Data acquisition system, method, electronic device, medium and product for manual welding torch
Patent Information
- Application Number
- CN202611469312.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-20
- Publication Date
- 2026-11-03
AI Technical Summary
Existing technologies struggle to obtain data on the molten pool morphology during manual welding, resulting in poor training performance for robotic welding torches.
A vision sensor, an interaction module, and a data processing module are integrated into a manual welding torch to collect images of the molten pool and process parameters in real time. The data is then bound and stored using a unified timestamp and task tag to form a complete dataset.
This improved the integrity and consistency of training data for the robot welding model, thereby enhancing the training effect.
Smart Images
Figure CN123024974A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of manual welding torch technology, specifically to data acquisition systems, methods, electronic devices, media, and products for manual welding torches. Background Technology
[0002] Welding robots for GMAW (Gas Metal Arc Welding), especially those using the Vision-Language-Action Model (VLA), generally follow an imitation learning approach: a human demonstrates the welding with a manual welding torch, and the robot then replicates the actions in the same scenario. This approach requires a large amount of multimodal data (molten pool image, tool posture, process parameters) from manual welding demonstrations as training samples. Traditional manual welding torches serve only as welding execution tools, capable of performing the welding task itself, without any sensing or data recording capabilities. Robotic welding torches, on the other hand, focus on solving the deployment-side engineering problem of "how to stably mount the welding torch onto the robot and accurately reproduce the predetermined trajectory." The design goal of robotic welding torches is to accurately execute the defined welding task, rather than collecting sensory data during the welding process.
[0003] In related technologies, the method of training robotic welding guns involves programmers dragging the robotic welding gun point by point with a handheld teach pendant to record its joint poses and motion trajectories, and then having the robotic welding gun reproduce the same behavior, or completely simulating it offline in a virtual environment. Both of these methods record the robot's motion trajectory, but it is difficult to obtain data on the molten pool morphology during manual welding, as well as process data related to the electric arc and welding wire, resulting in poor training effects for robotic welding guns. Summary of the Invention
[0004] This invention provides a data acquisition system, method, electronic device, medium, and product for manual welding torches, in order to solve the problem in related technologies that it is difficult to obtain molten pool morphology data during manual welding, resulting in poor training effect for robotic welding torches.
[0005] In a first aspect, the present invention provides a data acquisition system for a manual welding torch, comprising: a manual welding torch, a vision sensor, an interaction module, and a data processing module; the vision sensor is fixed on the manual welding torch by a mounting structure, and the optical axis of the vision sensor is aligned with the molten pool area; the interaction module is mounted on the manual welding torch, and the data processing module is configured in the manual welding torch; the vision sensor is used to acquire molten pool images of the molten pool area in response to welding task instructions input by the user according to the interaction module; the data processing module is used to create a task package in response to welding task instructions input by the user according to the interaction module, and to acquire process parameters during the execution of the welding task instructions, store the molten pool images and process parameters in the task package, so as to train a robot welding model based on the molten pool images and process parameters.
[0006] The data acquisition system for a manual welding torch of this invention includes a manual welding torch, a vision sensor, an interaction module, and a data processing module. Integrating the vision sensor, interaction module, and data processing module into the manual welding torch forms a complete welding data acquisition system capable of simultaneously acquiring molten pool images and process parameters at the manual welding work site. The vision sensor is fixed to the manual welding torch based on an installation structure, with its optical axis precisely aligned with the molten pool area, enabling direct and clear acquisition of real-time images of the weld pool from a fixed shooting angle and a stable field of view. Both the interaction module and the data processing module are directly mounted on the manual welding torch body, allowing users to perform tasks at any time without leaving their workstation. Furthermore, the compact layout of each module allows for easy portability and use at the welding site. The vision sensor, responding to welding task commands input by the user according to the interaction module, acquires images of the molten pool area. The vision sensor possesses anti-arc interference capabilities and can stably capture the molten pool morphology under arc light, achieving molten pool image acquisition by the manual welding torch. The data processing module is used to respond to welding task instructions input by the user based on the interaction module, create task packages, and collect process parameters during the execution of welding task instructions. It stores the molten pool image and process parameters into the task package to train the robot welding model based on the molten pool image and process parameters. The data processing module automatically creates task packages after the welding task instruction is issued and synchronously collects process parameters during task execution, storing the molten pool image and process parameters together into the corresponding task package. This binds and archives image information and process parameter information in terms of time and content, forming a complete, searchable, and reusable dataset. This allows the robot welding model to be trained with high quality based on the corresponding molten pool images and process parameters within the same task package, significantly improving the integrity, consistency, and training effect of the training data.
