A method, apparatus, system, device, medium, and product for optimizing a welding process

Abstract

This invention relates to the field of welding process technology, and discloses a welding process optimization method, apparatus, system, equipment, medium, and product. The method includes: collecting welding parameter data and welding image data corresponding to the target welding workpiece; using the welding parameter data to annotate the welding image data to obtain a welding sample dataset; training a welding process optimization model based on the welding sample dataset to obtain a trained welding process optimization model; wherein the welding process optimization model includes a visual detection sub-model and a process parameter regression sub-model; optimizing the welding process using the trained welding process optimization model to obtain welding process optimization parameters; and using the welding process optimization parameters to control a welding machine and an industrial robot to perform welding operations. This invention reduces the user's steps in calling the welding process, realizes automatic calling of the welding process, and achieves optimization of the welding process.

Description

Technical Field

[0001] This invention relates to the field of welding process technology, specifically to a welding process optimization method, apparatus, system, equipment, medium, and product. Background Technology

[0002] Related process software typically provides users with process retrieval functionality by integrating a welding process library. Specifically, this involves developing a process library software that provides interfaces for adding, deleting, modifying, and querying processes. Users can add welding process data to the database using the process addition function, manage processes through modification and deletion functions, and retrieve processes through the query function.

[0003] However, the aforementioned welding process software suffers from cumbersome parameter adjustments and poor adaptability, making it difficult to meet the modern industrial demand for high-efficiency, automated welding. Summary of the Invention

[0004] This invention provides a welding process optimization method, apparatus, system, equipment, medium, and product to solve the problems of cumbersome adjustment of welding process software parameters and poor adaptability.

[0005] In a first aspect, the present invention provides a welding process optimization method, the method comprising: Welding parameter data and welding image data corresponding to the target welding workpiece are collected. The welding image data is annotated using the welding parameter data to obtain a welding sample dataset. Based on the welding sample dataset, the welding process optimization model is trained to obtain the trained welding process optimization model; the welding process optimization model includes a visual detection sub-model and a process parameter regression sub-model. The welding process optimization model after training is used to optimize the welding process and obtain the welding process optimization parameters. Welding process optimization parameters are used to control welding machines and industrial robots to perform welding operations.

[0006] This embodiment provides a welding process optimization method that uses welding parameter data to annotate welding image data, then uses a welding sample dataset to train a welding process optimization model, and uses the trained welding process optimization model to optimize the welding process. This reduces the user's operation steps in calling the welding process, realizes the automatic calling of the welding process, optimizes the welding process, and improves welding efficiency and applicability.

[0007] In one optional implementation, welding image data is annotated using welding parameter data to obtain a welding sample dataset, including: The welding parameter data is standardized; this includes welding process data, welding machine parameter data, industrial robot trajectory data, and environmental data. Weld seam and defect annotations are performed on the welding image data to obtain the annotation file; Image preprocessing is performed on the welding image data to obtain the image of the welded workpiece; According to the timestamp, the annotation files and standardized welding parameter data are matched one by one with the images of the welded workpiece to obtain the welding annotation data; Welding sample datasets were obtained based on images of welded workpieces and welding annotation data.

[0008] This embodiment provides a welding process optimization method that standardizes welding parameter data to make it more accurate, annotates weld seams and defects in welding image data, and establishes a one-to-one correspondence between the annotation files and the standardized welding parameter data and the welded workpiece image. The welding workpiece image and the welding annotation data are then combined to provide accurate sample data for the subsequent training of the welding process optimization model, thereby improving the prediction accuracy of the welding process optimization model.

[0009] In one optional implementation, image preprocessing is performed on the welding image data to obtain a welded workpiece image, including: The welding image data is denoised and pixel values ​​are normalized to obtain preprocessed welding image data. The preprocessed welding image data is augmented to obtain the image of the welded workpiece.

[0010] This embodiment provides a welding process optimization method that improves the quality of welding images by performing noise reduction, pixel value normalization, and data augmentation on the welding image data. This ensures the stability of the welding process optimization model training and expands the welding image data, enabling diversification of welding conditions and providing key support for tasks such as welding defect detection and weld identification.

