Weld seam positioning tracking method and system

Abstract

The present application relates to the technical field of weld positioning, and particularly relates to a weld positioning and tracking method and system, comprising the following steps: S01, a small line laser sensor is used to project a laser line to a welding area on a workpiece surface, and multiple intersection point data are collected at a preset frequency, wherein the intersection point data comprises intersection point coordinates, reflection time and reflection intensity; S02, intersection point coordinates belonging to the weld area are selected based on the reflection time and the reflection intensity, and are marked as welding point coordinates; and S03, the real-time collected welding point coordinates are compared with a preset path, and it is determined whether a welding path adjustment instruction is generated. The weld positioning and tracking method provided by the present application can realize high-precision positioning of a weld even in a harsh welding environment by using a small line laser sensor in combination with an intelligent processing algorithm, and welding quality and consistency are improved.

Description

Technical Field

[0001] This invention relates to the field of weld seam positioning technology, specifically to a weld seam positioning and tracking method and system. Background Technology

[0002] In the field of welding automation, precise positioning and tracking of weld seams are crucial for achieving high-quality welding. Currently, most mainstream weld seam positioning technologies on the market employ vision sensors or structured light sensors.

[0003] Chinese patent application CN109304552A discloses a welding system and a weld seam tracking method. The system includes a vision sensor, an image processing module, a control module, a drive module, and a welding torch. The vision sensor is positioned in front of the workpiece and is used to illuminate the workpiece with positioning light to acquire current weld seam information. The image processing module is electrically connected to the output of the vision sensor and the input of the control module, respectively, and is used to convert the current weld seam information into position information and output the position information to the control module. The control module is electrically connected to the input of the drive module and the input of the welding torch, respectively, and is used to calculate the deviation based on the position information and the position information acquired at the previous sampling frequency, and use the deviation to control the drive module to correct the position of the welding torch.

[0004] As mentioned in the above application, existing weld seam positioning mostly uses visual sensors as a tracking and positioning method. Although they can provide rich image information, they are easily interfered with in harsh welding environments such as strong light, smoke or high temperature, which affects the positioning accuracy. While structured light sensors can improve positioning stability to a certain extent, their equipment is large in size, complicated to install and debug, and sensitive to the surface material and color of the weld seam, which limits their applicability in certain specific application scenarios. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a weld seam positioning and tracking method and system.

[0006] This invention employs the following technical solution: a weld seam positioning and tracking method, comprising:

[0007] Step S01: Using a small-line laser sensor, a laser line is projected onto the welding area on the workpiece surface, and multiple intersection point data are collected at a preset frequency. The intersection point data includes intersection point coordinates, reflection time, and reflection intensity.

[0008] Step S02: Based on the reflection time and reflection intensity, the coordinates of the intersection points belonging to the weld area are selected and marked as the welding point coordinates;

[0009] Step S03: Compare the real-time collected welding point coordinates with the preset path to determine whether to generate a welding path adjustment command.

[0010] Step S04: When a welding path adjustment command is generated, the obtained offset coefficient, offset amount, offset coordinates and preset path are input into the pre-built second machine learning model, the welding path adjustment parameters are output, and the welding path adjustment parameters are sent to the welding equipment to control the movement of the welding equipment.

[0011] As a further description of the above technical solution: step S02 also includes establishing a correspondence between the intersection coordinates and the corresponding reflection time and reflection intensity, creating an intersection data set, filtering out the intersection coordinates belonging to the weld area based on the intersection data set, and marking them as welding point coordinates.

[0012] As a further description of the above technical solution: the method for selecting the intersection coordinates belonging to the weld area based on reflection time and reflection intensity, and marking them as weld point coordinates, includes:

[0013] Multiple sets of reflection intensities and reflection times in the substrate area were obtained experimentally.

[0014] Based on multiple sets of reflection intensity and reflection time, the standard intensity range value of the substrate reflection intensity and the standard time range value of the reflection time are obtained.

[0015] Obtain intersection data within the intersection data set where neither the reflection intensity nor the reflection time is within the standard intensity range or the standard time range. Obtain the intersection coordinates within the intersection data set and mark these intersection coordinates as the welding point coordinates.

[0016] Obtain intersection data points within the intersection data set where both reflection intensity and reflection time are within the standard intensity and time ranges, and remove such intersection data points from the intersection data set.

[0017] Obtain intersection data from the intersection data set where either the reflection intensity or the reflection time falls within the standard intensity range or the standard time range. Input the intersection data and the preset path coordinate range into the pre-built first machine learning model and output whether it is a weld area.

[0018] When the output is a weld area, the coordinates in the intersection data are marked as the weld point coordinates; otherwise, when the output is not a weld area, the intersection data is removed from the intersection data set.

[0019] As a further description of the above technical solution: the method for obtaining the standard intensity range value of the substrate reflection intensity based on multiple sets of reflection intensities includes:

[0020] Calculate the mean μ and standard deviation σ of multiple sets of reflection intensities;

[0021] The standard strength range is set based on the mean and standard deviation, and the setting method is: standard range = [μ−2σ,μ+2σ].

[0022] As a further description of the above technical solution: the method for determining whether to generate a welding path adjustment command includes:

[0023] Obtain the width and centerline of the preset path, wherein the width of the preset path is the actual width of the weld.

[0024] Obtain the vertical distance from the welding point coordinates to the path centerline, and record it as the coordinate distance;

[0025] Based on the coordinate distance and path width, determine whether the coordinates of the welding point are within the width of the path. If the coordinate distance is greater than half of the path width, it is considered that the coordinates of the welding point deviate from the preset path. The coordinates of the welding point are marked as offset coordinates, an offset coordinate set is established, and the offset of each offset coordinate in the offset coordinate set is calculated to establish an offset set.