[0007] In one optional implementation, the data processing module includes a metadata and time synchronization submodule; the metadata and time synchronization submodule is used to configure a unified timestamp for the melt pool image with the local clock as the time reference; the metadata and time synchronization submodule is also used to set corresponding task tags for the melt pool image configured with the unified timestamp and process parameters; the metadata and time synchronization submodule is also used to mark defect events in response to user event marking instructions.
[0008] The data acquisition system for manual welding torches of this invention includes a metadata and time synchronization submodule that uses a local clock as the time reference to configure a unified timestamp for the molten pool image. This ensures that molten pool images from different times and sources remain consistent and aligned in the time dimension. The metadata and time synchronization submodule is also used to set corresponding task tags for the molten pool images configured with unified timestamps and process parameters. This allows data from the same task to be quickly located and distinguished during subsequent retrieval, classification, and use. Furthermore, it is used to mark defect events in response to user event marking commands, enabling users to immediately record any defects discovered during the welding operation. This accurately associates the time of defect occurrence with the corresponding molten pool image and process parameters.
[0009] In one alternative implementation, the interaction module is a physical button and / or a touch interface, which is used to indicate the equipment status and to control the welding process.
[0010] In one alternative implementation, the data acquisition system for the manual welding torch further includes a voice recording module; the voice recording module, configured on the manual welding torch, is used to record the welding task execution process via voice.
[0011] The data acquisition system for the manual welding torch of the present invention adds a voice recording module and is configured on the manual welding torch. It can record voice synchronously during the execution of welding tasks, and retain the user's dictated information along with the task, thus supplementing the molten pool image and process parameters with rich contextual information.
[0012] In one alternative implementation, the data acquisition system for the manual welding torch further includes a power supply module; the power supply module is used to power the vision sensor and the data processing module.
[0013] In one alternative implementation, the manual welding torch data acquisition system further includes a data interface module; the data interface module is used to export task packages to a preset server so that the preset server can train a robot welding model based on the molten pool image and process parameters.
[0014] The data acquisition system for manual welding torches of the present invention has a data interface module that can export task packages to a preset server, enabling the server to train robot welding models offline and centrally based on the molten pool images and process parameters therein. This not only reduces the data processing burden on the manual welding torch end, but also enables the training data to be uniformly aggregated and utilized on a large scale, thereby improving the overall training efficiency.
[0015] In one optional implementation, the data acquisition system for the manual welding torch further includes an attitude acquisition module; the attitude acquisition module is used to determine the relative motion information of the manual welding torch based on the molten pool image of the manual welding torch; the relative motion information is used to characterize the displacement and attitude changes of the manual welding torch; the attitude acquisition module is also used to send the relative motion information to the data processing module so that the data processing module stores the relative motion information in the task package.
[0016] In one optional implementation, the data acquisition system of the manual welding torch further includes an acoustic sensor; the acoustic sensor is configured on the manual welding torch and is used to acquire the arc sound signal during the welding process and send the arc sound signal to the data processing module so that the data processing module stores the arc sound signal in the task package.
[0017] In one alternative implementation, the data acquisition system of the manual welding torch further includes one or more of a temperature sensor, a light intensity sensor, and a gas sensor.
[0018] Secondly, the present invention provides a data acquisition method for a manual welding torch, comprising: in response to a welding task instruction input by a user, acquiring a molten pool image of the molten pool area; in response to a welding task instruction input by a user, creating a task package, acquiring process parameters during the execution of the welding task instruction, and storing the molten pool image and process parameters in the task package, so as to train a robot welding model based on the molten pool image and process parameters.
[0019] The data acquisition method for manual welding torches of the present invention achieves an automated closed loop from on-site data acquisition to robot welding model training by responding to welding task instructions to acquiring molten pool images, creating task packages, collecting and uniformly storing process parameters during execution, and finally training a robot welding model based on images and parameters. It binds and archives image information and process parameter information in terms of time and content, forming a complete, searchable, and reusable dataset. This enables the robot welding model to perform high-quality training based on corresponding molten pool images and process parameters within the same task package, significantly improving the completeness, consistency, and training effect of the training data.
[0020] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the data acquisition method for the manual welding torch described in the second aspect.
[0021] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the data acquisition method for a manual welding torch described in the second aspect.
[0022] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the data acquisition method for a manual welding torch described in the second aspect. Attached Figure Description
[0023] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0024] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the first working process of a data acquisition system for a manual welding torch according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the visual sensor mounting structure according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the interaction module according to an embodiment of the present invention; Figure 5 This is a schematic diagram of a second workflow of a data acquisition system for a manual welding torch according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the data acquisition and control process according to an embodiment of the present invention; Figure 7 This is a flowchart illustrating a data acquisition method using a manual welding torch according to an embodiment of the present invention. Figure 8 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation
[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0026] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.
