Welding path planning method and apparatus

Abstract

The embodiment of the application provides a kind of welding path planning method and device, method includes: by machine vision algorithm to the object to be processed is three-dimensional scanning, obtains three-dimensional point cloud data, by the segmentation algorithm based on deep learning to three-dimensional point cloud data is data analysis, and the geometry feature is identified;With the geometry feature is constructed into a undirected graph, and the optimal connected path of the weld center line is found on the graph by minimum spanning tree algorithm, and the optimal connected path is constructed to obtain the fine three-dimensional model of processing object;According to the three-dimensional coordinates of weld search site, current welding gun attitude and target weld length, determine the optimal search height, search depth, inclination angle and offset of welding gun, generate corresponding welding path planning based on search height, search depth, inclination angle and offset by welding process simulation algorithm;The application can effectively improve the precision and efficiency of welding path planning.

Description

Technical Field

[0001] This application relates to the field of industrial robots, specifically to a welding path planning method and apparatus. Background Technology

[0002] Welding path planning, a crucial step in industrial automated welding, has a direct and profound impact on welding quality and efficiency. With the rapid development of Industry 4.0 and intelligent manufacturing, welding path planning technology is constantly advancing, but it still faces many challenges and limitations. Traditional welding path planning methods mainly rely on manual programming or simple geometric algorithms. While these methods still have applications in certain scenarios, they are no longer sufficient to meet the demands of modern manufacturing for high precision, high efficiency, and high adaptability.

[0003] In recent years, with the development of computer vision and artificial intelligence technologies, some new welding path planning methods have emerged. Sensor-based adaptive planning methods dynamically adjust the welding path by acquiring weld information in real time, improving the system's adaptability. However, the effectiveness of this method is largely limited by the accuracy and response speed of the sensors. On the other hand, offline programming using CAD / CAM software improves planning efficiency, but often requires accurate workpiece models and is difficult to adapt to various changes in actual production.

[0004] While these methods have improved welding path planning to varying degrees, some common problems remain. First, there is insufficient accuracy, especially for workpieces with complex shapes or irregular surfaces, where existing methods struggle to achieve high-precision weld identification and path planning. Second, there is poor adaptability; most methods cannot quickly adapt to changes in workpieces or dynamic situations during the welding process. Furthermore, inefficiency is also a common problem; many methods either require significant manual intervention or complex calculations, making it difficult to meet the demands of high-efficiency production.

[0005] Current welding path planning technology still faces the problem of insufficient comprehensiveness. Many methods tend to focus only on one aspect of path planning, such as geometric features, while neglecting other important factors, such as welding process parameters and heat-affected zones. This one-sidedness limits the practicality and reliability of the planning results. At the same time, poor real-time performance is also an urgent problem to be solved. Many complex planning algorithms are difficult to calculate and adjust in real time, limiting their application in dynamic production environments.

[0006] In recent years, AI-based welding path planning methods have become a research hotspot. These methods utilize technologies such as machine learning and deep learning to improve the intelligence of the planning process. However, these methods are still in the exploratory stage and face many challenges in practical applications, such as how to effectively process large amounts of 3D data, how to accurately identify complex weld features, and how to achieve real-time path optimization.

[0007] In summary, although welding path planning technology has made some progress, many problems and challenges still exist. Developing an intelligent welding path planning method that can comprehensively consider multiple factors and possesses high precision, high efficiency, strong adaptability, and good real-time performance remains an important research direction. This method needs to be able to effectively process complex 3D data, intelligently identify weld features, optimize the welding path, and dynamically adjust it according to actual conditions. Simultaneously, it is also necessary to consider how to effectively integrate advanced algorithm technologies with practical industrial robot systems to achieve a leap from theory to practice, ultimately meeting the demands of modern manufacturing for intelligent and automated welding. Summary of the Invention

[0008] To address the problems in the prior art, this application provides a welding path planning method and apparatus, which can effectively improve the accuracy and efficiency of welding path planning.

[0009] To solve at least one of the above problems, this application provides the following technical solution:

[0010] In a first aspect, this application provides a welding path planning method, including:

[0011] The machine vision algorithm is used to perform a 3D scan of the object to be processed to obtain the 3D point cloud data of the object. The 3D point cloud data is then analyzed by a deep learning-based segmentation algorithm to identify the geometric features of the weld area on the surface of the object.

[0012] The extracted geometric features are constructed into an undirected graph, the weld centerline is discretized into graph nodes, the weld width is used as the graph edge weight, the minimum spanning tree algorithm is used to find the optimal connected path of the weld centerline on the graph, and the fine three-dimensional model of the processing object is constructed with the optimal connected path, wherein the three-dimensional coordinates of the weld finding point are marked in the fine three-dimensional model.

[0013] Based on the three-dimensional coordinates of the weld seam locating point, the current welding torch posture, and the target weld seam length, the optimal locating height, locating depth, tilt angle, and offset of the welding torch are determined. Based on the locating height, locating depth, tilt angle, and offset, a corresponding welding path plan is generated through a welding process simulation algorithm. The welding path plan is then output to the industrial robot control system to drive the industrial robot to perform welding operations along the welding path.

[0014] Furthermore, the step of performing a 3D scan of the object to be processed using a machine vision algorithm to obtain the 3D point cloud data of the object includes:

[0015] Machine vision technology is used to perform 3D scanning of the object to be processed, and 3D point cloud data is extracted and organized from the raw data obtained by scanning.

[0016] The acquired 3D point cloud data is preprocessed by data cleaning and noise reduction to obtain 3D point cloud data after the preprocessing operation.

[0017] Furthermore, the step of analyzing the 3D point cloud data using a deep learning-based segmentation algorithm to identify the geometric features of the weld area on the surface of the processed object includes:

[0018] The three-dimensional point cloud data is input into a set deep learning-based semantic segmentation algorithm for fine segmentation;

[0019] Extract the geometric features of the weld region from the output of the algorithm, wherein the geometric features include at least one of the centerline, width, and shape of the weld.

[0020] Further, the step of constructing an undirected graph from the extracted geometric features, discretizing the weld centerline into graph nodes, using the weld width as the graph edge weight, and finding the optimal connected path of the weld centerline on the graph using the minimum spanning tree algorithm includes:

[0021] The centerline and width of the geometric features are constructed into an undirected graph, the centerline is discretized and represented as a node of the graph, and the width is used as the weight of the graph edge;

[0022] The minimum spanning tree algorithm is applied to the constructed undirected graph model to determine the shortest path connecting all nodes as the optimal connected path for the weld centerline.

[0023] Furthermore, the detailed three-dimensional model of the processed object is constructed using the optimal connectivity path, wherein the detailed three-dimensional model is marked with the three-dimensional coordinates of the weld seam locating points, including:

[0024] The obtained optimal connection path of the weld centerline is integrated into the original three-dimensional geometric model of the processed object to generate a fine three-dimensional model, wherein the fine three-dimensional model contains geometric details of the weld path;

[0025] The three-dimensional coordinate information of the weld seam locating point is marked in the detailed three-dimensional model.

[0026] Further, determining the optimal positioning height, positioning depth, tilt angle, and offset of the welding torch based on the three-dimensional coordinates of the weld seam positioning point, the current welding torch posture, and the target weld seam length includes:

[0027] Extract the three-dimensional coordinates of the weld seam locating point marked in the fine three-dimensional model, and analyze the geometric relationship between the locating point coordinates, the welding gun posture, and the weld seam length by combining the actual posture of the current welding gun and the expected weld seam length target.

[0028] Based on the analysis results of the geometric relationship, the optimal positioning height, optimal positioning depth, tilt angle, and offset of the welding torch are calculated using an optimization algorithm.

[0029] Further, the step of generating the corresponding welding path plan based on the positioning height, the positioning depth, the tilt angle, and the offset using a welding process simulation algorithm includes:

[0030] The positioning height, positioning depth, tilt angle, and offset in the obtained welding torch positioning parameters are input into a preset welding process simulation algorithm for calculation.

[0031] Output a welding path plan that meets the requirements based on the simulation results.

[0032] Secondly, this application provides a welding path planning device, comprising:

[0033] The 3D construction module is used to perform 3D scanning of the object to be processed using machine vision algorithms to obtain 3D point cloud data of the object. The 3D point cloud data is then analyzed using a deep learning-based segmentation algorithm to identify the geometric features of the weld area on the surface of the object.

[0034] The coordinate determination module is used to construct an undirected graph from the extracted geometric features, discretize the weld centerline into graph nodes, use the weld width as the graph edge weight, find the optimal connected path of the weld centerline on the graph using the minimum spanning tree algorithm, and construct a fine three-dimensional model of the processing object using the optimal connected path, wherein the fine three-dimensional model is marked with the three-dimensional coordinates of the weld finding point.

[0035] The path planning module is used to determine the optimal positioning height, positioning depth, tilt angle, and offset of the welding torch based on the three-dimensional coordinates of the weld seam positioning point, the current welding torch posture, and the target weld seam length. Based on the positioning height, the positioning depth, the tilt angle, and the offset, the module generates a corresponding welding path plan through a welding process simulation algorithm and outputs the welding path plan to the industrial robot control system to drive the industrial robot to perform welding operations along the welding path.

[0036] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the welding path planning method.

[0037] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the welding path planning method described above.

[0038] Fifthly, this application provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the welding path planning method described above.

