construction and coordinate alignment of the virtual scene, the data reading unit 1052 acquires the complete data record of the target weld, the synchronous drive unit 1053 drives the virtual robot to move based on the time series data, and the visualization overlay unit 1054 synchronously overlays and displays the process parameters, sensing features and quality information. Together, the four achieve a complete spatiotemporal reproduction of the welding process.

[0060] The model loading unit 1051 is used to load a three-dimensional virtual model corresponding to the target weld from the digital twin modeling module in response to the playback command; and restore the transformation relationship between the robot base coordinate system, the workpiece coordinate system and the tooling coordinate system according to the coordinate system calibration information, so as to align the virtual coordinate system in the three-dimensional virtual model with the physical coordinate system space of the welding scene.

[0061] After receiving the playback command, the model loading unit 1051 retrieves and loads a matching 3D virtual model from the digital twin modeling module 103 based on the workpiece model and welding program version corresponding to the target weld. After loading, the model loading unit 1051 reads the coordinate system calibration information saved during the current welding from the data record. Based on the transformation matrix between the robot base coordinate system, workpiece coordinate system, and tooling coordinate system recorded therein, it restores the relative pose relationships of each coordinate system in the virtual scene, eliminating spatial deviations introduced by differences in workpiece clamping in different batches of production, and ensuring that the motion reproduction of the virtual robot is precisely consistent with the spatial environment during actual welding execution.

[0062] The data reading unit 1052 is used to read data records associated with the target weld identifier from the data management module 104.

[0063] The data reading unit 1052 uses the workpiece identifier and weld identifier of the target weld as the joint query key to initiate a data retrieval request to the data management module 104. It reads the robot motion data time series, process parameter data time series, sensor feature data and quality result data corresponding to the weld in one go. After organizing the above data in the order of timestamps, it transmits them to the synchronization drive unit 1053 and the visualization overlay unit 1054 respectively for subsequent motion reproduction and visualization display.

[0064] The synchronous drive unit 1053 is used to drive the virtual robot in the three-dimensional virtual model in time sequence according to the time series data, so as to reproduce the pose changes of each joint of the robot and the motion trajectory of the tool center point during the welding process.

[0065] The synchronization drive unit 1053 uses the timestamp sequence in the data record as the playback timeline. Following the original sampling interval or the playback rate set by the user, it inputs the joint angle values ​​from the robot motion data frame by frame into the link kinematics model. Through forward kinematics calculation, it updates the spatial pose of each link of the virtual robot. Simultaneously, it continuously overlays the calculated tool center point position onto the 3D scene as a trajectory line, forming an actual motion trajectory that can be intuitively compared with the weld seam spatial path. During playback, the user can adjust the playback rate or drag the timeline to any point in time. The synchronization drive unit 1053 responds to this operation and positions the virtual robot to the corresponding pose at that moment.

[0066] The visualization overlay unit 1054 is used to display, in conjunction with the synchronous drive unit, the waveform of process parameter execution, the sensing characteristic curve, and the quality inspection marks extracted from the quality result data for identifying defect locations, in a spatiotemporally synchronized visualization overlay of the motion process of the three-dimensional virtual model on the display interface.

[0067] The visualization overlay unit 1054 adopts a multi-view linkage layout on the display interface: the three-dimensional view area displays the motion process of the virtual robot and the weld path, and the waveform area synchronously scrolls to display the corresponding process parameter execution waveform (such as current and voltage waveforms) and sensing characteristic curves (such as weld side deviation curves and height deviation curves) at the current moment, and uses a vertical cursor to mark the position of the current playback moment in the waveform.

[0068] Quality inspection marks are displayed as icons overlaid on the corresponding positions of the weld path in the 3D view. When the user clicks on the mark, the visualization overlay unit 1054 will automatically position the playback time axis to the corresponding welding period according to the mapping relationship between the defect location and the weld arc length, so as to realize the rapid association and positioning of the defect location and process data.

[0069] The statistical analysis module 106, connected to the data management module 104, is used to receive statistical conditions, retrieve and statistically analyze the data records of multiple target welds, and output the correlation analysis results between the process parameter set values, process parameter execution values ​​and quality result data under different working conditions.