[0011] In one optional implementation, a welding process optimization model is trained based on a welding sample dataset to obtain a trained welding process optimization model, including: Based on the image of the welded workpiece, a visual inspection sub-model is used to detect weld seams and defects, thereby obtaining welding features. Based on welding characteristics, parameter regression is performed using a process parameter regression sub-model to obtain welding optimization parameters; Welding features and welding optimization parameters are compared with welding annotation data, and the model parameters are iteratively updated based on the comparison results to obtain the trained welding process optimization model.

[0012] This embodiment provides a welding process optimization method. Previously, welding processes required manual matching by workers or matching according to simple rules resulted in numerous welding process details, complex management, and process redundancy. This method uses visual acquisition of workpiece-related data, intelligent analysis of weld information, and automatic matching of welding processes. It utilizes a visual inspection sub-model and a process parameter regression sub-model for welding process optimization. On one hand, it can recommend welding processes for identified working conditions based on welding process data. On the other hand, leveraging the model's generalization ability, it can calculate welding process data for working conditions not included in the process library. Process parameters with good welding results can be saved in the process library, optimizing the welding process library and achieving welding process optimization, thereby improving the quality and reliability of welding process data.

[0013] In one alternative implementation, it further includes: Real-time images of the weld seam after welding are acquired. The welding process optimization model is adaptively updated based on the real-time images of the weld seam after welding. The welding process optimization model of the target welded workpiece is then optimized using the adaptively updated welding process optimization model.

[0014] This embodiment provides a welding process optimization method that improves the accuracy of the welding process optimization model by adaptively updating the model, making the generated welding process optimization parameters more applicable.

[0015] In a second aspect, the present invention provides a welding process optimization apparatus, the apparatus comprising: The acquisition module is used to acquire welding parameter data and welding image data corresponding to the target welding workpiece, and to use the welding parameter data to annotate the welding image data to obtain a welding sample dataset. The training module is used to train the welding process optimization model based on the welding sample dataset to obtain the trained welding process optimization model; the welding process optimization model includes a visual detection sub-model and a process parameter regression sub-model. The optimization module is used to optimize the welding process using the trained welding process optimization model to obtain welding process optimization parameters. The execution module is used to control the welding machine and industrial robot to perform welding operations by optimizing welding process parameters.

[0016] Thirdly, the present invention provides a welding process optimization system, comprising: an industrial control computer, a structured light camera, a welding machine, and an industrial robot; the industrial control computer is connected to the structured light camera, the welding machine, and the industrial robot respectively; the industrial control computer is used to execute the welding process optimization method of the first aspect above or any corresponding embodiment thereof.

[0017] 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 welding process optimization method of the first aspect or any corresponding embodiment described above.

[0018] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the welding process optimization method of the first aspect or any corresponding embodiment thereof.

[0019] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the welding process optimization method of the first aspect or any corresponding embodiment described above. Attached Figure Description

[0020] 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.

[0021] 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 process of a welding process optimization method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a second process of a welding process optimization method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the third process of a welding process optimization method according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the closed-loop management and control process according to an embodiment of the present invention; Figure 6 This is a structural block diagram of a welding process optimization device according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0022] 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.

[0023] 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.

[0024] 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.

[0025] The current welding process library consists of database software and welding process data. The database software is used to store and manage the welding process data, which includes indexes (keywords, such as 30mm thick steel plate) and data content (such as welding current and voltage). When using the data, the process is searched based on the index.

[0026] However, the aforementioned welding process software suffers from cumbersome parameter adjustments and poor adaptability, making it difficult to meet the demands of modern industry for high-efficiency and automated welding.

[0027] This invention provides a welding process optimization method and an integrated solution of visual algorithm and intelligent process software, which realizes closed-loop management of welding process parameter optimization and process call by visual algorithm, thereby reducing the debugging difficulty of welding process.

[0028] As an optional application scenario of this invention, such as Figure 1 As shown, a welding process optimization system includes: an industrial control computer 101, a structured light camera 102, a welding machine 103, and an industrial robot 104; the industrial control computer 101 is connected to the structured light camera 102, the welding machine 103, and the industrial robot 104 respectively; the industrial control computer 101 is used to execute the welding process optimization method.

[0029] The structured light camera 102 is used to acquire welding image data of the target welding workpiece, and the industrial control computer 101 is connected to the welding machine 103 through a PLC (Programmable Logic Controller).