[0039] As can be seen from the above technical solution, this application provides a welding path planning method and apparatus. It uses a machine vision algorithm to perform a three-dimensional scan of the object to be processed, obtaining three-dimensional point cloud data. A deep learning-based segmentation algorithm is then used to analyze the three-dimensional point cloud data and identify geometric features. These geometric features are then used to construct an undirected graph. A minimum spanning tree algorithm is used to find the optimal connected path for the weld centerline on this graph, and a detailed three-dimensional model of the object to be processed is constructed using this optimal connected path. Based on the three-dimensional coordinates of the weld search point, the current welding torch posture, and the target weld length, the optimal search height, search depth, tilt angle, and offset of the welding torch are determined. Based on the search height, search depth, tilt angle, and offset, a corresponding welding path plan is generated using a welding process simulation algorithm. This effectively improves the accuracy and efficiency of welding path planning. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is one of the flowcharts illustrating the welding path planning method in the embodiments of this application;

[0042] Figure 2 This is a second flowchart illustrating the welding path planning method in the embodiments of this application;

[0043] Figure 3 This is the third flowchart illustrating the welding path planning method in the embodiments of this application;

[0044] Figure 4 This is the fourth flowchart illustrating the welding path planning method in the embodiments of this application;

[0045] Figure 5 This is the fifth flowchart illustrating the welding path planning method in the embodiments of this application;

[0046] Figure 6 This is the sixth flowchart illustrating the welding path planning method in the embodiments of this application;

[0047] Figure 7 This is the seventh flowchart illustrating the welding path planning method in the embodiments of this application;

[0048] Figure 8 This is a structural diagram of the welding path planning device in the embodiments of this application;

[0049] Figure 9 This is a schematic diagram of the structure of the electronic device in the embodiments of this application.

[0050] Figure label:

[0051] Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver storage unit 9144, antenna 9111, speaker 9131, microphone 9132. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0053] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0054] In view of the problems existing in the prior art, this application provides a welding path planning method and apparatus. The method involves using a machine vision algorithm to perform a 3D scan of the object to be processed, obtaining 3D point cloud data. A deep learning-based segmentation algorithm is then used to analyze the 3D point cloud data and identify geometric features. These geometric features are then used to construct an undirected graph. A minimum spanning tree algorithm is then used to find the optimal connected path for the weld centerline on this graph, and a detailed 3D model of the object to be processed is constructed using this optimal connected path. Based on the 3D coordinates of the weld search point, the current welding torch posture, and the target weld length, the optimal search height, search depth, tilt angle, and offset of the welding torch are determined. Based on the search height, search depth, tilt angle, and offset, a corresponding welding path plan is generated using a welding process simulation algorithm. This effectively improves the accuracy and efficiency of welding path planning.

[0055] To effectively improve the accuracy and efficiency of welding path planning, this application provides an embodiment of a welding path planning method, see [link to embodiment]. Figure 1 The welding path planning method specifically includes the following:

[0056] Step S101: The object to be processed is scanned in three dimensions using a machine vision algorithm to obtain the three-dimensional point cloud data of the object. The three-dimensional point cloud data is then analyzed using a deep learning-based segmentation algorithm to identify the geometric features of the weld area on the surface of the object.

[0057] Optionally, in this embodiment, step S101 is a key step in the intelligent welding system, involving multiple advanced technical fields such as machine vision, 3D scanning technology, point cloud processing, and deep learning. The main goal of this step is to acquire detailed 3D information of the object to be processed and accurately identify the areas that need to be welded. This process provides a foundation for subsequent welding parameter configuration and path planning, improving the accuracy and efficiency of the entire welding process.

[0058] First, we will discuss the technical process of 3D scanning in detail. 3D scanning typically employs structured light scanning or laser triangulation. Taking structured light scanning as an example, the system projects a specific grating pattern onto the surface of the object to be processed, and then captures these deformed grating patterns using a high-resolution camera. By analyzing the degree of grating deformation, the system can calculate the 3D coordinates of each point on the object's surface. This process usually requires scanning from multiple angles to obtain complete 3D information.

[0059] In practice, we might use multiple scanning heads working simultaneously to improve scanning speed and coverage. For example, in a large welding workstation, we might install 4-6 scanning heads at different locations, which can simultaneously scan the workpiece. The obtained data is then fused to obtain a complete 3D model. This method can significantly shorten scanning time, especially for large or complex-shaped workpieces.

[0060] After obtaining the raw 3D point cloud data, the next step is to analyze the data using deep learning segmentation algorithms. Here, "segmentation" refers to classifying each point in the point cloud into different semantic categories, such as weld areas, non-weld areas, etc. Deep learning algorithms, especially models based on graph neural networks (GNNs) or point cloud-specific networks (such as PointNet++), perform exceptionally well in this type of task.

[0061] These algorithms work by converting 3D point cloud data into high-dimensional feature representations, and then using multi-layer neural networks to learn how to classify each point based on these features. Specifically, the algorithm first preprocesses the point cloud, including downsampling and normal vector estimation. Then, it extracts local and global features. Local features may include geometric information about the surrounding shape of each point, while global features may include the shape information of the entire object.

[0062] Network architecture typically includes multiple feature extraction layers and feature aggregation layers. Feature extraction layers may use convolution-like operations, but are specifically designed for point cloud data to handle its disorder and irregularity. Feature aggregation layers are responsible for combining local and global features to form a richer representation.

[0063] During training, we need a large amount of labeled data, including various types of welded workpieces and their corresponding weld location labels. During training, the network attempts to predict the category of each point, then compares the predictions with the true labels to calculate the loss function. Through backpropagation, the network parameters are continuously adjusted to minimize the loss function.

[0064] To improve the model's generalization ability and reduce the amount of labeled data required, we might employ transfer learning. First, a model is pre-trained on a large amount of general-purpose 3D point cloud data, and then fine-tuned using relatively little welding-specific data. This approach allows the model to learn general 3D features and then adapt to the specific requirements of the welding task.

[0065] During the inference phase, once a new workpiece is scanned, the point cloud data is fed into the trained network. The network classifies each point and outputs the probability that each point belongs to the weld area. Then, we can obtain the final weld area by setting a threshold or using post-processing algorithms (such as region growing, contour extraction, etc.).

[0066] This process solved several key technical problems. First, it addressed how to quickly and accurately acquire 3D information of workpieces with complex shapes. Traditional 2D image processing methods often fall short when dealing with complex 3D structures, while 3D scanning technology can provide more comprehensive and accurate information. Second, it solved the problem of accurately identifying weld seam areas from large amounts of 3D point cloud data. Weld seams typically occupy only a small portion of the entire workpiece surface and can have various complex shapes and locations. By using deep learning algorithms, the system can learn various complex feature patterns, thus accurately identifying weld seam areas on a wide variety of workpieces.

[0067] In terms of technical effectiveness, this method brings significant improvements. The accuracy of identification is greatly enhanced. Compared to traditional rule-based methods, deep learning algorithms can handle more complex situations, potentially increasing the identification accuracy from 90% to over 99%. This means that in actual production, the rework rate caused by misidentification can be significantly reduced. Furthermore, this method greatly improves the system's adaptability and flexibility. Traditional methods may require writing separate rules for each new workpiece type, while deep learning-based methods only need fine-tuning with new data to quickly adapt to new workpiece types.

[0068] In practical applications, this technology can bring about revolutionary changes. For example, in a welding workshop of an aerospace parts manufacturing plant, when a complex metal component (such as a turbine blade or engine nacelle) needs to be welded, the system can complete the entire 3D scanning and weld seam identification process within minutes. This process not only accurately locates each position that needs to be welded but also identifies various special welding requirements, such as weld depth and angle. This information can be directly used to configure welding robots, ensuring that every weld point is accurately positioned and every weld seam is perfectly welded according to design requirements. This high-precision welding is crucial for the performance and safety of aerospace parts.

[0069] In summary, this weld seam recognition method based on 3D scanning and deep learning represents an important development direction for intelligent manufacturing. It not only improves production efficiency and product quality but also lays the foundation for more flexible and intelligent manufacturing systems. The application of this technology is driving the entire manufacturing industry towards higher levels of intelligence and automation.

[0070] Step S102: Construct an undirected graph from the extracted geometric features, discretize the weld centerline into graph nodes, use the weld width as the graph edge weight, find the optimal connected path of the weld centerline on the graph using the minimum spanning tree algorithm, and construct a fine three-dimensional model of the processing object using the optimal connected path, wherein the fine three-dimensional model is marked with the three-dimensional coordinates of the weld finding point.

[0071] Optionally, in this embodiment, step S102 is a key step in the intelligent welding system, involving technical fields such as graph theory, minimum spanning tree algorithm, and 3D modeling. The main goal of this step is to transform the previously identified weld geometry features into an operable mathematical model, and then use this model to find the optimal welding path, ultimately constructing a refined 3D model. This process not only optimizes the welding path but also provides precise guidance for subsequent welding operations.

[0072] First, we need to construct an undirected graph from the extracted geometric features. This process essentially discretizes a continuous geometric problem into a graph theory problem. Specifically, we sample uniformly along the weld centerline, taking a point at regular intervals (e.g., 1 mm) as a node in the graph. These nodes represent key locations on the weld centerline. Then, we connect these nodes to form edges. Typically, we connect each node to other nodes within a certain radius (e.g., 5 mm). The purpose of this is to preserve the local geometric features of the weld while also providing sufficient choice space for subsequent path optimization.

[0073] In constructing the edges, we need to assign a weight to each edge. Here, we use the weld width as the edge weight. Specifically, for each edge connecting two nodes, we calculate the average weld width at those two nodes and use this value as the edge weight. The principle behind this weight setting is that weld width is generally proportional to welding difficulty and precision requirements; narrower welds often require more precise welding. In this way, we encode the important factor of welding difficulty into the graph model.

[0074] After constructing the undirected graph, the next step is to find the optimal connected path along the weld centerline. Here, we use the minimum spanning tree algorithm. The core idea of ​​the minimum spanning tree algorithm is to find the subgraph with the minimum total weight while ensuring the graph remains connected. In our scenario, this is equivalent to finding a path that connects all welding points while selecting the narrower (i.e., more difficult) sections of the weld.