[0070] The statistical analysis module 106 reads multiple weld data records that meet the statistical conditions from the data management module 104 in batches. It performs quantitative analysis on the deviation distribution between the set and executed values ​​of process parameters, the occurrence of defects under different process parameter ranges, etc., and establishes a statistical correlation between process parameters and quality results. The results of this correlation analysis help identify process parameters that have a significant impact on welding quality and their critical ranges, providing data support for process optimization.

[0071] The statistical analysis module 106 is also used to support multi-dimensional condition combination filtering based on workpiece batch number, weld geometry type, welding spatial posture, parameter range corresponding to process parameter settings, and defect type; it outputs defect incidence rate, rework rate, and process capability index for the filtering results, and generates statistical reports in the form of process parameter distribution histograms and defect Pareto charts; the defect incidence rate, rework rate, and process capability index constitute the quality indicators obtained after statistical analysis of the quality result data; based on historical statistical results and feature matching of weld geometry type and working conditions, it generates initial process parameter value recommendations for newly imported workpieces with similar weld types and geometric features.

[0072] The statistical analysis module 106 supports free combination and filtering of multi-dimensional conditions. For example, it can simultaneously specify the workpiece batch number range, specific weld geometry type (such as fillet weld, butt weld), welding spatial posture (such as flat welding, vertical welding, overhead welding), process parameter range and defect type. The system will retrieve weld records that meet all conditions from the database, calculate and output the defect occurrence rate, rework rate and process capability index, and automatically generate a statistical report containing a process parameter distribution histogram and a defect Pareto chart, which intuitively reflects the distribution pattern and main influencing factors of various defects under the current process status.

[0073] Regarding the initial value recommendation of process parameters, the statistical analysis module 106 performs feature matching between the weld geometry features and working conditions of the newly imported workpiece and the weld records in the historical database, extracts the process parameter combinations corresponding to welds with high quality pass rates under similar working conditions in the past, and outputs them as the initial value recommendation of process parameters for the new workpiece, which helps to shorten the process debugging cycle.

[0074] As can be seen from the above technical solutions, the beneficial effects of this embodiment are:

[0075] This application provides a welding process quality traceability and playback system based on digital twins, comprising: a physical welding module, a quality inspection module, a digital twin modeling module, a data management module, a virtual-real fusion playback module, and a statistical analysis module. The physical welding module includes a robot, a welding power source, and process sensing units, generating multi-source real-time data such as robot motion data, process parameter setpoints and execution values, and sensor feature data. The quality inspection module outputs quality result data including defect type, location, and level. The digital twin modeling module constructs a three-dimensional virtual model, including a robot kinematic model, a workpiece geometric model, and a weld space path. The data management module assigns a unique identifier to the workpiece to be welded and the weld, using the identifier as a joint index to associate and store multi-source real-time data, quality result data, program version, and coordinate system calibration information, forming data records based on a single weld. The virtual-real fusion playback module loads the virtual model and data records, drives the virtual model to reproduce the welding process, and overlays and displays the process parameter execution waveform, sensor feature curves, and quality results. The statistical analysis module retrieves and statistically analyzes weld records, outputting a correlation analysis of process parameter setpoints, execution values, and quality results. It enables unified association and storage of multi-source data, as well as refined quality traceability and virtual-real playback.

[0076] like Figure 4 The image shows a specific embodiment of a welding process quality traceability and playback method based on digital twins according to this application.

[0077] In this embodiment, a welding process quality traceability and playback method based on digital twins includes the following steps:

[0078] Step 401: Establish a data model for the welding process; the data model uses the workpiece identifier and weld identifier as a joint index, and writes the welding program version information and coordinate system calibration information together with the workpiece identifier and weld identifier into the database; the data model specifies the associated storage format of robot motion data, process parameter data, sensor feature data and quality result data.

[0079] The data model uses a combination of workpiece identifier and weld identifier as the primary key, with each data record corresponding to the complete welding lifecycle information of a single weld to be welded. Fixed fields are set in the data model to store welding program version information and coordinate system calibration information. The welding program version information records the program version number executed in the current welding operation, while the coordinate system calibration information stores the transformation matrix between the robot base coordinate system, workpiece coordinate system, and tooling coordinate system. Both are automatically written by the system when the welding task starts, serving as the basis for reconstructing the welding execution environment during subsequent playback.