[0030] The industrial control computer 101 utilizes C++ (a high-level computer programming language) or LabVIEW (a program development environment) to develop a unified software platform (i.e., an integrated platform). This platform includes: a data acquisition module, an image processing module, a visual model prediction module, a parameter closed-loop adjustment module, and an alarm and fault handling module. Specifically, the platform mainly integrates a data acquisition module (acquiring workpiece image data), an AI vision process module (intelligently calculating weld information and matching welding process data based on image data and process library data), and a welding process library (used to maintain welding process data). Data between modules is transmitted using predefined interface protocols (TCP / IP or CAN bus), and the data format and protocol strictly adhere to relevant standards. A graphical user interface provides a real-time display showing the status of the welding machine 103, industrial robot 104, and structured light camera 102; it displays the predicted parameters output by the model and real-time weld images, with an interface data refresh cycle of 1 second. Manual intervention is also supported, allowing operators to manually adjust parameters and record adjustment history upon receiving an abnormal alarm.

[0031] According to an embodiment of the present invention, a welding process optimization method embodiment 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 welding process optimization method, which can be used in the aforementioned industrial control computer 101. Figure 2 This is a flowchart of a welding process optimization method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps: Step S201: Collect welding parameter data and welding image data corresponding to the target welding workpiece, and use the welding parameter data to annotate the welding image data to obtain a welding sample dataset.

[0033] Step S202: Based on the welding sample dataset, the welding process optimization model is trained to obtain the trained welding process optimization model; wherein, the welding process optimization model includes a visual detection sub-model and a process parameter regression sub-model.

[0034] Specifically, the visual detection sub-model can adopt the network architecture of YOLOv3 (a single-stage object detection algorithm), and the process parameter regression sub-model can adopt a recurrent neural network.

[0035] Step S203: Optimize the welding process using the trained welding process optimization model to obtain the welding process optimization parameters.

[0036] Step S204: Optimize welding process parameters to control the welding machine and industrial robot to perform welding operations.

[0037] Specifically, the welding process optimization parameters are packaged into control commands through the closed-loop control module and sent to the control system of the welding machine 103 and the control system of the industrial robot. The welding machine 103 executes the welding action according to the welding process optimization parameters, and the robot makes path fine-tuning based on the updated trajectory, which can be the welding trajectory of the industrial robot.

[0038] This embodiment provides a welding process optimization method that uses welding parameter data to annotate welding image data, then uses a welding sample dataset to train a welding process optimization model, and uses the trained welding process optimization model to optimize the welding process. This reduces the user's operation steps in calling the welding process, realizes the automatic calling of the welding process, optimizes the welding process, and improves welding efficiency and applicability.

[0039] This embodiment provides a welding process optimization method, which can be used in the aforementioned industrial control computer 101. Figure 3 This is a flowchart of a welding process optimization method according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps: Step S301: Collect welding parameter data and welding image data corresponding to the target welding workpiece, and use the welding parameter data to annotate the welding image data to obtain a welding sample dataset.

[0040] Specifically, step S301 includes: Step S3011: Standardize the welding parameter data; wherein, the welding parameter data includes welding process data, welding machine parameter data, industrial robot trajectory data, and environmental data.

[0041] Specifically, welding process data is stored in an industrial control computer, and the standardization processing of welding process data includes: 1) Determine the data format and data packet structure: The parameter acquisition data includes: current: unit A, recorded to two decimal places (e.g., 15.25A); voltage: unit V, recorded to two decimal places (e.g., 32.50V); wire feed rate: unit mm / s, recorded to one decimal place (e.g., 5.3mm / s); welding path coordinates: recorded in the workpiece local coordinate system (unit: millimeters), with each coordinate point in the format (X,Y,Z) and a precision of one decimal place; preheating temperature: unit ℃, recorded as an integer. Among these, the current and voltage data are key data in the welding process, closely related to the material, plate thickness, and weld type of the welded workpiece. This data is integrated into the AI ​​model. After the workpiece's weld type, steel plate thickness, and other information are acquired visually, the model outputs the optimal current and voltage (process parameters) under the current conditions.

[0042] The data communication packet between the industrial control computer and other devices includes: Data header: fixed identifier 0xAA55 (2 bytes); Data length: fixed 2 bytes, indicating the number of bytes in the data body; Data body: arranged in a fixed order (such as current, voltage, wire feed rate, preheating temperature, welding coordinate data sequence); Check code: CRC16, used to check all subsequent bytes of the data body, with a length of 2 bytes.