[0075] Common minimum spanning tree algorithms include Kruskal's algorithm and Prim's algorithm. Taking Kruskal's algorithm as an example, its specific steps are as follows: First, sort all edges in the graph by weight in ascending order. Then, starting with the edge with the smallest weight, add each edge to the minimum spanning tree one by one. When adding each edge, check if it will form a cycle. If it will form a cycle, skip that edge. Repeat this process until all nodes are connected. The time complexity of this algorithm is O(ElogE) or O(ElogV), where E is the number of edges and V is the number of vertices. In practical implementations, we usually use a union-find data structure to accelerate the algorithm's operation.

[0076] Using the minimum spanning tree algorithm, we obtained the optimal connected path for the weld centerline. This path has several important characteristics: First, it connects all weld points, ensuring the integrity of the weld; second, it avoids possible repeated welding, improving efficiency; and finally, it tends to select paths with narrower weld widths, which usually implies higher welding precision requirements, helping to guarantee weld quality.

[0077] After obtaining the optimal connected path, we need to construct a detailed 3D model of the object to be processed based on this path. This process includes several key steps: the first is path smoothing. Since the path obtained by the minimum spanning tree algorithm is composed of discrete points, we need to use techniques such as spline interpolation or Bézier curves to connect these discrete points into a smooth curve. This ensures the continuity and smoothness of the welding path, which is beneficial for the motion control of the welding robot.

[0078] Next is surface reconstruction. We need to reconstruct the workpiece surface using algorithms such as Poisson surface reconstruction, based on the original point cloud data and the optimized weld path. The purpose of this step is to obtain a complete and accurate 3D model of the workpiece, providing a foundation for subsequent welding simulation and path planning.

[0079] Next comes weld modeling. Following the optimized path, we build a 3D model of the weld based on the weld width information at each location. This model includes not only the weld's location information but also its shape and dimensions, which are crucial for subsequent welding parameter settings.

[0080] Finally, we need to annotate the welding locating points in the model. These locating points typically include key locations such as the start point and corners of the weld. We need to accurately record the three-dimensional coordinates of these points, as this information will directly guide the movement of the welding robot.

[0081] This technical solution addresses several key technical challenges. First, it solves the problem of transforming complex geometric features into a mathematically optimized model. By converting weld features into a graph model, mature graph theory algorithms can be applied to optimize the welding path. Second, it addresses the problem of optimizing the welding sequence and path while ensuring weld integrity. The minimum spanning tree algorithm guarantees that all points requiring welding are connected, while also minimizing the overall welding difficulty. Finally, it solves the problem of transforming the optimized path into a detailed 3D model, providing precise guidance for subsequent welding operations.

[0082] In terms of technical effectiveness, this method brings significant improvements. Firstly, it enhances weld quality. By optimizing the welding path, the system prioritizes welding more challenging sections (such as narrower weld seams), helping to ensure consistent overall weld quality. Secondly, it improves efficiency. The optimal connection path avoids unnecessary re-welding and minimizes the distance the weld head travels, significantly increasing welding efficiency. For example, in welding a complex workpiece, this optimization might reduce total welding time by 20% to 30%.

[0083] Furthermore, this method improves the system's adaptability and flexibility. Regardless of the workpiece's complexity, the system can automatically generate the optimal welding path, significantly reducing the need for manual programming. For example, in a scenario of high-variety, low-batch production, when new workpieces require welding, the system can automatically generate an optimized welding path within minutes, whereas traditional methods might require hours or even days of manual programming.

[0084] Step S103: Based on the three-dimensional coordinates of the weld seam locating point, the current welding torch posture, and the target weld seam length, determine the optimal locating height, locating depth, tilt angle, and offset of the welding torch. Based on the locating height, locating depth, tilt angle, and offset, generate the corresponding welding path plan through a welding process simulation algorithm. Output the welding path plan to the industrial robot control system to drive the industrial robot to perform welding operations along the welding path.

[0085] Optionally, in this embodiment, step S103 is a key step in the intelligent welding system, involving technical fields such as welding torch posture optimization, welding path planning, and industrial robot control. The main objective of this step is to determine the optimal welding parameters based on the three-dimensional coordinates of the weld seam location obtained in previous steps, combined with information such as the current welding torch posture and the target weld seam length, and to generate the corresponding welding path plan, ultimately driving the industrial robot to perform the welding operation. This process not only optimizes welding quality but also improves welding efficiency and accuracy.

[0086] First, we need to determine the optimal positioning height of the welding torch. Positioning height refers to the vertical distance between the welding torch and the workpiece surface. Choosing this parameter requires weighing several factors: if the height is too low, it may cause the torch to collide with the workpiece; if the height is too high, it may affect welding accuracy and efficiency. Generally, we select a moderate height based on the geometry of the weld and the welding process requirements. For example, for thin plate welding, a lower positioning height (e.g., 5-10mm) might be chosen to improve accuracy; while for thick plate welding, a higher positioning height (e.g., 15-20mm) might be chosen to avoid collisions.

[0087] Next is determining the positioning depth. Positioning depth refers to the distance between the welding torch and the weld centerline in the horizontal direction. This parameter directly affects the quality and efficiency of the weld. If the depth is too shallow, it may lead to incomplete welding; if the depth is too deep, it may result in insufficient penetration. Typically, we determine the optimal positioning depth based on the width and depth of the weld, as well as the characteristics of the welding material. For example, for a V-weld, we might choose to position the welding torch at one-third of the weld opening.

[0088] The tilt angle of the welding torch is also an important parameter. The tilt angle affects the formation of the weld pool and the heat distribution. Typically, we select an appropriate tilt angle based on the welding position and weld type. For example, for flat welding, a forward tilt angle of 10-15 degrees might be chosen; while for vertical welding, a tilt angle of 0-5 degrees might be selected. In some cases, we also need to consider the side tilt angle to accommodate complex weld shapes.

[0089] Offset refers to the distance between the welding torch centerline and the weld centerline in the direction perpendicular to welding. Adjusting this parameter can compensate for weld position deviations caused by thermal deformation or other factors. Typically, we dynamically adjust the offset based on real-time feedback during the welding process. For example, if a 1mm offset is detected in the weld position, the system will automatically adjust the welding torch position to compensate for this deviation.

[0090] Once these parameters are determined, the next step is to generate a welding path plan using welding process simulation algorithms. This process involves complex numerical calculations and optimizations. First, we need to establish a physical model of the welding process, including heat conduction, fluid dynamics, and material phase transformations. Then, we use methods such as finite element analysis or computational fluid dynamics to solve this model to simulate the temperature distribution, stress distribution, and deformation during the welding process.

[0091] Based on the simulation results, we can further optimize the welding parameters and path. For example, if the simulation results show that certain areas are prone to porosity or cracking, we might adjust the welding speed or heat input, or change the welding sequence. This optimization process is usually iterative; we continuously adjust the parameters until we obtain satisfactory results.

[0092] The generated welding path plan includes not only the spatial trajectory of the welding torch, but also the changes in parameters such as welding speed, current, and voltage over time. This information is encoded into a sequence of instructions that the industrial robot can understand, and then output to the industrial robot control system.

[0093] After receiving the welding path plan, the industrial robot control system converts it into specific motion commands. This process involves complex kinematic and dynamic calculations. For example, for a 6-axis robot, the system needs to calculate the angle and angular velocity of each joint to achieve the specified trajectory of the end effector (welding torch). Simultaneously, the system also needs to consider the robot's dynamic characteristics, such as inertia and friction, to ensure the smoothness and precision of the motion.

[0094] During the welding process, the control system monitors welding parameters and robot status in real time and makes adjustments as needed. For example, if a deviation in the weld position is detected, the system adjusts the robot's trajectory in real time; if fluctuations in the welding current are detected, the system adjusts the power output accordingly. This real-time adjustment capability greatly improves the stability and quality of the welding process.

[0095] This technical solution addresses several key technical challenges. First, it solves the problem of determining the optimal welding torch posture in complex three-dimensional space. By comprehensively considering multiple parameters, the system can find the most suitable welding torch position and posture for each welding point. Second, it solves the problem of translating theoretical calculations into practically executable robot commands. Through precise kinematic and dynamic calculations, the system can convert complex welding paths into specific robot motion commands. Finally, it solves the problem of real-time adjustments during the welding process, improving the system's adaptability to environmental changes and process fluctuations.

[0096] In terms of technical effectiveness, this method brings significant improvements. Firstly, it enhances welding quality. By precisely controlling the welding torch posture and welding parameters, the system can significantly reduce welding defects and improve the strength and aesthetics of the weld. For example, in one application, the welding defect rate was reduced from 5% to less than 1% after adopting this method.

[0097] Secondly, there is the improvement in efficiency. Optimized welding paths and parameters can reduce unnecessary movement and rework, thereby improving production efficiency. In one practical case, the welding time for a complex workpiece was reduced from 2 hours to 1.5 hours, an increase in efficiency of 25%.

[0098] Furthermore, this method significantly improves the system's flexibility and adaptability. Whether welding new workpieces or responding to process changes, the system can quickly generate new welding solutions. This is particularly valuable in multi-variety, low-batch production. For example, on an automotive parts production line, when switching to a new product model, traditional methods might require several days of programming and debugging, while this intelligent system can complete the switch in just a few hours.

[0099] As described above, the welding path planning method provided in this application can perform three-dimensional scanning of the object to be processed using machine vision algorithms to obtain three-dimensional point cloud data. A deep learning-based segmentation algorithm is then used to analyze the three-dimensional point cloud data and identify geometric features. These geometric features are then used to construct an undirected graph. A minimum spanning tree algorithm is then used to find the optimal connected path for the weld centerline on this graph, and a detailed three-dimensional model of the object to be processed is constructed using this optimal connected path. Based on the three-dimensional coordinates of the weld search point, the current welding torch posture, and the target weld length, the optimal search height, search depth, tilt angle, and offset of the welding torch are determined. Based on the search height, search depth, tilt angle, and offset, a corresponding welding path plan is generated using a welding process simulation algorithm. This effectively improves the accuracy and efficiency of welding path planning.