[0080] The data model specifies that robot motion data, process parameter data, and sensor feature data are stored in time-series format. Each time-series data entry carries a data timestamp and a quality flag. Quality result data records defect type, defect location, and defect level in structured fields and is associated with the time-series data through workpiece and weld identifiers. The predefined data model ensures the consistency of data format in subsequent steps of data acquisition, storage, and retrieval, providing a prerequisite for cross-sample statistical analysis and traceability playback.

[0081] Step 402: During the welding process of the weld seam to be welded, multi-source real-time data is continuously collected with a unified time reference, and a data timestamp, corresponding workpiece identifier and weld seam identifier are added to each collected multi-source real-time data.

[0082] When a welding task begins, the data acquisition unit simultaneously initiates data acquisition processes from the robot controller, power control unit, and process sensing unit. It also calibrates the local clocks of each data source using a precise time synchronization protocol to establish a unified time reference. Throughout the welding process, the acquisition unit continuously reads robot motion data, process parameter data, and sensor feature data from each data source. It adds a timestamp based on the unified time reference to each acquired data record and simultaneously binds it to the workpiece identifier and weld identifier of the current workpiece to be welded, ensuring that each piece of data can be accurately traced to a specific weld on a specific workpiece.

[0083] Buffer queues are established for real-time data from multiple sources collected from different data sources; a time synchronization window length is set, with the timestamp sequence of the data source with the lowest sampling frequency as the reference time axis; within the time synchronization window length, the high-frequency data source is downsampled, and the time point closest to the reference time axis is selected for alignment; linear interpolation is used to compensate for data packet loss or timeout that occurs during the collection process; when the duration of continuous packet loss exceeds the preset threshold, previous value hold compensation is used, and a quality flag bit is added.

[0084] Since the sampling frequencies of the robot controller, power control unit and process sensing unit are usually inconsistent, the acquisition unit maintains an independent buffer queue for each data source to cache the original data stream and avoid data loss due to differences in processing speed.

[0085] During time alignment, the timestamp sequence of the data source with the lowest sampling frequency is used as the reference time axis. Within the set time synchronization window, the sampling point whose timestamp is closest to the reference time axis is selected from the buffer queue for high-frequency data sources to complete the alignment, and multiple data from different frequencies are uniformly mapped to the same time axis.

[0086] For occasional data packet loss caused by communication jitter or device response delay, linear interpolation between adjacent valid sampling points is used for compensation to maintain the continuity of the data sequence. When the duration of continuous packet loss from a certain data source exceeds a preset threshold, it is determined that the data segment cannot be reliably recovered by interpolation. Instead, the previous value preservation method is used for compensation, that is, the missing interval is filled with the last valid sampling value. At the same time, a quality flag is added to the compensated data in this interval. In subsequent analysis, the measured data and the compensated data can be identified and distinguished, avoiding interference of the compensated data with the statistical analysis results.

[0087] Step 403: After the target workpiece is welded, obtain the corresponding quality result data of the target workpiece; the quality result data includes the defect type, defect location and defect level; establish the association between the quality result data and the collected multi-source real-time data according to the corresponding workpiece identifier and weld identifier.

[0088] After all welds on the target workpiece are completed, the workpiece enters the quality inspection process. The quality inspection module 102 inspects each weld on the workpiece one by one, outputting the inspection results in structured data format. Defect locations are represented by weld arc length coordinates, facilitating subsequent time-dimension mapping and conversion with welding sequence data. Upon completion of the quality inspection, the system writes the inspection results into the database using the workpiece identifier and weld identifier as indexes. An explicit association is established with the multi-source real-time data collected during the welding process for that weld, forming a complete correspondence between process data and quality results. This provides complete input data for subsequent data recording, construction, and traceability analysis.

[0089] Step 404: Combine the multi-source real-time data and quality result data with the data format defined by the data model, using a single weld seam as the smallest data organization unit, construct a data record containing complete welding lifecycle information, and store it in the database.