[0043] 2) Data exchange protocols and database design: Protocol: Data transmission between devices will be conducted using TCP / IP (Transmission Control Protocol / Internet Protocol) or CAN bus (Controller Area Network) to ensure communication latency does not exceed 100ms; Database Design: A MySQL (Relational Database Management System) database will be established on the industrial control computer. The main data table fields include: param_id (primary key, auto-incrementing integer), parameter_name (string, such as "current"), value (floating-point number, such as 15.25), unit (string, such as "A"), timestamp (date and time, accurate to milliseconds), and device_id (string, such as welding machine, robot, or camera identification number).

[0044] All data is recorded according to a fixed timestamp and a strategy of scheduled full backup (every 24 hours) and incremental backup (every hour) is adopted.

[0045] Furthermore, a data acquisition module is embedded in the welding machine controller, with a sampling period of 1000ms. The output data includes: welding machine operating status, current current, voltage, and wire feeding rate. The data is sent to the industrial control computer via the RS232 interface in a predefined data packet format.

[0046] Further, industrial robot trajectory acquisition: enable the motion recording function in the robot controller to record joint angles (accuracy 0.1°) and end effector coordinates; the sampling rate is set to 50Hz; the data is transmitted to the industrial computer via the internal CAN bus in a predefined data structure (including machine number, joint sequence, and timestamp).

[0047] Furthermore, environmental data collection: Temperature and humidity sensors are installed to collect environmental data every 5 seconds and record it synchronously to the aforementioned database.

[0048] Step S3012: Add weld seam and defect annotations to the welding image data to obtain the annotation file.

[0049] Specifically, for structured light camera image acquisition: after installation, the camera's installation position and angle are fixed to ensure complete coverage of the welding area; the image resolution is set to 1920×1080, and the acquisition frame rate is fixed at 1fps; the image data is transmitted through the GigE (camera interface standard developed using the Gigabit Ethernet communication protocol) interface and saved on the industrial control computer as BMP (Bitmap-File, a graphic file format) or PNG (Portable Network Graphics, a bitmap image format using a lossless compression algorithm), with the filename including a timestamp and device number.

[0050] Furthermore, LabelImg (an open-source image annotation tool) or custom annotation software is used to annotate the images captured by the structured light camera. The annotation content includes weld boundaries (described in pixel coordinates), weld center lines, and defect areas (specifically, "incomplete fusion" areas). The annotated areas within the image are rectangular or polygonal, and the coordinate points are recorded (e.g., [(x1,y1),(x2,y2),...]).

[0051] Furthermore, the labeled data is stored in XML file format (Extensible Markup Language format), which must include image file name, image size, coordinates of the upper left and lower right corners of each target box, and the corresponding process parameter ID; all labeled files correspond one-to-one with the original images and welding parameter data through timestamps to ensure complete data association during subsequent training.

[0052] Step S3013: Perform image preprocessing on the welding image data to obtain the image of the welded workpiece.

[0053] In some optional implementations, step S3013 above includes: Step a1: Denoise the welding image data and normalize the pixel values ​​to obtain preprocessed welding image data.

[0054] Specifically, the acquired image is denoised (median filtering, kernel size 3×3), light equalized (histogram equalization), and geometric distortion corrected (perspective transformation using calibration parameters) before being saved.

[0055] Furthermore, the pixel values ​​of all training images are normalized (0~1).

[0056] Step a2 involves performing data augmentation on the preprocessed welding image data to obtain an image of the welded workpiece.

[0057] Specifically, data augmentation includes: random rotation (range ±10°), horizontal flipping (probability 50%), and random adjustment of brightness and contrast (±10% variation); while simultaneously updating the corresponding annotation information (coordinate transformation).

[0058] Step S3014: According to the timestamp, match the annotation file and the standardized welding parameter data with the image of the welded workpiece one by one to obtain the welding annotation data.

[0059] Step S3015: Obtain a welding sample dataset based on the welding workpiece image and welding annotation data.

[0060] Specifically, 80% of the collected data is used as the training set, 10% as the validation set, and 10% as the test set, with a fixed random seed to ensure reproducibility.