[0100] In one embodiment of the welding path planning method of this application, see [link to relevant documentation]. Figure 2 It can also specifically include the following:

[0101] Step S201: Perform a 3D scan of the object to be processed using machine vision technology, and extract and organize 3D point cloud data from the raw data obtained from the scan.

[0102] Step S202: Perform data cleaning and data noise reduction preprocessing operations on the acquired 3D point cloud data to obtain 3D point cloud data after the preprocessing operations.

[0103] Optionally, in this embodiment, steps S201 and S202 are key steps in the intelligent manufacturing system, involving technical fields such as machine vision, 3D scanning, and point cloud processing. The main objective of these two steps is to acquire accurate 3D data of the object to be processed and to perform preliminary processing on this data, providing a foundation for subsequent analysis and processing. This process not only improves the accuracy and efficiency of data acquisition but also lays the foundation for subsequent intelligent processing.

[0104] First, in step S201, we use machine vision technology to perform a 3D scan of the object to be processed. This process typically involves the combined application of multiple scanning technologies to obtain comprehensive and accurate information about the object. Commonly used 3D scanning technologies include structured light scanning, laser triangulation, and time-of-flight (TOF).

[0105] Taking structured light scanning as an example, the system projects a specific grating pattern onto the object's surface and then captures the deformation of this pattern on the surface using a camera. By analyzing these deformations, the system can calculate the three-dimensional coordinates of each point on the object's surface. This method is advantageous due to its high speed and accuracy, making it particularly suitable for scanning objects with complex shapes.

[0106] For objects made of special materials or with complex shapes, it may be necessary to combine multiple scanning techniques. For example, for highly reflective metal surfaces, special coatings or multi-angle scanning may be required to reduce interference from reflections. For objects with internal structures, X-ray computed tomography (CT) may be used to obtain internal information.

[0107] During the scanning process, the system typically employs a multi-angle, multiple-scan strategy to ensure comprehensive information about the object is acquired. For example, a complex industrial part may require scanning from 20-30 different angles, each generating a set of point cloud data. These point cloud data from different angles then need to be aligned and fused using a registration algorithm (such as the ICP algorithm) to form a complete 3D point cloud model.

[0108] After scanning is complete, the system performs preliminary processing on the raw data, extracting and organizing it into a standard 3D point cloud data format. This process includes operations such as coordinate system transformation and data format conversion. For example, it may be necessary to transform data in the device coordinate system to the global coordinate system, or convert proprietary scan data to a common point cloud data format (such as .ply or .pcd format).

[0109] In step S202, we perform preprocessing operations on the acquired 3D point cloud data, mainly including data cleaning and noise reduction. The purpose of this step is to improve data quality, remove various noises and outliers generated during the scanning process, and provide a more reliable data foundation for subsequent analysis and processing.

[0110] Data cleaning primarily addresses the following issues: First, it removes outliers. These outliers are typically caused by random errors during the scanning process or environmental interference, and they are often far from the actual object surface. Common methods include Statistical Outlier Removal (SOR) and Radius Outlier Removal (ROR). The SOR method determines whether a point is an outlier based on the average distance from each point to its neighbors, while the ROR method determines outliers by checking the number of neighbors within a given radius for each point.

[0111] Secondly, handling duplicate points is crucial. During multiple scans and registrations, duplicate sampling may occur, leading to abnormally high point density in certain areas. We typically use methods such as voxelization or uniform sampling to address this issue. Voxelization divides the space into fixed-size cubes (voxels), then uses the average value of all points within each voxel to represent that voxel. This not only removes duplicate points but also reduces the amount of data to some extent.

[0112] Finally, we fill in the voids. Due to occlusion or surface material, small voids may appear in the scan results. These voids can be filled using methods such as interpolation or surface fitting. Commonly used methods include Moving Least Squares (MLS) and Poisson Surface Reconstruction.

[0113] Data denoising primarily addresses random noise caused by equipment precision limitations or environmental interference. Commonly used denoising methods include bilateral filtering and Gaussian filtering. The core idea behind these methods is to utilize neighborhood information to smooth the data while preserving the edges and details of objects as much as possible. For example, bilateral filtering considers not only the spatial distance between points but also the similarity of their values, thus smoothing noise while maintaining edge sharpness.

[0114] In practical applications, we typically choose an appropriate combination of preprocessing methods based on the specific circumstances. For example, for a complex industrial part, we might first use the SOR method to remove outliers, then apply voxelization to reduce the data volume, and finally use bilateral filtering for noise reduction. The entire preprocessing process may require multiple iterations and parameter adjustments to achieve the best results.

[0115] This technical solution addresses several key technical challenges. First, it solves the problem of how to quickly and accurately acquire the geometric information of complex 3D objects. By combining multiple scanning techniques and multi-angle scanning, the system can obtain comprehensive and accurate information about the object, including complex shapes and internal structures that are difficult to obtain using traditional measurement methods.

[0116] Secondly, it solves the problem of handling large-scale, noisy point cloud data. Through a series of data cleaning and denoising operations, the system can significantly improve the quality of point cloud data, providing a reliable data foundation for subsequent analysis and processing. This is crucial for improving the accuracy and efficiency of subsequent processing.

[0117] Finally, it addresses the challenge of reducing data volume while maintaining data accuracy. Through appropriate data compression and sampling strategies, the system can significantly reduce data volume while preserving key geometric features, which is crucial for improving the efficiency of subsequent processing.

[0118] In terms of technical effectiveness, this method brings significant improvements. Firstly, it improves data acquisition accuracy. Compared to traditional manual measurement, this method can acquire more comprehensive and accurate three-dimensional data. For example, in measuring a precision mechanical part, this method can control the measurement error to within 0.01 mm, while the error of traditional manual measurement may be around 0.1 mm.

[0119] Secondly, there is a significant increase in efficiency. Automated 3D scanning and data processing can greatly reduce manual operation time. In one practical case, the measurement and data processing time for a complex workpiece was reduced from several hours to just a few minutes, improving efficiency by tens of times.

[0120] Furthermore, this method improves the availability and reliability of the data. Preprocessed point cloud data is clearer and more accurate, and can be directly used for subsequent analysis and processing. This not only improves the quality of subsequent processing but also reduces rework and waste caused by data issues.

[0121] In one embodiment of the welding path planning method of this application, see [link to relevant documentation]. Figure 3 It can also specifically include the following:

[0122] Step S301: Input the three-dimensional point cloud data into a set deep learning-based semantic segmentation algorithm for fine segmentation;

[0123] Step S302: Extract the geometric features of the weld area from the output of the algorithm, wherein the geometric features include at least one of the centerline, width, and shape of the weld.

[0124] Optionally, in this embodiment, steps S301 and S302 are key steps in the intelligent manufacturing system, involving technical fields such as deep learning, point cloud processing, semantic segmentation, and geometric feature extraction. The main goal of these two steps is to perform fine segmentation of the 3D point cloud data and extract key geometric features of the weld area from it. This process not only improves the accuracy and efficiency of weld identification but also provides important basic data for subsequent welding planning and quality control.

[0125] In step S301, we first input the preprocessed 3D point cloud data into a pre-defined deep learning-based semantic segmentation algorithm. This algorithm typically employs deep neural network structures, such as PointNet++, DGCNN, or the latest Transformer-based models. These models can directly process irregular point cloud data without converting it into a regular voxel mesh.

[0126] Taking PointNet++ as an example, the algorithm first organizes point cloud data into a multi-layered structure through sampling and grouping operations. Then, it uses a shared multilayer perceptron (MLP) at each layer to extract local features. These features are then aggregated through max pooling to obtain global features. Finally, semantic segmentation is achieved by classifying each point through point-by-point prediction.

[0127] During the training phase, we need to prepare a large amount of labeled point cloud data. This data should include various types of welds, as well as other features that may be confused with welds (such as edges, grooves, etc.). Through the backpropagation algorithm, the model can learn to distinguish different semantic categories. To improve the model's generalization ability, we usually use data augmentation techniques, such as random rotation, translation, scaling, etc.

[0128] In practical applications, to further improve segmentation accuracy, we might employ ensemble learning methods. For example, we can train multiple models with different structures or initializations, and then fuse their predictions through voting or weighted averaging. This approach can effectively reduce the error and bias of individual models.

[0129] Furthermore, to handle large-scale point cloud data, we may need to adopt a block processing strategy. This involves dividing the entire point cloud data into multiple overlapping smaller blocks, processing them separately, and then merging the results. This approach not only reduces memory consumption but also enables parallel processing, significantly improving processing speed.

[0130] In step S302, we need to extract the geometric features of the weld region from the output of semantic segmentation. This process typically includes the following sub-steps:

[0131] The first step is accurate localization of the weld area. Although semantic segmentation algorithms have already assigned a category to each point, there may be some noise and misclassifications. We can use morphological operations (such as dilation and erosion) to optimize the segmentation results, remove isolated misclassified points, and fill in small gaps.

[0132] Next is the extraction of the weld centerline. For straight welds, we can use the RANSAC (Random Sampling Consensus) algorithm to fit a straight line. For curved welds, we can use the principal curve algorithm or B-spline fitting. In practical applications, we may need to incorporate prior knowledge to improve the accuracy of the fitting. For example, if we know that the weld is roughly arc-shaped, we can directly use an arc fitting algorithm.

[0133] Measuring weld width typically involves cross-sectional analysis. We can take multiple cross-sections perpendicular to the weld centerline and analyze each one. At each cross-section, we can measure the width by locating the boundary points of the weld region. To improve stability, we usually take the average of multiple cross-sections as the final width estimate.

[0134] Identifying weld shape requires more complex analysis. We can determine the weld type (e.g., V-shaped, U-shaped, butt-joint, etc.) by analyzing features such as point distribution and curvature changes in the weld area. In some cases, we may need to use template matching to compare the extracted weld contour with a predefined template to determine the best matching shape type.