[0027] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.
[0028] As an optional application scenario of this invention, such as Figure 1 As shown, the data acquisition system of the manual welding torch may include a manual welding torch 101, a vision sensor 102, an interaction module 103, and a data processing module 104. The user holds the manual welding torch 101 to perform welding tasks. The vision sensor 102 is used to acquire images of the molten pool during the welding process. The interaction module 103 is used to provide the user with an interactive method so that the user can control the data acquisition system of the manual welding torch 101. The data processing module 104 is used to acquire process parameters and store the molten pool images and process parameters.
[0029] In related technologies, there are several problems with methods for training welding gun robots. First, data acquisition during welding is difficult. While data on the working conditions outside the arc-starting stage is easy to collect, what is truly lacking is the molten pool morphology data collected synchronously during manual welding, as well as process data related to the arc and welding wire. Due to the lack of this data, the welder's adjustments to techniques and operational decisions for different molten pool states cannot be recorded synchronously, making it difficult to construct an end-to-end mapping of "visual perception - action execution" in the welding process. Therefore, it cannot meet the current training requirements of vision-language-action models for high-quality demonstration samples. Second, experience is difficult to structure. The "feel / experience" of skilled welders in arc starting, arc ending, track changing, and defect judgment cannot be structured and retained in a timestamp format, making it difficult to reuse for robot strategy training or enterprise process archiving. Finally, integration is difficult. Adding observation equipment to the handheld welding gun can easily obstruct the welder's view, disrupt the center of gravity, and affect the grip. Moreover, there is often no network / host computer at the welding site, making real-time recording and local storage difficult.
[0030] This invention provides a data acquisition system for a manual welding torch. By configuring a vision sensor on the manual welding torch, the system collects process data of the manual welding torch performing welding tasks, thereby improving the training effect of the robot welding model.
[0031] According to an embodiment of the present invention, a data acquisition system for a manual welding torch is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.
[0032] This embodiment provides a data acquisition system for a manual welding torch. Figure 2 This is a first workflow diagram of a data acquisition system for a manual welding torch according to an embodiment of the present invention, such as... Figure 2 As shown, the system includes: a manual welding torch 201, a vision sensor 202, an interaction module 203, and a data processing module 204.
[0033] Among them, the manual welding gun 201 is a hand-held arc welding tool, which consists of a handle, gun neck / bend, conductive nozzle, nozzle, gas distributor, trigger switch, cable bundle, and interface for connecting to the welding machine. When performing welding tasks, the user holds the manual welding gun 201 to perform the welding task. The user can be a professional welder or welder. The manual welding gun 201 can also be installed on a robot through a specific fixture so that the robot can control the manual welding gun 201 to play back and confirm the recorded motion data and perform welding tasks.
[0034] The vision sensor 202 is an industrial camera resistant to arc light interference. It can be used as a molten pool camera, suppressing the arc light background through narrow-band filtering, and outputting the original molten pool image through the camera interface.
[0035] The interaction module 203 is used to provide users with a human-computer interaction method for starting and stopping recording and marking operating conditions.
[0036] The data processing module 204 is a computing processing unit responsible for receiving, organizing, storing and exporting data. It can be an embedded edge computing unit responsible for receiving camera data, timestamp marking, lightweight preprocessing and local disk writing, including local storage media.
[0037] The vision sensor 202 is fixed on the manual welding torch 201 by the mounting structure, and the optical axis of the vision sensor 202 is aligned with the molten pool area; the interaction module 203 is mounted on the manual welding torch 201, and the data processing module 204 is configured in the manual welding torch 201.
[0038] For example, such as Figure 3 The diagram shown is a schematic of the visual sensor mounting structure. Figure 3In this design, the vision sensor 302 is fixed between the neck and the conductive nozzle of the manual welding torch 301 via a mounting structure 303. The mounting structure 303 is a mechanical connector used to fix the vision sensor 302 to the manual welding torch 301, and can be a bracket, clamp, flange, or other structure. The vision sensor 302 is fixed to the front end of the neck, with its optical axis pointing towards the molten pool area in front of the conductive nozzle. The neck is the bent tube / neck section connecting the torch body and the conductive nozzle. The conductive nozzle is the metal tip at the end of the torch that supplies power to and guides the welding wire. The optical axis is the axis of the light ray from the center of the camera lens. The molten pool area is the spatial range where the welding wire and the base material melt to form a liquid molten pool. The vision sensor 302 maintains an unobstructed geometric relationship with the nozzle. The mounting structure is lightweight, has a small center of gravity offset, does not obstruct the user's view, and does not affect the grip and feel of the manual welding torch 301. The mounting structure 303 can be a ring clamp fitted onto the outside of the neck and secured with screws to fix the camera bracket. Alternatively, the mounting structure can be a metal bracket and bolts welded to the neck. The camera's optical axis should be aligned with the gun's neck axis at a 45°-60° angle to the molten pool area to avoid obstructing the welding wire.