[0090] After the multi-source real-time data acquisition is completed and the quality result data is associated and written, the storage unit archives the robot motion data time series, process parameter data time series, sensor feature data, quality result data, welding program version information and coordinate system calibration information corresponding to the same weld seam in accordance with the data model format predefined in the steps, constructs a complete data record with the weld seam as the unit and persists it in the database.

[0091] The data records cover the entire lifecycle information from the start of welding to the completion of quality inspection. Using workpiece identification and weld identification as a joint index, it supports accurate retrieval by identification during the traceability stage, and obtains all the data of the target weld at once without the need for data aggregation across multiple subsystems.

[0092] Step 405: In response to the playback command of the target weld, load the three-dimensional virtual model corresponding to the target weld; the three-dimensional virtual model includes the geometric model of the workpiece to be welded and the weld space path as a spatial comparison reference; and according to the coordinate system calibration information, spatially align the virtual coordinate system in the three-dimensional virtual model with the physical coordinate system of the welding scene.

[0093] When a user initiates a playback command using workpiece and weld identifiers, the model loading unit retrieves and loads a matching 3D virtual model from the digital twin modeling module based on the workpiece model and welding program version corresponding to the target weld. The weld space path in the 3D virtual model is pre-generated by offline programming software, recording the spatial coordinates of each path point and the welding torch's designed posture, serving as a spatial comparison benchmark for the virtual robot's actual motion trajectory during playback.

[0094] After the model is loaded, the model loading unit reads the coordinate system calibration information saved during the current welding from the data record, and restores the relative pose relationship of the robot base coordinate system, workpiece coordinate system and tooling coordinate system in the virtual scene according to the transformation matrix therein. This eliminates the spatial deviation caused by the workpiece clamping difference in different batches of production, and ensures that the motion reproduction of the virtual robot is accurately consistent with the spatial environment during the actual welding execution.

[0095] Step 406: Read the robot motion data from the data record in chronological order, drive the virtual robot in the 3D virtual model to reproduce the pose changes of each joint of the robot and the motion trajectory of the tool center point during the welding process; and read the corresponding process parameter data and sensor feature data from the data record, and display them on the display interface in the form of process parameter execution waveforms and sensor feature curves in sync with the motion process of the virtual robot.

[0096] After the data reading unit reads the complete data record from the database at once using the joint identifier of the target weld, the synchronous drive unit uses the timestamp sequence in the data record as the playback time axis. According to the original sampling interval or the playback rate set by the user, it inputs the joint angle values ​​in the robot motion data frame by frame into the link kinematic model. The spatial pose of each link of the virtual robot is updated through forward kinematics calculation, and the calculated tool center point position is continuously superimposed and drawn in the three-dimensional scene in the form of trajectory lines to form the actual motion trajectory that can be intuitively compared with the weld spatial path.

[0097] At the same time, the visualization overlay unit synchronously scrolls and displays the waveforms of the process parameters (including welding current, voltage, etc.) and sensing characteristic curves (including weld side deviation curves, height deviation curves, etc.) corresponding to the current moment in the waveform area of ​​the display interface, and marks the position of the current playback moment in the waveform with a vertical cursor, so that the three-dimensional motion reproduction and the process data waveform are strictly synchronized in the time dimension. This allows the time axis to be dragged to any moment for fixed-point viewing at any time during playback.

[0098] Step 407: During playback, in response to the triggering operation of the quality inspection mark extracted from the quality result data on the display interface, the welding time period corresponding to the quality inspection mark is calculated according to the arc length mapping relationship between the defect location and the weld path, and the playback time axis is automatically positioned to the start time of the welding time period to display the robot motion data, process parameter data and sensor feature data within the welding time period.

[0099] Quality inspection markers are displayed as overlaid icons at corresponding positions on the weld path in the 3D view. When a user clicks on a quality inspection marker, a positioning operation is triggered. Based on the arc length coordinates of the defect and the correspondence between the cumulative arc length of the tool center point and the data timestamp in the data record, the system converts the defect location into a specific time period on the welding timeline and automatically jumps the playback timeline to the start time of that time period. The virtual robot in the 3D view is simultaneously positioned to the corresponding posture, and the waveform area synchronously displays the process parameter execution waveforms and sensor characteristic curves within that time period. This allows users to quickly focus on all process data during the defect occurrence period without the need for manual frame-by-frame retrieval.