[0135] In practical applications, to improve the robustness of feature extraction, we often employ multiple methods combined with statistical analysis. For example, for weld width measurement, we might simultaneously use methods based on cross-section analysis and region growing, and then fuse the results of different methods through median filtering or weighted averaging.

[0136] This technical solution addresses several key technical challenges. First, it solves the problem of accurately identifying and locating weld seam areas from complex 3D point cloud data. Traditional rule-based methods often struggle with complex industrial scenarios, while deep learning-based methods can learn more complex and abstract features, thus achieving higher accuracy in identification.

[0137] Secondly, it addresses the challenge of automatically extracting key geometric features of welds. These features are crucial for subsequent welding planning and quality control. By combining deep learning with traditional image processing techniques, the system can extract these features efficiently and accurately, significantly reducing the need for manual intervention.

[0138] Finally, it addresses the challenge of handling welds of various complex shapes and types. By employing flexible deep learning models and diverse feature extraction algorithms, the system can adapt to diverse weld conditions, improving its versatility and adaptability.

[0139] In terms of technical effectiveness, this method brings significant improvements. Firstly, the recognition accuracy is greatly enhanced. Compared to traditional rule-based methods, this deep learning-based method can more accurately identify and segment weld seam areas. In one practical case, the accuracy of weld seam recognition increased from 85% to 98%, significantly reducing false positives and false negatives.

[0140] Secondly, there is a significant improvement in processing efficiency. The automated feature extraction process greatly reduces the need for manual intervention. In a large industrial project, using this method reduced the time for weld identification and feature extraction from several hours to just a few minutes, improving efficiency by tens of times.

[0141] Furthermore, this approach enhances the system's flexibility and adaptability. Due to the powerful generalization capabilities of deep learning models, the system can better adapt to different types of welds and varying working environments. This is particularly valuable when handling new workpieces or facing production line changes.

[0142] Finally, this method provides more reliable and accurate input data for subsequent intelligent welding. Accurate weld geometry information enables more precise welding path planning and parameter optimization, thereby improving welding quality and reducing rework and material waste.

[0143] In one embodiment of the welding path planning method of this application, see [link to relevant documentation]. Figure 4 It can also specifically include the following:

[0144] Step S401: Construct an undirected graph from the centerline and width of the geometric features, discretize the centerline as a node of the graph, and use the width as the weight of the graph edge;

[0145] Step S402: Apply the minimum spanning tree algorithm to the constructed undirected graph model to determine the shortest path connecting all nodes as the optimal connected path for the weld centerline.

[0146] Optionally, in this embodiment, steps S401 and S402 are key steps in the intelligent welding system, involving technical fields such as graph theory, path planning, and optimization algorithms. The main goal of these two steps is to transform the geometric features of the weld into a computable mathematical model, and on this basis, determine the optimal welding path. This process not only improves the efficiency and accuracy of welding path planning, but also provides important guidance for subsequent welding operations.

[0147] In step S401, we first construct an undirected graph model of the weld's geometric features (centerline and width). This process essentially discretizes a continuous geometric problem into a discrete mathematical problem, allowing us to apply mature graph theory algorithms to solve the path planning problem.

[0148] Specifically, we first need to discretize the centerline of the weld. This is typically achieved through equally spaced sampling. For example, we can take a point every 1 mm along the centerline as a node in the graph. The sampling interval needs to be determined based on the complexity of the weld and the required accuracy. For complex curved welds, denser sampling may be needed; while for simple straight welds, relatively sparse sampling can be used.

[0149] Each sampling point becomes a node in the graph. Adjacent nodes are connected by an edge, which represents a small segment of the weld. The weight of the edge is determined by the weld width at the corresponding location. Here, we can use the average width, or we can use more complex methods, such as considering the rate of change of width.

[0150] In practical applications, we may need to consider other factors to determine the edge weights. These include weld depth, variations in material thickness, and welding difficulty. These factors can be integrated into a single weighting function, allowing the graphical model to more comprehensively reflect the complexity of the welding process.

[0151] Furthermore, to improve the robustness of the model, we may need to handle some special cases. For example, for welds with branches, we need to add additional connections in the graph; for welds with discontinuities, we may need to add virtual high-cost edges to connect the different parts.

[0152] In step S402, we apply the minimum spanning tree algorithm to the constructed undirected graph model to determine the shortest path connecting all nodes. This path will serve as the optimal connected path for the weld centerline, providing guidance for subsequent welding operations.

[0153] The minimum spanning tree algorithm is a classic algorithm in graph theory. It finds the tree structure with the minimum total weight while ensuring that all nodes are connected. Commonly used minimum spanning tree algorithms include Kruskal's algorithm and Prim's algorithm. In this application scenario, Prim's algorithm may be more suitable because it is easier to handle weighted graphs.

[0154] The basic idea of ​​Prim's algorithm is to start from a primary node and, at each step, select the edge with the smallest weight that connects to the current tree, until all nodes are included in the tree. In our application, this means that the algorithm will prioritize weld segments with smaller widths (or smaller overall weights), which typically correspond to sections that are easier to weld or require more precise control.

[0155] In practical applications, we may need to modify and optimize the standard Prim algorithm. For example, we might need to add constraints, such as the start and end points of the weld, or avoid certain areas. This can be achieved by adding additional decision conditions to the algorithm or modifying the edge weights.

[0156] Furthermore, to handle large-scale weld diagram models, we may need to use efficient data structures, such as Fibonacci heaps, to improve the algorithm's efficiency. For particularly complex welds, we may also need to consider using parallel computing techniques to accelerate the computation process.

[0157] This technical solution addresses several key technical challenges. First, it solves the problem of transforming complex weld geometry into a computable mathematical model. By discretizing the continuous geometric problem into a graph theory problem, we can leverage established graph theory algorithms to solve the path planning problem, significantly simplifying its complexity.

[0158] Secondly, it addresses the problem of determining the optimal welding path when considering multiple factors (such as weld width, depth, and material properties). By incorporating these factors into the edge weights of the graph, we can obtain the optimal path using a single minimum spanning tree algorithm.

[0159] Finally, it solves the path planning problem for handling complex welds (such as those with branches or discontinuities). By adding extra nodes and edges to the graph model, we can flexibly handle various complex situations.

[0160] In terms of technical effectiveness, this method brings significant improvements. Firstly, the efficiency of path planning is greatly enhanced. Compared to traditional manual planning or simple geometric methods, this graph theory-based method can obtain the globally optimal path in a very short time. In one practical case, the welding path planning time for a complex workpiece was reduced from several hours to just a few seconds, improving efficiency by hundreds of times.

[0161] Secondly, welding quality is improved. Because the path planning takes into account characteristics such as weld width, the welding process can better adapt to changes in the weld. For example, in narrower weld sections, the welding speed can be appropriately reduced to ensure sufficient penetration; while in wider weld sections, the speed can be appropriately increased to improve efficiency. This refined control greatly improves the quality and consistency of the welding.

[0162] Furthermore, this approach enhances the system's flexibility and adaptability. Because graphical models allow for the easy addition of various constraints, the system can readily handle diverse special requirements, such as specifying start and end points or avoiding certain areas. This is particularly valuable when dealing with complex workpieces or special welding requirements.

[0163] Finally, this method provides a solid foundation for subsequent intelligent welding systems. The optimal connection path can not only be directly used to guide the movement of the welding robot, but also serve as a basis for further optimizing welding parameters. For example, parameters such as welding current, voltage, and wire feed speed can be dynamically adjusted based on the characteristics of different positions along the path, achieving more intelligent and precise welding control.

[0164] In one embodiment of the welding path planning method of this application, see [link to relevant documentation]. Figure 5 It can also specifically include the following:

[0165] Step S501: Integrate the obtained optimal connection path of the weld centerline into the original three-dimensional geometric model of the processing object to generate a fine three-dimensional model, wherein the fine three-dimensional model contains geometric details of the weld path;

[0166] Step S502: Mark the three-dimensional coordinate information of the weld seam locating point in the refined three-dimensional model.

[0167] Optionally, in this embodiment, steps S501 and S502 are key steps in the intelligent welding system, involving technical fields such as 3D modeling, data fusion, and coordinate annotation. The main goal of these two steps is to fuse the previously obtained weld path information with the original 3D geometric model and mark the key weld locating points in the fused model. This process not only improves the accuracy and usability of the model but also provides more detailed and reliable guidance information for subsequent welding operations.

[0168] In step S501, we first need to integrate the obtained optimal connectivity path of the weld centerline into the original three-dimensional geometric model of the workpiece. This process is actually a complex three-dimensional data fusion process, which requires consideration of multiple technical details and challenges.

[0169] First, we need to ensure that the weld path is precisely aligned spatially with the original model. This typically requires point cloud registration techniques. For example, we can use the Iterative Closest Point (ICP) algorithm to achieve accurate spatial alignment. In some cases, if the weld path is directly extracted from the original model, this step may be omitted.

[0170] Next, we need to "embed" the weld path into the original model. This process typically involves local modifications and reconstructions of the model. Specifically, we might need to "groove" the surface of the original model to accommodate the weld path. This process can be achieved using Boolean operations or more complex geometric processing algorithms.

[0171] In practical applications, we may also need to consider the depth and shape of the weld. For example, for a V-weld, we need to create the corresponding V-groove in the model; for a lap weld, we may need to adjust the relative positions of the two components. Handling these details is crucial for generating a realistic and usable, detailed model.

[0172] Furthermore, to improve the accuracy and usability of the model, we may need to perform some additional processing. For example, we might need to smooth the fused model to eliminate any potential discontinuities or sharp edges. We might also need to perform mesh optimization to ensure that the model has sufficient detail in the weld area while maintaining a reasonable level of simplification in other areas.

[0173] When dealing with large or complex models, we may need to use a chunked processing strategy. This involves dividing the entire model into multiple smaller chunks, processing each chunk separately, and then merging the results. This approach not only reduces memory consumption but also enables parallel processing, significantly improving processing speed.