[0039] The interaction module 203 is located on the side of the gun handle within thumb reach (button / touch area). The data processing module 204 is encapsulated in the internal cavity of the gun handle, which is equipped with heat dissipation holes and a dust cover.
[0040] The vision sensor 202 is used to acquire images of the molten pool area in response to welding task instructions input by the user according to the interaction module 203.
[0041] Among them, the welding task instructions are the instructions issued by the user to start, end, and select the welding mode for the welding task; the molten pool image is the sequence of molten pool images / video frames captured by the vision sensor 202.
[0042] For example, when the user turns on the power switch or connects an external power source, the vision sensor 202 is activated and enumerated, local storage is mounted, and after a self-test, the status indicator light on the manual welding torch 201 turns to "ready," and the screen displays the capacity and clock. The user selects a welding task and issues a start recording command through the interaction module 203. For example, the user selects / creates a welding task number by pressing a button and presses the start recording button to create a task package (including an image directory, process parameter file, relative motion information file, arc sound signal file, and sensor data file, etc.), enters the recording state, the indicator light changes, and the screen displays the real-time recording duration.
[0043] In another possible implementation, after the user presses the start button, the user holds the manual welding torch 201 to perform the welding task, and the vision sensor 202 begins recording the molten pool video at a set frame rate (e.g., 30fps / 60fps). After the user clicks to create a new task via the touchscreen of the interaction module 203, the vision sensor 202 enters a continuous shooting mode (e.g., capturing N frames per second). The vision sensor 202 only acquires data while the welding task command is valid, and automatically stops when the task ends to avoid invalid data. The command triggers the high dynamic range or infrared / thermal imaging mode of the vision sensor 202 to adapt to strong arc light environments.
[0044] The data processing module 204 is used to respond to the welding task instructions input by the user according to the interaction module 203, create a task package, and collect process parameters during the execution of the welding task instructions, store the molten pool image and process parameters into the task package, so as to train the robot welding model according to the molten pool image and process parameters.
[0045] The task package is a data container corresponding to a welding task, packaging the molten pool image and process parameters of that task into a unified dataset. The task package includes an image directory, process parameter files, relative motion information files, arc acoustic signal files, and sensor data files. The image directory stores the molten pool image, and the process parameter files store the process parameters. Process parameters are measurable parameters of the welding process, such as current, voltage, wire feed speed, welding speed, gas flow rate, and temperature. The relative motion information files store the relative motion information determined by the attitude acquisition module, the arc acoustic signal files store the arc acoustic signals during the welding process, and the sensor data files store multi-source sensor data acquired by additional sensors. The execution period of the welding task instruction is a time window from the start to the end of the welding task instruction. The start time of the welding task instruction can be the moment the user presses the start button, and the end time can be the moment the user presses the end button. The robot welding model is an automated model trained based on the acquired data to achieve automatic welding.
[0046] In some optional implementations, the data processing module 204 is used to write the molten pool image and process parameters into a data packet for local storage, producing a structured welding process multimodal dataset organized by task, namely an image sequence or video and a corresponding process parameter file.
[0047] For example, the data processing module 204 is used to automatically generate folders named with task numbers and timestamps as task packages in response to welding task instructions input by the user according to the interaction module 203. Molten pool images and process parameters are stored in separate files and indexed. Process parameters are read in real time by the welding power source / welding machine through a data acquisition interface and written into the task package synchronously with the molten pool images. A database is used to record task package metadata; images are stored in files, and parameters are stored in tables, linked by timestamps. Parameters within the task package are aligned with image frames according to fixed time windows, forming frame-by-frame corresponding training sample pairs. During robot welding model training, task packages are input in batches into the robot welding model.
[0048] In some optional implementations, the data processing module 204 includes a metadata and time synchronization submodule 2041; the metadata and time synchronization submodule 2041 is used to configure a unified timestamp for the melt pool image with the local clock as the time reference; the metadata and time synchronization submodule 2041 is also used to set corresponding task tags for the melt pool image configured with the unified timestamp and process parameters; the metadata and time synchronization submodule 2041 is also used to mark defect events in response to user event marking instructions.