[0061] Step S302: Based on the welding sample dataset, train the welding process optimization model to obtain the trained welding process optimization model; the welding process optimization model includes a visual detection sub-model and a process parameter regression sub-model. For details, please refer to [link to details]. Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0062] Step S303: Optimize the welding process using the trained welding process optimization model to obtain the optimized welding process parameters. For details, please refer to [link to relevant documentation]. Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0063] Step S304 involves using optimized welding process parameters to control the welding machine and industrial robot to perform the welding operation. For details, please refer to [link to relevant documentation]. Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0064] This embodiment provides a welding process optimization method that standardizes welding parameter data to make it more accurate, annotates weld seams and defects in welding image data, and establishes a one-to-one correspondence between the annotation files and the standardized welding parameter data and the welded workpiece image. The welding workpiece image and the welding annotation data are then combined to provide accurate sample data for the subsequent training of the welding process optimization model, thereby improving the prediction accuracy of the welding process optimization model.

[0065] This embodiment provides a welding process optimization method, which can be used in the aforementioned industrial control computer. Figure 4 This is a flowchart of a welding process optimization method according to an embodiment of the present invention, such as... Figure 4 As shown, the process includes the following steps: Step S401: Collect welding parameter data and welding image data corresponding to the target welding workpiece. Use the welding parameter data to annotate the welding image data to obtain a welding sample dataset. For details, please refer to [link to relevant documentation]. Figure 3 Step S301 of the illustrated embodiment will not be described again here.

[0066] Step S402: Based on the welding sample dataset, train the welding process optimization model to obtain the trained welding process optimization model; wherein, the welding process optimization model includes a visual detection sub-model and a process parameter regression sub-model.

[0067] Specifically, step S402 includes: Step S4021: Based on the image of the welded workpiece, the weld seam and defects are detected using a visual inspection sub-model to obtain the welding features.

[0068] Specifically, the visual inspection sub-model adopts a YOLOv3-based network architecture, and the input image size is uniformly adjusted to 640×480. The input of the visual inspection sub-model, namely the welding workpiece image, is in the format of RGB (a color system) image (usually uniform in size), and the type is a two-dimensional image (2D image). Its purpose is to perform target detection (the position and category of the detection box) in the image.

[0069] Furthermore, the network hierarchy of the visual inspection sub-model includes three convolutional modules. Each convolutional module contains a convolutional layer, a Batch Normalization layer, and a ReLU activation layer, followed by three detection layers to achieve real-time detection of weld boundaries and defects. The output data includes the coordinates of the target detection box, the category probability, etc.

[0070] Furthermore, 3×3 small-sized convolutional kernels are configured for the convolutional layers, with the number of kernels increasing with network depth (e.g., 64 in the first layer, 128 in the middle layer, and 256 in the deep layer). The stride is set to 1, and the padding (used to add boundary values ​​around the input data to maintain the size of the feature map after convolution or pooling) is set to 1. The convolutional kernels are multiplied element-wise with the local regions of the input welded workpiece image to generate a feature map containing local features. The weld edge features are extracted by low-level convolutional kernels, and the texture of the defect area is extracted by high-level convolutional kernels. Multi-scale fusion is performed on the local feature map to capture weld defect features at different scales, resulting in an unnormalized feature map.

[0071] Furthermore, the unnormalized feature map output from the convolutional layer is input into the Batch Normalization layer to calculate the mean and variance of each feature channel in the unnormalized feature map; then, the feature map is normalized by channel based on the mean and variance; the normalized feature map is scaled and shifted to preserve feature diversity.

[0072] Furthermore, the feature map output from the Batch Normalization layer is input into the ReLU activation layer. The ReLU activation function is used to set the pixel values ​​less than 0 in the feature map to 0, retaining non-negative features, and outputting the current feature map. The current feature map has the same dimension as the welding workpiece image input to the convolutional layer.

[0073] Furthermore, the feature maps output by the convolution module are downsampled through convolutional layers or pooling layers with a stride of 2 to generate three sets of feature maps at different scales. For example, small-scale feature maps retain fine-grained information and focus on weld boundaries and minor defects (such as micro-pores and scratches); medium-scale feature maps balance fine-grained and global information and focus on medium-sized defects (such as medium-sized cracks and slag inclusions); large-scale feature maps have the largest receptive field and focus on large-sized defects (such as large areas of incomplete penetration and severe undercut).