[0174] In step S502, we need to annotate the 3D coordinates of the weld seam locating points in the generated detailed 3D model. This process is crucial for subsequent actual welding operations because it provides the key information required for welding robot positioning and path planning.

[0175] First, we need to determine the selection strategy for weld locator points. Generally, we distribute some points evenly along the weld path as locator points. However, in practical applications, we may need to consider more factors. For example, denser locator points may be needed at the start, end, and corners of the weld path; additional locator points may also be needed where the weld width or depth changes significantly.

[0176] Next, we need to accurately calculate the 3D coordinates of these locating points. This typically involves precise geometric calculations on a 3D model. For straight line segments, we can use simple linear interpolation; for curved segments, we may need to use more complex curve parametric methods.

[0177] In practical applications, we may need to calculate additional information for each locating point. For example, the weld seam normal vector at that point, the weld seam width, and recommended welding parameters. This information helps the welding robot to more accurately position and adjust welding parameters.

[0178] To improve the reliability of positioning, we may need to add some auxiliary positioning points around each primary positioning point. These auxiliary points can help the welding system correct itself when the primary positioning point is difficult to identify.

[0179] Furthermore, to facilitate subsequent use, we need to store and output this location information in a suitable format. This may involve specific data structure design to ensure the integrity and usability of the information.

[0180] This technical solution addresses several key technical challenges. First, it solves the problem of accurately integrating one-dimensional weld path information into a three-dimensional geometric model. This process not only requires consideration of spatial alignment but also the handling of complex geometric modifications, making it a highly challenging problem.

[0181] Secondly, it solves the problem of accurately marking weld location points in complex 3D models. This requires considering multiple factors such as the geometric characteristics of the weld and welding process requirements, and is a key step in achieving precise welding.

[0182] Finally, it solves the problem of how to generate a comprehensive 3D model that includes both original geometric information and welding details. This model can be used not only for welding planning but also for visualization, simulation, and many other applications, greatly enhancing its value and scope of use.

[0183] In terms of technical effectiveness, this method brings significant improvements. Firstly, it greatly enhances model accuracy. The refined 3D model, incorporating weld path information, can more accurately reflect the geometric features of the actual workpiece. In one practical case, the geometric error between the model generated using this method and the actual workpiece was reduced by 80%, from millimeter-level error to sub-millimeter level.

[0184] Secondly, the efficiency and accuracy of welding planning are significantly improved. With precise weld seam location information, the welding robot can position the weld seam more quickly and accurately. In a large-scale engineering project, using this method reduced welding positioning time by 70% and improved positioning accuracy by 50%.

[0185] Furthermore, this method significantly improves the controllability and repeatability of the welding process. Because the model contains detailed weld geometry information, the welding system can dynamically adjust welding parameters based on this information. For example, where the weld width changes, the system can automatically adjust the welding speed or power, thereby ensuring consistent weld quality.

[0186] Finally, this detailed 3D model provides a unified information foundation for the entire manufacturing process. It can be used not only for welding planning but also for quality inspection, process optimization, and many other aspects. For example, after welding is completed, this model can be used as a reference for 3D scanning and comparison to quickly identify potential welding defects.

[0187] In one embodiment of the welding path planning method of this application, see [link to relevant documentation]. Figure 6 It can also specifically include the following:

[0188] Step S601: Extract the three-dimensional coordinates of the weld search point marked in the fine three-dimensional model, and analyze the geometric relationship between the search point coordinates, the welding gun posture and the weld length by combining the actual posture of the current welding gun and the expected weld length target.

[0189] Step S602: Based on the analysis results of the geometric relationship, calculate the optimal positioning height, optimal positioning depth, tilt angle, and offset of the welding torch using an optimization algorithm.

[0190] Optionally, in this embodiment, steps S601 and S602 are core components of the intelligent welding system, involving technical fields such as spatial geometry analysis, posture optimization, and path planning. The main objective of these two steps is to calculate the optimal welding torch positioning parameters based on the weld seam positioning information in the detailed 3D model and the actual conditions of the welding torch. This process requires consideration not only of spatial geometric relationships but also of welding process requirements and equipment limitations, making it a crucial step in achieving high-quality welding.

[0191] In step S601, we first need to extract the 3D coordinates of the weld seam location points from the detailed 3D model. This process typically involves database queries or file parsing operations, depending on the storage format of the model data. For example, if using common CAD file formats such as STEP or IGES, we may need to use a specialized CAD parsing library to extract this information.

[0192] After extracting the coordinates, we need to analyze the geometric relationship between the location coordinates, the welding torch posture, and the weld length, combining the current actual welding torch posture and the expected weld length target. This analysis process involves complex spatial geometric calculations.

[0193] First, we need to establish a unified coordinate system. Typically, we choose the workpiece coordinate system as the reference and then transform the coordinates and orientation information of the welding torch into this coordinate system. This transformation process involves calculating the rigid body transformation matrix, which requires consideration of both translation and rotation.

[0194] Next, we need to analyze the relative position and orientation between the welding torch and the weld. This typically involves vector calculations. For example, we can calculate the angle between the welding torch axis and the weld tangent, or the shortest distance from the welding torch tip to the weld. These calculations will provide the foundational data for subsequent optimization.

[0195] During the analysis, we also need to consider the geometric characteristics of the weld. For example, we may need to use different analysis methods for planar welds and curved welds. For curved welds, we may need to calculate the local normal vector and curvature of the surface, which may involve knowledge of differential geometry.

[0196] In addition, we need to consider the target weld length. This typically involves calculating path integrals. We need to integrate along the weld path, calculate the cumulative length, and determine the endpoint where the length requirement is met.

[0197] In step S602, based on the analysis results of S601, we need to calculate the optimal positioning parameters of the welding torch using an optimization algorithm, including height, depth, tilt angle, and offset. This is a multi-objective optimization problem that requires balancing multiple factors.

[0198] First, we need to define the optimization objective function. This function typically consists of several parts, such as:

[0199] Minimize the distance deviation between the welding torch and the weld.

[0200] Minimize the angular deviation between the welding torch posture and the ideal posture

[0201] Maximize the stability of the welding process (e.g., avoid collisions between the welding torch and the workpiece).

[0202] Minimize the amount of movement of the welding torch (to improve efficiency).

[0203] Next, we need to define the constraints. These constraints may include:

[0204] Welding torch movement range limitation

[0205] Angle range required for welding process

[0206] Space constraints to avoid collisions with workpieces or other equipment

[0207] Based on these objective functions and constraints, we can apply various optimization algorithms to solve for the optimal parameters. Commonly used algorithms include gradient descent, genetic algorithms, and particle swarm optimization. The choice of algorithm usually depends on the specific characteristics of the problem and the limitations of computational resources.

[0208] In practical applications, we may need to perform segmented optimization. For complex welds, we can divide them into multiple segments, optimize each segment separately, and then comprehensively consider the overall continuity and smoothness.

[0209] Furthermore, to improve the efficiency and robustness of optimization, we may need to employ some heuristic strategies. For example, we can set some initial values ​​based on experience, or use multiple random initializations to avoid getting trapped in local optima.

[0210] This technical solution addresses several key technical challenges. First, it solves the problem of accurately analyzing the geometric relationship between the welding torch and the weld seam in complex three-dimensional space. This involves complex spatial geometric calculations and coordinate transformations, which are fundamental to achieving precise welding.

[0211] Secondly, it addresses the problem of finding optimal welding parameters under multiple objectives and constraints. This is a typical multi-objective optimization problem, requiring a balance of factors such as efficiency, quality, and stability.

[0212] Finally, it addresses the problem of how to translate theoretical calculations into practically usable welding torch control parameters. This requires considering the characteristics and limitations of actual equipment and is a key aspect of combining theory with practice.

[0213] In terms of technical effectiveness, this method brings significant improvements. Firstly, there is a substantial increase in weld quality. Through precise geometric analysis and parameter optimization, the welding torch can always maintain the optimal position and orientation. In a welding project for an aerospace component, using this method reduced the weld defect rate by 90% and increased weld strength by 30%.

[0214] Secondly, welding efficiency is significantly improved. The optimized welding torch path minimizes unnecessary movement while ensuring weld quality. On an automotive manufacturing production line, applying this method reduced welding time by 25% and energy consumption by 20%.

[0215] Furthermore, this method significantly improves the adaptability and robustness of the welding process. Because the system can analyze and optimize welding torch parameters in real time, it is better able to handle minor workpiece deformations or positioning errors. In a large steel structure project, using this method improved the weld pass rate by 40%, even under harsh field conditions.

[0216] Finally, this method provides an important technological foundation for achieving fully automated welding. Through precise geometric analysis and parameter optimization, the system can automatically adapt to different welding tasks, significantly reducing the need for manual intervention. On a fully automated welding production line, applying this method reduced manual supervision time by 80% and increased production efficiency by 50%.

[0217] In one embodiment of the welding path planning method of this application, see [link to relevant documentation]. Figure 7 It can also specifically include the following:

[0218] Step S701: Input the positioning height, positioning depth, tilt angle and offset from the obtained welding torch positioning parameters into the preset welding process simulation algorithm for calculation;

[0219] Step S702: Output a welding path plan that meets the requirements based on the simulation results.

[0220] Optionally, in this embodiment, steps S701 and S702 are crucial steps in the intelligent welding system, involving technical fields such as welding process simulation, path planning, and optimization. The main goal of these two steps is to verify the previously optimized welding torch positioning parameters through simulation calculations and ultimately generate a welding path plan that meets the requirements. This process not only needs to consider the complexity of the welding process but also the feasibility of actual operation, making it a key step in ensuring high-quality welding.

[0221] In step S701, we first need to input the obtained welding torch positioning parameters, including positioning height, positioning depth, tilt angle, and offset, into a preset welding process simulation algorithm for calculation. This process involves complex physical models and numerical calculation methods.