[0049] Among them, the local clock is the event reference built into the metadata and time synchronization submodule 2041, such as the RTC (Real-Time Clock); the unified timestamp is to assign a unified and alignable time stamp to each frame of the molten pool image; the task tag is a tag that identifies the task to which the data belongs, including user identifier, process number, task identifier, etc.; the event tag instruction is the marking instruction issued by the user when an anomaly is found; the defect event tag is the mark of the defect occurrence point marked on the molten pool image and process parameters.
[0050] Specifically, the metadata and time synchronization submodule 2041 is used to maintain a unified time base with a local clock, stamp each frame of molten pool image with a unified timestamp, and write task tags (user identifier, workpiece / process number), arc and welding wire related process parameters, and event markers (arc start / arc end / defect, the event content is recorded in text or voice, voice recording requires a microphone to be configured on the device; supports real-time marking during welding or video playback and supplementation of marking after recording), and stores them in pairs with the image sequence using metadata files.
[0051] For example, welders can instantly mark events during the welding process by pressing the "event mark" key at points such as arc initiation, arc termination, track change, and defect occurrence. They can also review the video after recording and add additional marks for key events. The data acquisition system for manual welding torches also includes a voice recording module 205. The voice recording module 205 is configured on the manual welding torch 201 and is used to record the welding task execution process by voice. During the defect event marking process, the defect event content is recorded in text or voice. Both methods insert event tags with timestamps into the metadata, so that the dataset comes with manual working condition annotations.
[0052] The metadata and time synchronization submodule 2041 is also used for optional lightweight preprocessing of the melt pool image.
[0053] The data acquisition system for manual welding torches in this embodiment of the invention uses a local clock as the time reference to configure a unified timestamp for the molten pool image, ensuring that molten pool images from different times and sources remain consistent and aligned in the time dimension. The metadata and time synchronization submodule is also used to set corresponding task tags for the molten pool images configured with unified timestamps and process parameters, enabling data from the same task to be quickly located and distinguished during subsequent retrieval, classification, and use. It is also used to mark defect events in response to user event marking instructions, allowing users to immediately mark and record any defects discovered during the welding operation, thereby accurately associating the time of defect occurrence with the corresponding molten pool image and process parameters.
[0054] In some alternative implementations, the interaction module 203 is a physical button and / or a touch interface, which is used to indicate the equipment status and to control the welding process.
[0055] The interaction module 203 can be a separate physical button, a separate touch interface, or a combination of a physical button and a touch interface. The physical button is a physical mechanical button, including start, stop, and marker buttons, used to control the welding process. The physical button also includes indicator lights to display the working status of the equipment. The touch interface is a display operation interface with a touch screen, including a display screen, recording status, remaining capacity, recording duration, and playback of recorded video data. The display screen is an LCD / OLED screen that displays text / graphics. The recording status indicates whether recording is currently in progress. The remaining capacity indicates the remaining storage space. The recording duration indicates the recorded time. The playback of recorded video data is the real-time recorded video, which is convenient for on-site inspection of data quality.
[0056] The interaction module 203 can use either physical buttons or a touch interface, and is equipped with status indicator lights and a display screen to show the recording status, remaining capacity, and recording duration. When using physical buttons, users can complete the operation without looking at the screen. Operating events can be recorded in text or voice, and a microphone is configured on the device for voice recording.
[0057] For example, Figure 4 This is a schematic diagram of an interactive module according to an embodiment of the present invention. The entire interactive module includes a display screen, which displays the recording status (standby / recording), remaining capacity, and recording duration. The interactive module also includes a start / stop recording control, an event marker (arc start / arc end / defect) control, and a status indicator (ready / recording / abnormal). Each control can be a physical button or a touch control on a touch screen.
[0058] In some alternative implementations, the manual welding torch data acquisition system also includes a power supply module 206; the power supply module 206 is used to power the vision sensor 202 and the data processing module 204.
[0059] The power supply module 206 is used to power the vision sensor 202 and the data processing module 204. It can be powered by a rechargeable battery pack or by an external power source, and includes over-discharge protection and multiple voltage regulation outputs.
[0060] For example, the power supply module 206 can be a built-in rechargeable battery pack and charging management circuit, and adopts a removable / replaceable battery design to support quick on-site replacement to extend battery life. The power supply module 206 includes overcurrent, overvoltage, and reverse connection protection circuits to ensure stable power supply to the vision sensor 202 and the data processing module 204.
[0061] In some alternative implementations, the manual welding torch data acquisition system also includes a data interface module 207; the data interface module 207 is used to export the task package to a preset server so that the preset server can train the robot welding model based on the molten pool image and process parameters.