[0074] Furthermore, each detection layer includes a classification branch and a regression branch, which respectively determine the category and location of defects / boundaries: The classification branch applies a 1×1 convolution kernel to the input feature map (small-scale feature map / medium-scale feature map / large-scale feature map), adjusting the number of feature channels to the number of categories × the number of anchor boxes; it uses an activation function (e.g., Sigmoid or Softmax) to output the probability of each category corresponding to each anchor box; scale adaptation: the anchor box size of the small-scale detection layer is smaller (e.g., 16×16), and the anchor box size of the large-scale detection layer is larger (e.g., 64×64); The regression branch applies a 1×1 convolution kernel to the input feature map, and the number of output channels is 4 × the number of anchor boxes, where 4 corresponds to the 4 coordinate offsets of the bounding box. The output result is the offset of the anchor box relative to the true bounding box, and the true coordinates are restored through the decoding formula.

[0075] Furthermore, the three detection layers output prediction results (category probability + bounding box coordinates) at their respective scales. The predicted bounding boxes at all scales are sorted by category probability, redundant boxes with overlap exceeding the threshold are removed, and the bounding boxes with the highest confidence are retained. The final weld boundary contour and defect detection results (including defect category, location, and confidence) are output to achieve the goal of real-time detection.

[0076] Step S4022: Based on welding characteristics, perform parameter regression using the process parameter regression sub-model to obtain welding optimization parameters.

[0077] Specifically, based on welding characteristics, parameter regression is performed through a 2-3 layer fully connected neural network to output optimized parameters (current, voltage, wire feed rate, etc.), which are the welding optimization parameters; the regression loss adopts the mean square error (MSE), and the target error is controlled within ±5%.

[0078] Furthermore, the visual inspection sub-model can identify information such as the location and type of the weld based on the input weld image data, and provide the best matching welding process. The process parameter regression sub-model is an objective evaluation of whether the given welding process is accurate, and is used to verify and evaluate the model.

[0079] Step S4023: Compare the welding features and welding optimization parameters with the welding annotation data respectively, and iteratively update the model parameters based on the comparison results to obtain the trained welding process optimization model.

[0080] Specifically, the training parameters were set as follows: initial learning rate 0.001, batch size 32, and Adam optimizer; the training period was set to 200 epochs; the loss was calculated using the validation set every 1000 iterations, the model changes were recorded, and the learning rate was decayed (reduced to 0.5 of the original value) when the loss plateaued.

[0081] Furthermore, the welding features output by the visual detection sub-model can be compared with the welding annotation data. If the difference between the welding features and the welding annotation data is large, the visual detection sub-model is iteratively trained to obtain a trained visual detection sub-model. Then, the current welding features are output by the trained visual detection sub-model and input into the process parameter regression sub-model. The welding optimization parameters output by the process parameter regression sub-model are compared with the welding annotation data. If the difference is large, the process parameter regression sub-model is iteratively trained to obtain a trained welding process optimization model. Finally, the trained visual detection sub-model and the trained welding process optimization model are integrated to form a trained welding process optimization model.

[0082] Furthermore, a real-time feedback mechanism is established. During the actual welding process, after each batch of new data is collected (e.g., once every 5 seconds), the data is sent to the model for retraining or parameter fine-tuning. The welding parameters predicted by the model are compared with the welding effect collected in real time (the weld quality image is obtained through a structured light camera). If the error exceeds the preset error range (±5%), the model parameter update algorithm is triggered (using an online learning method, fixing the most recent 50 data samples for small-batch retraining), and the update log is recorded.

[0083] Furthermore, the target detection accuracy (target target accuracy ≥ 95%) and parameter regression error (controlled within ± 5%) are evaluated using a test set; an evaluation report is output, including the average loss curve, error histogram, regression scatter plot, etc., to ensure that the technical effect data has reproducibility and statistical basis.

[0084] Step S403: Optimize the welding process using the trained welding process optimization model to obtain the welding process optimization parameters.

[0085] Specifically, such as Figure 5 As shown, the specific steps for optimizing the welding process using the trained welding process optimization model to obtain the welding process optimization parameters include: 1) Data acquisition stage: Each device uploads data to the industrial control computer according to the predetermined sampling cycle. The industrial control computer writes all data (welding machine parameters, robot trajectory, environmental data, image data) into the database and performs real-time caching.

[0086] 2) Image processing and feature extraction: After receiving the image data, it first undergoes denoising, equalization and geometric correction, and then enters the pre-trained YOLOv3 detection module (i.e. visual detection sub-model) to extract features such as weld boundaries and defect areas.

[0087] 3) Model prediction and parameter regression: Using the features output by the visual inspection sub-model, the process parameter regression sub-model calculates the optimized parameters for the next cycle and outputs updated values ​​for current, voltage, wire feed rate, etc.