[0222] Welding process simulation algorithms are typically based on methods such as finite element analysis (FEA) or computational fluid dynamics (CFD). These algorithms need to consider various physical processes, including heat conduction, phase change, fluid flow, and electromagnetic fields. For example, when simulating arc welding, we need to consider the heat input of the arc, the formation and flow of the molten pool, and the melting and solidification processes of the metal.

[0223] In practice, we may need to use specialized welding simulation software, such as SYSWELD, ABAQUS, or a self-developed simulation program. These software programs typically offer rich material libraries and process parameter libraries, which can be configured according to the specific welding task.

[0224] During the simulation, we need to set a series of parameters, including:

[0225] Material properties: such as thermal conductivity, specific heat capacity, density, etc., these parameters may change with temperature. Welding process parameters: such as current, voltage, welding speed, etc.

[0226] Environmental conditions: such as ambient temperature, heat dissipation conditions, etc.

[0227] Geometric model: includes the geometry of the workpiece and the specific location of the weld.

[0228] Simulation calculations are typically an iterative process. We may need to perform multiple simulations, adjusting parameters until we obtain satisfactory results. During this process, we need to pay attention to several indicators, such as weld penetration, weld width, heat-affected zone size, and residual stress distribution.

[0229] Furthermore, to improve the efficiency and accuracy of the simulation, we may need to employ some optimization strategies. For example, we can use adaptive meshing techniques, using a finer mesh in critical areas and a coarser mesh in non-critical areas. We can also use parallel computing techniques, leveraging multi-core processors or GPUs to accelerate the computation process.

[0230] In step S702, we need to output a welding path plan that meets the requirements based on the simulation results. This process involves multiple stages, including result analysis, path optimization, and feasibility verification.

[0231] First, we need to analyze the simulation results to evaluate the welding quality. This typically includes checking whether the penetration depth is sufficient, whether the weld pool shape is appropriate, and whether there are any defects (such as porosity or cracks). If the simulation results do not meet the requirements, we may need to return to step S701, adjust the parameters, and resimulate.

[0232] If the simulation results meet the requirements, we can generate welding paths based on these results. Path generation needs to consider several factors:

[0233] Weld geometry: The path must accurately follow the shape of the weld.

[0234] Welding speed: The speed may need to be adjusted at different locations to ensure uniform heat input.

[0235] Welding torch posture: Appropriate working angle and forward tilt angle need to be maintained.

[0236] Arc initiation and termination: Special consideration needs to be given to the start and end points of the welding.

[0237] Multi-pass welding: For welding thick plates, it may be necessary to plan the sequence and location of multiple weld passes.

[0238] When generating paths, we typically use mathematical methods such as spline curves or piecewise linear interpolation to describe them. To ensure the smoothness and continuity of the paths, we may need to apply some smoothing algorithms.

[0239] The generated path also needs to be verified for feasibility. This includes checking whether the path is within the welding robot's working range, whether there are potential collision points, and whether the welding torch's posture changes too drastically. If problems are found, the path needs to be adjusted.

[0240] Finally, we need to convert the generated path into a sequence of instructions that the welding robot can execute. This typically involves inverse kinematics and trajectory planning algorithms for the robot.

[0241] This technical solution addresses several key technical challenges. First, it solves the problem of accurately predicting welding results before actual welding. Through sophisticated physical models and numerical simulations, we can evaluate and optimize welding parameters in a virtual environment, significantly reducing the need for actual testing.

[0242] Secondly, it addresses the problem of generating the optimal welding path based on simulation results. This process requires comprehensive consideration of multiple factors, including welding quality, efficiency, and feasibility, making it a complex multi-objective optimization problem.

[0243] Finally, it addresses the challenge of translating theoretical models and optimization results into practically executable welding commands. This requires considering the limitations and characteristics of actual equipment and is a key aspect of combining theory with practice.

[0244] In terms of technical effectiveness, this method brings significant improvements. Firstly, there is a substantial increase in welding quality. Through precise simulation and optimization, we can predict and address potential problems before actual welding. In a welding project for an aero-engine component, using this method reduced the welding defect rate by 95% and increased the welding strength by 40%.

[0245] Secondly, welding efficiency is significantly improved. The optimized welding path minimizes unnecessary movement and stoppages while ensuring weld quality. On an automotive body welding production line, applying this method reduced welding time by 30% and energy consumption by 25%.

[0246] Furthermore, this method significantly reduces the time and cost of welding parameter debugging. Traditional parameter debugging typically requires multiple actual tests, which is time-consuming and labor-intensive. However, through simulation optimization, we can complete this process quickly in a virtual environment. In a new product development project, applying this method reduced welding parameter debugging time by 80% and material waste by 70%.

[0247] Finally, this method provides an important technological foundation for achieving intelligent and flexible production. Through simulation and optimization, the system can quickly adapt to different welding tasks, greatly improving production flexibility. In a factory producing multiple varieties in small batches, applying this method reduced product changeover time by 60% and increased production efficiency by 45%.

[0248] In summary, this simulation-based optimization-based welding path planning method represents an important development direction for intelligent welding systems. It not only improves welding quality and efficiency but also enables more intelligent and flexible production processes. The application scope of this technology is very broad, benefiting everything from precision electronics manufacturing to large-scale engineering structure construction, and it is driving the entire manufacturing industry towards higher levels of intelligence and automation.

[0249] To effectively improve the accuracy and efficiency of welding path planning, this application provides an embodiment of a welding path planning device for implementing all or part of the aforementioned welding path planning method. See [link to embodiment]. Figure 8 The welding path planning device specifically includes the following components:

[0250] The 3D construction module 10 is used to perform 3D scanning of the object to be processed using machine vision algorithms to obtain 3D point cloud data of the object to be processed, and to perform data analysis on the 3D point cloud data by setting a segmentation algorithm based on deep learning to identify the geometric features of the weld area on the surface of the object to be processed.

[0251] The coordinate determination module 20 is used to construct an undirected graph from the extracted geometric features, discretize the weld centerline into graph nodes, use the weld width as the graph edge weight, find the optimal connected path of the weld centerline on the graph using the minimum spanning tree algorithm, and construct a fine three-dimensional model of the processing object using the optimal connected path, wherein the fine three-dimensional model is marked with the three-dimensional coordinates of the weld finding point.

[0252] The path planning module 30 is used to determine the optimal positioning height, positioning depth, tilt angle, and offset of the welding torch based on the three-dimensional coordinates of the weld seam positioning point, the current welding torch posture, and the target weld seam length. Based on the positioning height, the positioning depth, the tilt angle, and the offset, the module generates a corresponding welding path plan through a welding process simulation algorithm and outputs the welding path plan to the industrial robot control system to drive the industrial robot to perform welding operations along the welding path.

[0253] As described above, the welding path planning device provided in this application embodiment can perform three-dimensional scanning of the object to be processed using machine vision algorithms to obtain three-dimensional point cloud data. It then analyzes the three-dimensional point cloud data using a deep learning-based segmentation algorithm to identify geometric features. These geometric features are then used to construct an undirected graph. A minimum spanning tree algorithm is used to find the optimal connected path for the weld centerline on this graph, and a detailed three-dimensional model of the object to be processed is constructed using this optimal connected path. Based on the three-dimensional coordinates of the weld search point, the current welding torch posture, and the target weld length, the optimal search height, search depth, tilt angle, and offset of the welding torch are determined. Based on the search height, search depth, tilt angle, and offset, a corresponding welding path plan is generated using a welding process simulation algorithm. This effectively improves the accuracy and efficiency of welding path planning.

[0254] From a hardware perspective, in order to effectively improve the accuracy and efficiency of welding path planning, this application provides an embodiment of an electronic device for implementing all or part of the welding path planning method, wherein the electronic device specifically includes the following:

[0255] The system comprises a processor, memory, a communications interface, and a bus; wherein the processor, memory, and communications interface communicate with each other via the bus; the communications interface is used to realize information transmission between the welding path planning device and core business systems, user terminals, and related databases and other related devices; the logic controller can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the logic controller can be implemented with reference to the embodiments of the welding path planning method and the welding path planning device in the embodiments, the contents of which are incorporated herein, and repeated details will not be described again.

[0256] It is understood that the user terminal may include smartphones, tablet computers, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), in-vehicle devices, smart wearable devices, etc. Among these, the smart wearable devices may include smart glasses, smartwatches, smart bracelets, etc.

[0257] In practical applications, the welding path planning method can be partially executed on the electronic device side as described above, or all operations can be completed in the client device. The choice can be made based on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed in the client device, the client device may further include a processor.

[0258] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0259] Figure 9 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 9 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 9 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0260] In one embodiment, the welding path planning method function can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following control:

[0261] Step S101: The object to be processed is scanned in three dimensions using a machine vision algorithm to obtain the three-dimensional point cloud data of the object. The three-dimensional point cloud data is then analyzed using a deep learning-based segmentation algorithm to identify the geometric features of the weld area on the surface of the object.

[0262] Step S102: Construct an undirected graph from the extracted geometric features, discretize the weld centerline into graph nodes, use the weld width as the graph edge weight, find the optimal connected path of the weld centerline on the graph using the minimum spanning tree algorithm, and construct a fine three-dimensional model of the processing object using the optimal connected path, wherein the fine three-dimensional model is marked with the three-dimensional coordinates of the weld finding point.

[0263] Step S103: Based on the three-dimensional coordinates of the weld seam locating point, the current welding torch posture, and the target weld seam length, determine the optimal locating height, locating depth, tilt angle, and offset of the welding torch. Based on the locating height, locating depth, tilt angle, and offset, generate the corresponding welding path plan through a welding process simulation algorithm. Output the welding path plan to the industrial robot control system to drive the industrial robot to perform welding operations along the welding path.