[0062] The data interface module 207 provides wired or wireless export channels to import task packages (image sequences and metadata) into a preset server (data storage server) and connect to the welding robot training pipeline; it supports breakpoint resume and task package verification.
[0063] After welding is completed, press the "Stop Recording" button to write the seal and generate a task package with verification. The indicator light changes. The task package is exported to the data storage server via the data interface module 207, and the welding robot enters the VLA imitation learning training process.
[0064] The data acquisition system for the manual welding torch in this embodiment of the invention has a data interface module that can export task packages to a preset server, enabling the server to train the robot welding model offline and centrally based on the molten pool image and process parameters therein. This not only reduces the data processing burden on the manual welding torch end, but also enables the training data to be uniformly aggregated and utilized on a large scale, thereby improving the overall training efficiency.
[0065] In some optional implementations, the data acquisition system for the manual welding torch further includes an attitude acquisition module 208; the attitude acquisition module 208 is used to determine the relative motion information of the manual welding torch based on the molten pool image of the manual welding torch; the relative motion information is used to characterize the displacement and attitude changes of the manual welding torch; the attitude acquisition module 208 is also used to send the relative motion information to the data processing module so that the data processing module stores the relative motion information in the task package.
[0066] The attitude acquisition module 208 determines the relative motion information of the manual welding torch based on the molten pool image of the manual welding torch using the odometer. The odometer adopts a visual odometer method, which calculates the relative motion of the manual welding torch based solely on the molten pool image through feature extraction and inter-frame matching to obtain the relative motion information. This method is suitable for scenarios with sufficient environmental features in the field of view of the visual sensor.
[0067] The odometer calculation unit can also employ visual or inertial odometer methods, fusing inertial measurement data acquired by the inertial measurement unit integrated into the visual sensor with the molten pool image collected by the visual sensor. During the welding process, when arc light or fumes cause insufficient visual environmental features, the calculation is primarily performed based on measurements from the inertial measurement unit to ensure continuous acquisition of motion information during welding.
[0068] The mileage calculation unit can be deployed on the end-side device and run in real time during the acquisition process, or it can be run offline after the acquisition is completed to obtain the relative motion information. When running offline, the mileage calculation unit can be deployed on the end-side device or other computing devices (such as data storage servers or host computers), thereby decoupling the computing burden from the welding site acquisition.
[0069] In some optional embodiments, the data acquisition system of the manual welding torch further includes an acoustic sensor 209; the acoustic sensor 209 is disposed on the manual welding torch 201, and the acoustic sensor 209 is used to collect the arc sound signal during the welding process and send the arc sound signal to the data processing module 204 so that the data processing module 204 stores the arc sound signal in the task package.
[0070] The acoustic sensor 209 is positioned on the manual welding torch 201 near the vision sensor 202 to collect arc sound signals during the welding process. The arc sound signals reflect the combustion stability of the welding arc and the state of the molten pool. The data processing module 204 uses the local clock as a unified time reference to align the arc sound signals with the molten pool image and process parameters within the same time window, and writes them into the task package. This allows the task package to simultaneously contain visual images, process parameters, and process acoustic information, forming a more multi-modal demonstration sample with richer dimensions.
[0071] In some alternative implementations, the data acquisition system of the manual welding torch may further include one or more of a temperature sensor, a light intensity sensor, and a gas sensor.
[0072] The types of sensors can be configured according to requirements, and are not limited to temperature sensors, light intensity sensors, and gas sensors. Other additional sensors can also be included. The signals collected by each additional sensor are aligned with the molten pool image and process parameters under a unified time reference and written into the task package. By optionally configuring different combinations of sensor devices, the system can acquire welding process data in a scalable multimodal manner without changing the unified packaging structure and time alignment mechanism of the task package, thereby adapting to different process monitoring needs.
[0073] The data acquisition system for a manual welding torch according to this invention includes a manual welding torch, a vision sensor, an interaction module, and a data processing module. Integrating the vision sensor, interaction module, and data processing module into the manual welding torch constitutes a complete welding data acquisition system, enabling simultaneous acquisition of molten pool images and process parameters at the manual welding work site. The vision sensor is fixed to the manual welding torch based on an installation structure, with its optical axis precisely aligned with the molten pool area, thus enabling direct and clear acquisition of real-time images of the weld pool from a fixed shooting angle and a stable field of view. Both the interaction module and the data processing module are directly mounted on the manual welding torch body, allowing users to perform tasks at any time without leaving their workstation. Furthermore, the compact layout of each module allows for easy portability and use at the welding site. The vision sensor responds to welding task commands input by the user according to the interaction module, acquiring images of the molten pool area. The vision sensor possesses anti-arc interference capabilities, stably capturing the molten pool morphology under arc light, thus realizing molten pool image acquisition by the manual welding torch. The data processing module is used to respond to welding task instructions input by the user based on the interaction module, create task packages, and collect process parameters during the execution of welding task instructions. It stores the molten pool image and process parameters into the task package to train the robot welding model based on the molten pool image and process parameters. The data processing module automatically creates task packages after the welding task instruction is issued and synchronously collects process parameters during task execution, storing the molten pool image and process parameters together into the corresponding task package. This binds and archives image information and process parameter information in terms of time and content, forming a complete, searchable, and reusable dataset. This allows the robot welding model to be trained with high quality based on the corresponding molten pool images and process parameters within the same task package, significantly improving the integrity, consistency, and training effect of the training data.