[0100] Based on the correspondence between the arc length coordinates of the weld space path and the data timestamp, the defect location corresponding to the quality inspection mark is converted into a defect time period. A dynamic time warping algorithm is used to compare the similarity of the process parameter execution waveform within the defect time period with a pre-stored normal waveform template to identify parameter fluctuation patterns that deviate from the normal range. The attitude angle change sequence within the defect time period is compared with the design attitude corresponding to the weld space path to calculate the attitude deviation. Based on the matching degree between the parameter fluctuation pattern and historical defect samples, the magnitude of the attitude deviation, and the degree of anomaly in the sensor feature data, a weighted fusion algorithm is used to calculate the confidence level of each candidate cause, generating a candidate list of defect causes sorted by confidence level, which is then displayed on the interface.

[0101] After locating the defect period, the system automatically initiates the defect cause analysis process. First, a dynamic time warping algorithm is used to compare the waveform of the process parameters executed during the defect period with the normal waveform template of the same type of weld that is pre-stored in the database, and to identify the fluctuation patterns of parameters such as current and voltage that deviate from the normal range during that period and their duration.

[0102] Secondly, the sequence of attitude angle changes of the robot end effector during the defect period is compared point by point with the design attitude of the weld space path in the corresponding arc length segment. The deviation between the actual executed attitude and the design attitude is calculated to determine whether there is a possibility that the abnormal behavior of the molten pool is caused by the attitude deviation.

[0103] Simultaneously, anomaly indicators of the sensor feature data within the defect period are extracted, such as the extent of weld lateral deviation or height deviation exceeding the limit. Finally, the above three types of analysis results—the matching degree between parameter fluctuation patterns and historical defect samples, attitude deviation, and sensor feature anomaly—are used as input features. A weighted fusion algorithm is employed to calculate the confidence score for each candidate cause (such as process parameter fluctuations, robot attitude deviation, weld tracking deviation, etc.), generating a candidate list of defect causes sorted from high to low confidence. This list is then displayed in a list format on the interface to help quickly identify the most likely defect cause and improve the efficiency and objectivity of quality problem attribution.

[0104] Step 408: In response to the statistical analysis command, perform cross-sample statistics on the data records of multiple target welds in the database, establish a quantitative relationship between the quality result data and the process parameter data including the process parameter set value and the process parameter execution value under different welding conditions, and determine the optimized process based on the quantitative relationship.

[0105] The statistical analysis module responds to user-initiated statistical analysis commands and retrieves weld data records that meet the specified multi-dimensional filtering conditions (including workpiece batch number, weld geometry type, welding spatial posture, process parameter range, and defect type, etc.) from the database in batches. It quantitatively calculates the deviation distribution between process parameter set values ​​and executed values, defect incidence rate, rework rate, and process capability index under different process parameter ranges, and generates statistical reports in the form of process parameter distribution histograms and defect Pareto charts.

[0106] Based on this, the statistical analysis module identifies the process parameters that significantly affect welding quality and their reasonable value ranges by comparing and analyzing the quantitative relationship between quality results and process parameters under multiple working conditions. Based on this, the optimization direction is determined, such as adjusting the current setting range under specific weld types or optimizing the welding speed parameters under specific spatial postures. The optimized process parameter combination is then used as the initial value recommendation output for the process parameters of new workpieces under similar working conditions, forming a closed-loop mechanism from data accumulation to continuous process optimization.