[0264] As described above, the electronic device provided in this application embodiment performs three-dimensional scanning of the object to be processed using a machine vision algorithm to obtain three-dimensional point cloud data. A deep learning-based segmentation algorithm is then used to analyze the three-dimensional point cloud data and identify geometric features. These geometric features are then used to construct an undirected graph. A minimum spanning tree algorithm is used to find the optimal connected path for the weld centerline on this graph, and a detailed three-dimensional model of the object to be processed is constructed using this optimal connected path. Based on the three-dimensional coordinates of the weld search point, the current welding torch posture, and the target weld length, the optimal search height, search depth, tilt angle, and offset of the welding torch are determined. Based on the search height, search depth, tilt angle, and offset, a corresponding welding path plan is generated using a welding process simulation algorithm, thereby effectively improving the accuracy and efficiency of welding path planning.

[0265] In another embodiment, the welding path planning device can be configured separately from the central processing unit 9100. For example, the welding path planning device can be configured as a chip connected to the central processing unit 9100, and the welding path planning method function can be implemented through the control of the central processing unit.

[0266] like Figure 9 As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 9 All components shown; in addition, the electronic device 9600 may also include Figure 9 For components not shown, please refer to existing technologies.

[0267] like Figure 9 As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0268] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0269] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0270] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0271] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device's communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0272] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processing unit 9100 to provide input signals and receive output signals, which is the same as in a conventional mobile communication terminal.

[0273] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module 9110 (transmitter / receiver) is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored audio via the speaker 9131.

[0274] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the welding path planning method with a server or client as the execution subject in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the welding path planning method with a server or client as the execution subject in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0275] Step S101: The object to be processed is scanned in three dimensions using a machine vision algorithm to obtain the three-dimensional point cloud data of the object. The three-dimensional point cloud data is then analyzed using a deep learning-based segmentation algorithm to identify the geometric features of the weld area on the surface of the object.

[0276] Step S102: Construct an undirected graph from the extracted geometric features, discretize the weld centerline into graph nodes, use the weld width as the graph edge weight, find the optimal connected path of the weld centerline on the graph using the minimum spanning tree algorithm, and construct a fine three-dimensional model of the processing object using the optimal connected path, wherein the fine three-dimensional model is marked with the three-dimensional coordinates of the weld finding point.

[0277] Step S103: Based on the three-dimensional coordinates of the weld seam locating point, the current welding torch posture, and the target weld seam length, determine the optimal locating height, locating depth, tilt angle, and offset of the welding torch. Based on the locating height, locating depth, tilt angle, and offset, generate the corresponding welding path plan through a welding process simulation algorithm. Output the welding path plan to the industrial robot control system to drive the industrial robot to perform welding operations along the welding path.

[0278] As described above, the computer-readable storage medium provided in this application embodiment performs three-dimensional scanning of the object to be processed using a machine vision algorithm to obtain three-dimensional point cloud data. A deep learning-based segmentation algorithm is then used to analyze the three-dimensional point cloud data and identify geometric features. These geometric features are then used to construct an undirected graph. A minimum spanning tree algorithm is used to find the optimal connected path for the weld centerline on this graph, and a detailed three-dimensional model of the object to be processed is constructed using this optimal connected path. Based on the three-dimensional coordinates of the weld search point, the current welding torch posture, and the target weld length, the optimal search height, search depth, tilt angle, and offset of the welding torch are determined. Based on the search height, search depth, tilt angle, and offset, a corresponding welding path plan is generated using a welding process simulation algorithm, thereby effectively improving the accuracy and efficiency of welding path planning.

[0279] Embodiments of this application also provide a computer program product capable of implementing all steps of the welding path planning method in the above embodiments, where the execution subject is a server or a client. When executed by a processor, this computer program / instruction implements the steps of the welding path planning method. For example, the computer program / instruction implements the following steps:

[0280] Step S101: The object to be processed is scanned in three dimensions using a machine vision algorithm to obtain the three-dimensional point cloud data of the object. The three-dimensional point cloud data is then analyzed using a deep learning-based segmentation algorithm to identify the geometric features of the weld area on the surface of the object.

[0281] Step S102: Construct an undirected graph from the extracted geometric features, discretize the weld centerline into graph nodes, use the weld width as the graph edge weight, find the optimal connected path of the weld centerline on the graph using the minimum spanning tree algorithm, and construct a fine three-dimensional model of the processing object using the optimal connected path, wherein the fine three-dimensional model is marked with the three-dimensional coordinates of the weld finding point.

[0282] Step S103: Based on the three-dimensional coordinates of the weld seam locating point, the current welding torch posture, and the target weld seam length, determine the optimal locating height, locating depth, tilt angle, and offset of the welding torch. Based on the locating height, locating depth, tilt angle, and offset, generate the corresponding welding path plan through a welding process simulation algorithm. Output the welding path plan to the industrial robot control system to drive the industrial robot to perform welding operations along the welding path.

[0283] As described above, the computer program product provided in this application uses a machine vision algorithm to perform a three-dimensional scan of the object to be processed, obtaining three-dimensional point cloud data. A deep learning-based segmentation algorithm is then used to analyze the three-dimensional point cloud data and identify geometric features. These geometric features are then used to construct an undirected graph. A minimum spanning tree algorithm is used to find the optimal connected path for the weld centerline on this graph, and a detailed three-dimensional model of the object to be processed is constructed using this optimal connected path. Based on the three-dimensional coordinates of the weld search point, the current welding torch posture, and the target weld length, the optimal search height, search depth, tilt angle, and offset of the welding torch are determined. Based on the search height, search depth, tilt angle, and offset, a corresponding welding path plan is generated using a welding process simulation algorithm, thereby effectively improving the accuracy and efficiency of welding path planning.

[0284] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0285] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0286] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0287] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0288] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method of weld path planning, characterized by, The method comprises: a three-dimensional scanning of the object to be processed is performed by a machine vision algorithm to obtain three-dimensional point cloud data of the object to be processed, and a segmentation algorithm based on deep learning is set to analyze the three-dimensional point cloud data to identify the geometric features of the weld area on the surface of the object to be processed; the extracted geometric features are constructed into an undirected graph, the weld centerline is discretized into graph nodes, and the weld width is taken as the weight of the graph edge, and the minimum spanning tree algorithm is used to find the optimal connected path of the weld centerline on the graph, and a fine three-dimensional model of the object to be processed is constructed based on the optimal connected path; based on the three-dimensional coordinates of the weld positioning point, the current welding gun posture, and the target weld length, the optimal positioning height, positioning depth, tilt angle, and offset of the welding gun are determined, a corresponding welding path planning is generated based on the positioning height, positioning depth, tilt angle, and offset by a welding process simulation algorithm, and the welding path planning is output to an industrial robot control system to drive the industrial robot to perform a welding operation along the welding path.

2. The weld path planning method of claim 1, wherein, The three-dimensional scanning of the object to be processed by the machine vision algorithm to obtain the three-dimensional point cloud data of the object to be processed comprises: a three-dimensional scanning of the object to be processed is performed by a machine vision algorithm to obtain three-dimensional point cloud data of the object to be processed, and a segmentation algorithm based on deep learning is set to analyze the three-dimensional point cloud data to identify the geometric features of the weld area on the surface of the object to be processed; the three-dimensional point cloud data obtained is preprocessed by data cleaning and data noise reduction to obtain the three-dimensional point cloud data after the preprocessing.

3. The weld path planning method of claim 1, wherein, The three-dimensional scanning of the object to be processed by the machine vision algorithm to obtain the three-dimensional point cloud data of the object to be processed comprises: the three-dimensional point cloud data is input into a semantic segmentation algorithm based on deep learning for fine segmentation; the geometric features of the weld area are extracted from the output results of the algorithm, wherein the geometric features include at least one of the centerline, width, and shape of the weld.

4. The weld path planning method of claim 1, wherein, The three-dimensional scanning of the object to be processed by the machine vision algorithm to obtain the three-dimensional point cloud data of the object to be processed comprises: the centerline and width in the geometric features are constructed into an undirected graph, the centerline is discretized to represent the nodes of the graph, and the width is taken as the weight of the graph edge; the minimum spanning tree algorithm is applied to the constructed undirected graph model to determine the shortest path connecting all nodes as the optimal connected path of the weld centerline.

5. The weld path planning method of claim 1, wherein, The fine three-dimensional model of the object to be processed is constructed based on the optimal connected path, wherein the fine three-dimensional model is marked with the three-dimensional coordinates of the weld positioning point. The weld centerline optimal connected path is fused into the original three-dimensional geometric model of the object to be processed to generate a fine three-dimensional model, wherein the fine three-dimensional model contains the geometric detail information of the weld path; the three-dimensional coordinate information of the weld positioning point is marked in the fine three-dimensional model.

6. The weld path planning method of claim 1, wherein, The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

7. The method of claim 1, wherein, The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

8. A weld path planning apparatus characterized by comprising: The device comprises: A three-dimensional construction module is configured to perform three-dimensional scanning on a machining object by a machine vision algorithm to obtain three-dimensional point cloud data of the machining object, and perform data analysis on the three-dimensional point cloud data by setting a segmentation algorithm based on deep learning to identify geometric features of a welding seam area on a surface of the machining object. A coordinate determination module is configured to construct the extracted geometric features into an undirected graph, discretize a welding center line into graph nodes, and take a welding seam width as a graph edge weight, and find an optimal connected path of the welding center line on the graph by a minimum spanning tree algorithm, and construct a fine three-dimensional model of the machining object based on the optimal connected path, wherein the fine three-dimensional model is marked with three-dimensional coordinates of a welding seam positioning point. A path planning module is configured to determine optimal positioning height, positioning depth, inclination angle and offset of a welding gun based on three-dimensional coordinates of the welding seam positioning point, a current welding gun posture and a target welding seam length, generate corresponding welding path planning by a welding process simulation algorithm based on the positioning height, the positioning depth, the inclination angle and the offset, and output the welding path planning to an industrial robot control system to drive the industrial robot to perform welding operation along the welding path.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the steps of the welding path planning method of any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program implements the steps of the welding path planning method of any one of claims 1 to 7 when executed by the processor.

Images