[0074] This embodiment provides a data acquisition system for a manual welding torch. Figure 5 This is a second workflow diagram of a data acquisition system for a manual welding torch according to an embodiment of the present invention, such as... Figure 5 As shown, the system includes: The components include a manual welding torch body, a vision sensor module, a camera mounting bracket / fixing structure, a portable computer storage device, an interaction module, a data interface module, and a data storage server.
[0075] The vision sensor module is fixed to the front end of the gun neck of the manual welding torch body via a camera mounting bracket / fixing structure. The vision sensor module acquires images of the molten pool and sends them to a portable computer storage device. The metadata and time synchronization submodule in the portable computer storage device is used to timestamp, determine task tags, and mark events at the same clock. The interaction module includes buttons or a touch screen, indicator lights, and a display screen. The data interface module is used to export the task package obtained by the portable computer storage device to the data storage server via wired or wireless connection. The power supply module is a battery or external power source and is used to power the portable computer storage device and the vision sensor module.
[0076] The data acquisition system for the manual welding torch of this invention has a structure in which a single anti-arc light molten pool camera is rigidly integrated at the front end of the manual welding torch neck near the conductive tip / nozzle, and the optical axis of the camera points to the molten pool without obstructing the welder's view. It is equipped with a portable computing and storage device, and simultaneously acquires molten pool images and event metadata with a unified clock and stores them locally. Through physical buttons or a touch interface, the welder can start and stop recording and mark the working conditions in text or voice for human-computer interaction.
[0077] This embodiment provides a data acquisition and control method, which can be used in the data acquisition system of the aforementioned manual welding torch. Figure 6 This is a flowchart of a data acquisition and control method according to an embodiment of the present invention, such as... Figure 6 As shown, the process includes the following steps: Power-on initialization; select task and start recording; capture molten pool during welding (each frame is timestamped); write data locally (image frame + timestamp + process parameters); mark key node events (arc start / arc end / track change / defect), triggered by the welder pressing the event mark key, and insert a timestamped event tag after the operation; determine if the welder has pressed the stop recording key, if not, continue welding (loop acquisition); if so, stop archiving and exporting (task package verification, export to data storage server).
[0078] The data acquisition and control method of this invention features a vision sensor with anti-arc interference capabilities, enabling stable capture of the molten pool morphology under arc light. Acquisition and welding actions are timestamped using a unified clock, meeting the alignment requirements for demonstration deployment observations. Single-device local storage eliminates the need for external storage, allowing independent operation without a host computer and covering continuous welding records for an entire shift. The camera is lightweight, with minimal centroid shift and low line-of-sight obstruction, offering a grip experience largely consistent with the original welding torch. Event marking supports both real-time marking during welding and post-weld review and supplementary marking. Event content can be recorded as text or voice (voice recording requires a microphone), both timestamped using a unified clock, allowing the dataset to inherently include human annotations, providing flexible and highly consistent annotation methods. This provides a structured, multimodal welding process dataset with human demonstration capabilities for direct training of welding robot VLA imitation learning, shortening the data preparation cycle from demonstration to strategy deployment.
[0079] This embodiment provides a data acquisition method for a manual welding torch, which can be used in the aforementioned data acquisition system for a manual welding torch. Figure 7 This is a flowchart of a data acquisition method for a manual welding torch according to an embodiment of the present invention, such as... Figure 7 As shown, the process includes the following steps: Step S701: In response to the welding task command input by the user, acquire the image of the molten pool area.
[0080] Step S702: In response to the welding task instruction input by the user, a task package is created, process parameters during the execution of the welding task instruction are collected, and the molten pool image and process parameters are stored in the task package to train the robot welding model based on the molten pool image and process parameters.