[0107] As can be seen from the above technical solutions, the beneficial effects of this embodiment are: establishing a data model for the welding process; the data model uses workpiece identification and weld identification as a joint index, and writes welding program version information and coordinate system calibration information together with workpiece identification and weld identification into the database; the data model specifies the associated storage format for robot motion data, process parameter data, sensor feature data, and quality result data; during the welding process of the weld to be welded, multi-source real-time data is continuously collected with a unified time reference, and each collected multi-source real-time data is appended with a data timestamp and the corresponding workpiece identification and weld identification; during the welding of the target workpiece... Upon completion, acquire the quality result data corresponding to the target workpiece; the quality result data includes defect type, defect location, and defect level; associate the quality result data with the collected multi-source real-time data according to the corresponding workpiece identifier and weld identifier; construct a data record containing complete welding lifecycle information by using a single weld seam as the smallest data organization unit, and store it in the database, according to the data format defined by the data model; in response to the playback command of the target weld seam, load the three-dimensional virtual model corresponding to the target weld seam; the three-dimensional virtual model includes the geometric model of the workpiece to be welded and serves as a spatial comparison. The system establishes a baseline weld space path and aligns the virtual coordinate system in the 3D virtual model with the physical coordinate system of the welding scene based on coordinate system calibration information. It reads robot motion data from data records in chronological order, driving the virtual robot in the 3D virtual model to reproduce the pose changes of each joint and the motion trajectory of the tool center point during the welding process. It also reads corresponding process parameter data and sensor feature data from data records and displays them synchronously with the virtual robot's motion process on the display interface as process parameter execution waveforms and sensor feature curves. During playback, in response to triggering operations on quality inspection marks extracted from quality result data on the display interface, it calculates the welding time period corresponding to the quality inspection mark based on the arc length mapping relationship between the defect location and the weld path, and automatically positions the playback timeline to the start time of the welding time period to display the robot motion data, process parameter data, and sensor feature data within the welding time period. In response to statistical analysis commands, it performs cross-sample statistics on data records of multiple target welds in the database, establishing a quantitative relationship between quality result data and process parameter data (including process parameter set values ​​and process parameter execution values) under different welding conditions, and determines the optimized process based on this quantitative relationship. It has achieved time-series aligned acquisition and full lifecycle associated storage of multi-source heterogeneous welding process data. With the help of the virtual-real fusion playback and automatic defect cause analysis of the digital twin model, the welding quality traceability has been transformed from passive inspection relying on human experience to data-driven proactive closed-loop management, which has significantly improved the accuracy and efficiency of weld-level quality traceability. Furthermore, it has continuously accumulated process optimization knowledge through cross-sample statistical analysis.

[0108] This application also proposes a readable medium storing execution instructions. When these instructions are executed by a processor of an electronic device, the device can perform a welding process quality traceability and playback method based on digital twins provided in any embodiment of this application, specifically for executing, for example... Figure 4 The method shown.

[0109] The electronic devices in the foregoing embodiments may be computers.

[0110] Those skilled in the art will understand that the embodiments of this application can be provided as methods or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or a combination of software and hardware.

[0111] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0112] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0113] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A welding process quality traceability and playback system based on digital twins, characterized in that, The system is applied to the welding process of industrial robots, and the system includes: The module includes a physical welding module, a quality inspection module, a digital twin modeling module, a data management module, a virtual-real fusion playback module, and a statistical analysis module. The physical welding module includes the industrial robot and its corresponding robot controller, the welding power supply and its corresponding power control unit, and a process sensing unit; it is used for the welding operation of the industrial robot and generates multi-source real-time data reflecting the welding process; the multi-source real-time data includes robot motion data, process parameter data, and sensor feature data; the process parameter data includes process parameter set values ​​and process parameter executed values. The quality inspection module is connected to the physical welding module in the process and is used to perform quality inspection on the target workpiece after welding, and output quality result data including defect type, defect location and defect level. The digital twin modeling module is used to construct and store a three-dimensional virtual model of the welding scene. The three-dimensional virtual model includes the linkage kinematic model of the industrial robot, the geometric model of the workpiece to be welded, and the weld space path corresponding to the weld seam on the workpiece to be welded, which is generated offline. The data management module is communicatively connected to the physical welding module and the quality inspection module, respectively. It is used to configure a unique workpiece identifier and weld identifier for each workpiece to be welded and each corresponding weld seam to be welded. Using the workpiece identifier and weld identifier as a joint index, it associates and stores the multi-source real-time data, the quality result data, and the corresponding welding program version information and coordinate system calibration information to construct a data record with a single weld seam to be welded as the smallest data organization unit. The virtual-real fusion playback module is connected to the digital twin modeling module and the data management module, respectively. In response to a playback command for the target weld, it loads the corresponding 3D virtual model from the digital twin modeling module and reads the data record associated with the identifier of the target weld from the data management module. Based on the time-series data in the data record, it drives the 3D virtual model to reproduce the welding process. On the display interface, it overlays and displays the process parameter execution waveform generated from the process parameter execution values, the quality result data, and the sensor feature curve generated from the sensor feature data with the motion process of the 3D virtual model in a spatiotemporally synchronized manner. The statistical analysis module, connected to the data management module, is used to receive statistical conditions, retrieve and statistically analyze the data records of multiple target welds, and output the correlation analysis results between the process parameter setting values, the process parameter execution values ​​and the quality result data under different working conditions.