[0088] 4) Parameter output and execution adjustment: The predicted parameters (i.e. welding process optimization parameters) are packaged into control commands (based on the CAN bus data structure) through the closed-loop control module and sent to the welding machine control system and the robot control system; the welding machine executes welding actions according to the new parameters, and the robot makes path fine adjustments based on the updated trajectory.

[0089] Step S404: Optimize welding process parameters to control the welding machine and industrial robot to perform welding operations.

[0090] Specifically, real-time images of the weld seam after welding are acquired, the welding process optimization model is adaptively updated based on the real-time images of the weld seam after welding, and the welding process of the target welded workpiece is optimized using the adaptively updated welding process optimization model.

[0091] Furthermore, real-time images of the weld seam after welding are acquired and compared with the predicted image (using edge overlap rate and morphological similarity calculations); if the difference exceeds the 5% error threshold, abnormal data is automatically recorded and recalculation is triggered: the newly acquired data and historical data are combined into a small batch sample for online model fine-tuning; the adjusted new parameters are output again in the next cycle, forming a complete closed-loop control feedback process, with the cycle controlled within 5 seconds to meet real-time requirements.

[0092] Furthermore, a complete experimental platform was built in the laboratory. Each module was debugged individually to confirm that the data acquisition, preprocessing, model prediction, and parameter output all met the preset standards. The overall system was then integrated and debugged, and the time delay, data transmission success rate, and weld image quality of each step were recorded.

[0093] This embodiment provides a welding process optimization method. Previously, welding processes required manual matching by workers or matching according to simple rules resulted in numerous welding process details, complex management, and process redundancy. This method uses visual acquisition of workpiece-related data, intelligent analysis of weld information, and automatic matching of welding processes. It utilizes a visual inspection sub-model and a process parameter regression sub-model for welding process optimization. On one hand, it can recommend welding processes for identified working conditions based on welding process data. On the other hand, leveraging the model's generalization ability, it can calculate welding process data for working conditions not included in the process library. Process parameters with good welding results can be saved in the process library, optimizing the welding process library and achieving welding process optimization, thereby improving the quality and reliability of welding process data.

[0094] This embodiment also provides a welding process optimization device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0095] This embodiment provides a welding process optimization device, such as... Figure 6 As shown, it includes: The acquisition module 601 is used to acquire welding parameter data and welding image data corresponding to the target welding workpiece, and to use the welding parameter data to annotate the welding image data to obtain a welding sample dataset. Training module 602 is used to train the welding process optimization model based on the welding sample dataset to obtain the trained welding process optimization model; wherein, the welding process optimization model includes a visual detection sub-model and a process parameter regression sub-model; Optimization module 603 is used to optimize the welding process using the trained welding process optimization model to obtain welding process optimization parameters; The execution module 604 is used to control the welding machine and industrial robot to perform welding operations by optimizing welding process parameters.

[0096] In some alternative implementations, the acquisition module 601 includes: The standardization processing unit is used to standardize welding parameter data, which includes welding process data, welding machine parameter data, industrial robot trajectory data, and environmental data. The annotation unit is used to annotate weld seams and defects in welding image data to obtain annotation files; The image preprocessing unit is used to preprocess the welding image data to obtain the image of the welded workpiece. The corresponding unit is used to match the annotation file and the standardized welding parameter data with the image of the welded workpiece one by one according to the timestamp to obtain the welding annotation data; The unit is defined to obtain a welding sample dataset based on the image of the welded workpiece and the welding annotation data.

[0097] In some optional implementations, the image preprocessing unit includes: The preprocessing subunit is used to perform noise reduction and pixel value normalization on the welding image data to obtain preprocessed welding image data. The data augmentation subunit is used to augment the preprocessed welding image data to obtain the welded workpiece image.

[0098] In some alternative implementations, training module 602 includes: The detection unit is used to detect weld seams and defects based on images of welded workpieces using a visual detection sub-model, thereby obtaining welding features. The parameter regression unit is used to perform parameter regression based on welding characteristics using the process parameter regression sub-model to obtain welding optimization parameters. The iterative update unit is used to compare the welding features and welding optimization parameters with the welding annotation data, and to iteratively update the model parameters based on the comparison results to obtain the trained welding process optimization model.