[0081] The manual welding torch data acquisition method of this invention realizes an automated closed loop from on-site data acquisition to robot welding model training by responding to welding task instructions to acquire molten pool images, create task packages, collect and uniformly store process parameters during execution, and finally train the robot welding model based on the images and parameters. It binds and archives image information and process parameter information in terms of time and content to form a complete, searchable, and reusable dataset, so that the robot welding model can be trained with high quality based on the corresponding molten pool images and process parameters in the same task package, which significantly improves the integrity, consistency and training effect of the training data.
[0082] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.
[0083] The following is a detailed reference. Figure 8This diagram illustrates a suitable structural schematic for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 801, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 802 or a program loaded from memory 808 into random access memory (RAM) 803. The RAM 803 also stores various programs and data required for the operation of the electronic device. The processor 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.
[0084] Typically, the following devices can be connected to I / O interface 805: input devices 806 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 807 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 808 including, for example, magnetic tapes, hard disks, etc.; and communication devices 809. Communication device 809 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 8 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.
[0085] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 809, or installed from a memory 808, or installed from a ROM 802. When the computer program is executed by the processor 801, it performs the functions defined in the data acquisition method for a manual welding torch according to embodiments of the present invention.
[0086] Figure 8 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.
[0087] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the data acquisition method for the manual welding torch shown in the above embodiments is implemented.
[0088] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.
[0089] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.
Claims
1. A data acquisition system for a manual welding torch, characterized in that, The system includes: a manual welding torch, a vision sensor, an interaction module, and a data processing module; The vision sensor is fixed to the manual welding torch by a mounting structure, and the optical axis of the vision sensor is aligned with the molten pool area; the interaction module is mounted on the manual welding torch, and the data processing module is configured in the manual welding torch; The vision sensor is used to acquire images of the molten pool in the molten pool area in response to the welding task command input by the user according to the interaction module. The data processing module is used to create a task package in response to the welding task instruction input by the user according to the interaction module, and to collect process parameters during the execution of the welding task instruction, store the molten pool image and the process parameters into the task package, so as to train the robot welding model according to the molten pool image and the process parameters.
2. The system according to claim 1, characterized in that, The data processing module includes a metadata and time synchronization submodule; The metadata and time synchronization submodule is used to configure a unified timestamp for the molten pool image using the local clock as the time reference. The metadata and time synchronization submodule is also used to set corresponding task tags for the molten pool image and the process parameters configured with the unified timestamp; The metadata and time synchronization submodule is also used to mark defect events in response to user event marking instructions.
3. The system according to claim 1 or 2, characterized in that, The interaction module consists of physical buttons and / or a touch interface, which are used to indicate the equipment status and to control the welding process.
4. The system according to claim 1 or 2, characterized in that, The data acquisition system for the manual welding torch also includes a voice recording module; The voice recording module is configured on the manual welding gun and is used to record the voice during the welding task execution process.
5. The system according to claim 1 or 2, characterized in that, The data acquisition system for the manual welding torch also includes a power supply module; The power supply module is used to supply power to the vision sensor and the data processing module.
6. The system according to claim 1 or 2, characterized in that, The data acquisition system for the manual welding torch also includes a data interface module; The data interface module is used to export the task package to a preset server, so that the preset server can train a robot welding model based on the molten pool image and the process parameters.
7. The system according to claim 1 or 2, characterized in that, The data acquisition system for the manual welding torch also includes an attitude acquisition module; The attitude acquisition module is used to determine the relative motion information of the manual welding torch based on the molten pool image of the manual welding torch; the relative motion information is used to characterize the displacement and attitude changes of the manual welding torch. The attitude acquisition module is further configured to send the relative motion information to the data processing module, so that the data processing module stores the relative motion information in the task package.
8. The system according to claim 1 or 2, characterized in that, The data acquisition system for the manual welding torch also includes an acoustic sensor; The acoustic sensor is configured on the manual welding torch. The acoustic sensor is used to collect the arc sound signal during the welding process and send the arc sound signal to the data processing module so that the data processing module stores the arc sound signal in the task package.
9. The system according to claim 1 or 2, characterized in that, The data acquisition system for the manual welding torch also includes one or more of a temperature sensor, a light intensity sensor, and a gas sensor.
10. A data acquisition method for a manual welding torch, characterized in that, The method includes: In response to the welding task command input by the user, the system acquires images of the molten pool area. In response to a welding task instruction input by the user, a task package is created, process parameters are collected during the execution of the welding task instruction, and the molten pool image and the process parameters are stored in the task package to train a robot welding model based on the molten pool image and the process parameters.
11. An electronic device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes these computer instructions to perform the data acquisition method for the manual welding torch as described in claim 10.
12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the data acquisition method for the manual welding torch as described in claim 10.
13. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute the data acquisition method for a manual welding torch as described in claim 10.