2. The system according to claim 1, characterized in that, The data management module includes: Acquisition unit, identification unit, and storage unit; The acquisition unit is used to acquire the multi-source real-time data from the physical welding module at a unified time reference during the welding process, and to add a data timestamp to the multi-source real-time data. The identification unit is used to configure a unique workpiece identifier and a weld identifier for each workpiece to be welded and the corresponding weld seam to be welded; the workpiece identifier is generated by combining the production order number, production date and serial number, and the weld seam identifier is generated according to the workpiece structural part and the weld seam sequence number. The storage unit is used to associate and store the multi-source real-time data, the quality result data, the welding program version information, and the coordinate system calibration information using the workpiece identifier and the weld identifier as a joint index, and to construct a data record with a single weld seam to be welded as the smallest data organization unit.

3. The system according to claim 2, characterized in that, The acquisition unit is connected to the robot controller and the power control unit via industrial Ethernet or fieldbus, and uses a precise time synchronization protocol to calibrate the local clocks of each data source corresponding to the multi-source real-time data.

4. The system according to claim 1, characterized in that, The virtual-real fusion playback module includes: Model loading unit, data reading unit, synchronization driving unit, and visualization overlay unit; The model loading unit is used to load the three-dimensional virtual model corresponding to the target weld from the digital twin modeling module in response to the playback command; and restore the transformation relationship between the robot base coordinate system, workpiece coordinate system and tooling coordinate system according to the coordinate system calibration information, so as to align the virtual coordinate system in the three-dimensional virtual model with the physical coordinate system space of the welding scene. The data reading unit is used to read the data record associated with the target weld identifier from the data management module; The synchronous drive unit is used to drive the virtual robot in the three-dimensional virtual model in chronological order according to the time series data, so as to reproduce the pose changes of each joint of the robot and the motion trajectory of the tool center point during the welding process. The visualization overlay unit is used to display, in conjunction with the synchronous drive unit, the process parameter execution waveform, the sensing characteristic curve, and the quality detection mark extracted from the quality result data to identify the defect location, in a spatiotemporally synchronized manner with the motion process of the three-dimensional virtual model on the display interface.

5. The system according to claim 1, characterized in that, The process sensing unit includes a line laser profile sensor and a signal processing subunit. The signal processing subunit is used to filter and extract features from the laser cross-section profile data collected frame by frame by the line laser profile sensor, and generate a weld center deviation curve and a height deviation curve with the arc length coordinate in the weld length direction as the horizontal axis, which constitute the weld geometric deviation component in the sensing feature data. The signal processing subunit outputs the weld geometric deviation component to the data management module, and the virtual-real fusion playback module maps the weld geometric deviation component to the weld path of the three-dimensional virtual model in a three-dimensional spatial superposition manner during playback.

6. The system according to claim 1, characterized in that, The statistical analysis module is also used to support multi-dimensional condition combination filtering based on workpiece batch number, weld geometry type, welding spatial posture, parameter range corresponding to process parameter setting value, and defect type. The screening results output the defect incidence rate, rework rate and process capability index, and generate a statistical report in the form of process parameter distribution histogram and defect Pareto chart. The defect incidence rate, the rework rate, and the process capability index constitute the quality indicators derived from the statistical analysis of the quality result data. Based on historical statistical results, and based on the feature matching of the weld geometry type and working conditions, initial values ​​of process parameters are recommended for new imported workpieces with similar weld types and geometric features.