[0099] In some alternative implementations, it also includes: The adaptive update module is used to acquire real-time images of the weld after welding, adaptively update the welding process optimization model based on the real-time images of the weld after welding, and use the adaptively updated welding process optimization model to optimize the welding process of the target welded workpiece.

[0100] The welding process optimization apparatus provided in this embodiment of the invention can execute a welding process optimization method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0101] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0102] The following is a detailed reference. Figure 7 This 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.) 701, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 702 or a program loaded from memory 708 into random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device. The processor 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0103] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 7 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.

[0104] 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 709, or installed from a memory 708, or installed from a ROM 702. When the computer program is executed by the processor 701, it performs the functions defined in a welding process optimization method according to an embodiment of the present invention.

[0105] Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0106] 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 computer code that can be recorded 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 then 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, a welding process optimization method shown in the above embodiments is implemented.

[0107] 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.

[0108] 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 welding process optimization method, characterized in that, The method includes: Collect welding parameter data and welding image data corresponding to the target welding workpiece, and use the welding parameter data to annotate the welding image data to obtain a welding sample dataset. Based on the welding sample dataset, the welding process optimization model is trained to obtain the trained welding process optimization model; wherein, the welding process optimization model includes a visual detection sub-model and a process parameter regression sub-model; The trained welding process optimization model is used to optimize the welding process and obtain the welding process optimization parameters. The welding process optimization parameters are used to control the welding machine and industrial robot to perform welding operations.

2. The method according to claim 1, characterized in that, The step of annotating the welding image data using the welding parameter data to obtain a welding sample dataset includes: The welding parameter data is standardized; wherein, the welding parameter data includes welding process data, welding machine parameter data, industrial robot trajectory data, and environmental data. Weld seam and defect annotations are performed on the welding image data to obtain an annotation file; The welding image data is preprocessed to obtain the image of the welded workpiece; According to the timestamp, the annotation file and the standardized welding parameter data are matched one-to-one with the image of the welded workpiece to obtain the welding annotation data; The welding sample dataset is obtained based on the image of the welded workpiece and the welding annotation data.

3. The method according to claim 2, characterized in that, The step of preprocessing the welding image data to obtain a welded workpiece image includes: The welding image data is subjected to denoising and pixel value normalization to obtain preprocessed welding image data. The preprocessed welding image data is augmented to obtain the image of the welded workpiece.

4. The method according to claim 2, characterized in that, The step of training the welding process optimization model based on the welding sample dataset to obtain the trained welding process optimization model includes: Based on the image of the welded workpiece, the visual inspection sub-model is used to detect weld seams and defects, thereby obtaining welding features. Based on the welding characteristics, parameter regression is performed using the process parameter regression sub-model to obtain welding optimization parameters; The welding features and welding optimization parameters are compared with the welding annotation data respectively. Based on the comparison results, the model parameters are iteratively updated to obtain the trained welding process optimization model.

5. The method according to claim 1, characterized in that, Also includes: Real-time images of the weld seam after welding are acquired, and the welding process optimization model is adaptively updated based on the real-time images of the weld seam after welding. The welding process optimization model of the target welded workpiece is then optimized using the adaptively updated welding process optimization model.

6. A welding process optimization device, characterized in that, The device includes: The acquisition module is used to acquire welding parameter data and welding image data corresponding to the target welding workpiece, and to use the welding parameter data to annotate the welding image data to obtain a welding sample dataset. The training module is used to train the welding process optimization model based on the welding sample dataset to obtain the trained welding process optimization model; wherein, the welding process optimization model includes a visual detection sub-model and a process parameter regression sub-model. The optimization module is used to optimize the welding process using the trained welding process optimization model to obtain welding process optimization parameters. The execution module is used to control the welding machine and industrial robot to perform welding operations using the welding process optimization parameters.

7. A welding process optimization system, characterized in that, include: Industrial control computers, structured light cameras, welding machines, and industrial robots; The industrial control computer is connected to the structured light camera, the welding machine, and the industrial robot, respectively. The industrial control computer is used to execute the welding process optimization method according to any one of claims 1 to 5.

8. An electronic device, characterized in that, include: A memory and a processor are interconnected, the memory storing computer instructions, and the processor executing the computer instructions to perform the welding process optimization method according to any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the welding process optimization method according to any one of claims 1 to 5.

10. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute the welding process optimization method according to any one of claims 1 to 5.

Images