7. A method for quality traceability and playback of welding processes based on digital twins, characterized in that, The method is applied to the system according to any one of claims 1-6, and the method includes: A data model for the welding process is established; the data model uses the workpiece identifier and weld identifier as a joint index, and writes the welding program version information and coordinate system calibration information together with the workpiece identifier and weld identifier into the database; the data model specifies the associated storage format of robot motion data, process parameter data, sensor feature data and quality result data; During the welding process of the weld seam to be welded, multi-source real-time data is continuously collected with a unified time reference, and a data timestamp, the corresponding workpiece identifier, and the weld seam identifier are added to each collected multi-source real-time data. After the target workpiece is welded, the quality result data corresponding to the target workpiece is obtained; the quality result data includes defect type, defect location and defect level; the quality result data is associated with the collected multi-source real-time data according to the corresponding workpiece identifier and weld identifier; The multi-source real-time data and the quality result data are organized according to the data format defined by the data model, with each weld seam to be welded as the smallest data organization unit, to construct a data record containing complete welding life cycle information, and stored in the database. In response to the playback command of the target weld, a three-dimensional virtual model corresponding to the target weld is loaded; the three-dimensional virtual model includes the geometric model of the workpiece to be welded and the weld space path as a spatial comparison reference; and according to the coordinate system calibration information, the virtual coordinate system in the three-dimensional virtual model is spatially aligned with the physical coordinate system of the welding scene. The robot motion data is read from the data record in chronological order to drive the virtual robot in the three-dimensional virtual model to reproduce the pose changes of each joint of the robot and the motion trajectory of the tool center point during the welding process; the corresponding process parameter data and sensor feature data are read from the data record and displayed synchronously with the motion process of the virtual robot on the display interface in the form of process parameter execution waveform and sensor feature curve; During playback, in response to the triggering operation of the quality inspection mark extracted from the quality result data on the display interface, the welding time period corresponding to the quality inspection mark is calculated according to the arc length mapping relationship between the defect location and the weld path, and the playback time axis is automatically positioned to the start time of the welding time period to display the robot motion data, the process parameter data and the sensing feature data within the welding time period; In response to a statistical analysis command, cross-sample statistics are performed on the data records of multiple target welds in the database to establish a quantitative relationship between the quality result data and the process parameter data, including the process parameter set value and the process parameter execution value, under different welding conditions, and to determine the optimized process based on the quantitative relationship.

8. The method according to claim 7, characterized in that, During the welding process of the weld seam to be welded, the continuous acquisition of multi-source real-time data at a unified time reference includes: Buffer queues are established for the multi-source real-time data collected from different data sources; Set the time synchronization window length, and use the timestamp sequence of the data source with the lowest sampling frequency as the reference time axis; Within the time synchronization window, the high-frequency data source is downsampled, and the time point closest to the reference time axis is selected for alignment. For data packet loss or timeouts that occur during the acquisition process, linear interpolation compensation is used; when the duration of continuous packet loss exceeds a preset threshold, previous value hold compensation is used, and a quality flag bit is added.

9. The method according to claim 7, characterized in that, The trigger operation in response to the quality inspection mark extracted from the quality result data on the display interface calculates the welding time period corresponding to the quality inspection mark based on the arc length mapping relationship between the defect location and the weld path, and automatically positions the playback time axis to the start time of the welding time period to display the robot motion data, process parameter data, and sensor feature data within the welding time period, including: Based on the correspondence between the arc length coordinates of the weld space path and the data timestamp, the defect location corresponding to the quality inspection mark is converted into a defect time period; A dynamic time warping algorithm is used to compare the waveform of the process parameters executed during the defect period with a pre-stored normal waveform template to identify parameter fluctuation patterns that deviate from the normal range. The attitude angle change sequence during the defect period is compared with the design attitude corresponding to the weld space path to calculate the attitude deviation. Based on the matching degree between the parameter fluctuation pattern and historical defect samples, the magnitude of the attitude deviation, and the degree of anomaly of the sensing feature data, a weighted fusion algorithm is used to calculate the confidence level of each candidate cause, generate a candidate list of defect causes sorted by confidence level, and present it in the display interface.

10. A computer-readable storage medium storing a computer program, characterized in that, The computer program is used to execute the welding process quality traceability and playback method based on digital twin as described in any one of claims 7-9.

Images