[0026] Extract feature data from the offset set and generate offset coefficients based on the feature data;

[0027] The offset coefficient is compared and analyzed with the preset offset coefficient threshold to determine whether to generate a welding path adjustment command.

[0028] As a further description of the above technical solution: the offset is equal to the coordinate distance minus half of the path width.

[0029] As a further description of the above technical solution: the method for extracting feature data from the offset set and generating offset coefficients based on the feature data includes:

[0030] Collect the number of offset coordinates in the offset set, the maximum value, the average value, and the standard deviation of the offset set;

[0031] The offset coefficient is generated by weighting the number of offset coordinates, the maximum value, the average value, and the standard deviation of the offset set.

[0032] As a further description of the above technical solution: the method for comparing and analyzing the offset coefficient with a preset offset coefficient threshold to determine whether to generate a welding path adjustment command includes:

[0033] When the offset coefficient is greater than or equal to the preset offset coefficient threshold, a welding path adjustment command is generated.

[0034] If the offset coefficient is less than the preset offset coefficient threshold, no welding path adjustment command will be generated.

[0035] The methods for obtaining the coordinate distance include:

[0036] The centerline is formed by a series of discrete points P. j (A) j B jThe path is composed of line segments described by the discrete points, with each pair of consecutive discrete points defining a path segment.

[0037] Obtain welding coordinates ( ， ), calculate the welding coordinates to the current path point P j (A) j B j A vector of ) is denoted as: ;

[0038] Calculate path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 A vector of ) is denoted as: ;

[0039] Calculate vectors exist The projection size is calculated using the following formula:

[0040] ;

[0041] Calculate the coordinates of the projection point based on the projection size. ;

[0042] Calculate whether the projection point lies on path segment P. j (A) j B j ) to P j+1 (A) j+1 B j+1 )superior;

[0043] When the projection point is on path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 When the coordinates are on the welding coordinates ( ) ， ) to path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 The distance is the perpendicular distance from the intersection point to the projection point;

[0044] When the projection point is not on path segment P j (A) jB j ) to P j+1 (A) j+1 B j+1 When on, welding coordinates ( ， The perpendicular distance from the path centerline is the welding coordinate ( ， ) to endpoint P j (A) j B j ) or P j+1 (A) j+1 B j+1 The Euclidean distance between the intersection points is calculated, and the minimum distance is selected as the perpendicular distance from the intersection point to the path.

[0045] A weld seam positioning and tracking system, implementing the aforementioned weld seam positioning and tracking method, the system comprising:

[0046] The data acquisition module projects a laser line onto the welding area on the workpiece surface using a small-line laser sensor, and acquires multiple intersection point data at a preset frequency. The intersection point data includes the intersection point coordinates, reflection time, and reflection intensity.

[0047] The coordinate acquisition module filters out the coordinates of intersection points belonging to the weld area based on reflection time and reflection intensity, and marks them as weld point coordinates.

[0048] The data analysis module compares the real-time collected welding point coordinates with the preset path to determine whether to generate a welding path adjustment command.

[0049] When a welding path adjustment command is generated, the path adjustment module inputs the obtained offset coefficients, offset amounts, offset coordinates, and preset paths into the pre-built second machine learning model, outputs welding path adjustment parameters, and sends the welding path adjustment parameters to the welding equipment to control the movement of the welding equipment.

[0050] Beneficial effects:

[0051] The weld seam positioning and tracking method provided by this invention, through the combination of a small-line laser sensor and intelligent processing algorithm, can achieve high-precision positioning of weld seams even in harsh welding environments, thereby improving welding quality and consistency.

[0052] By precisely projecting through a small-line laser sensor, the system can acquire real-time data on the intersection points of the welding area, ensuring accurate weld positioning. Regardless of dynamic changes during the welding process or minor deviations in equipment or workpiece position, the system can accurately track the weld path through high-frequency data acquisition and real-time analysis. Combined with intelligent algorithms, it can quickly assess the error between the intersection point and the preset path, thereby adjusting the welding position in real time. This precision directly improves welding quality and reduces the occurrence of defective welds.

[0053] Furthermore, by comparing the standard range of reflection intensity and reflection time of the substrate region obtained from the comparative experiment, the weld region and the substrate region can be effectively distinguished. The weld region usually has different optical characteristics (such as surface roughness, temperature influence, etc.), which are significantly different from the reflection signal characteristics of the substrate region. Through this screening method, the coordinates of the weld point can be accurately identified from a large amount of data, thereby improving the accuracy of weld positioning. By pre-screening the intersection data according to the intensity and time of the reflection signal, the amount of data that needs further analysis can be greatly reduced. Data that clearly belongs to the substrate region is directly removed, reducing the amount of computation and complexity of subsequent processing and improving processing speed. Attached Figure Description

[0054] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0055] Figure 1 This is a flowchart illustrating the weld seam positioning and tracking method provided in an embodiment of the present invention.

[0056] Figure 2 A flowchart of a method for filtering out the coordinates of intersection points belonging to the weld area based on reflection time and reflection intensity, and marking them as weld point coordinates, provided in an embodiment of the present invention;

[0057] Figure 3 A flowchart of a method for determining whether a welding path adjustment command has been generated, provided in an embodiment of the present invention;

[0058] Figure 4 This is a module connection diagram of the weld seam positioning and tracking system provided in an embodiment of the present invention. Detailed Implementation

[0059] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0060] Example 1

[0061] Please see Figures 1-2This invention provides a technical solution: a weld seam positioning and tracking method, comprising:

[0062] A laser line is projected onto the welding area of ​​the workpiece surface using a small-line laser sensor. During the welding process, the small-line laser sensor continuously collects data on the intersection points of the laser line and the workpiece surface, acquiring multiple intersection point data at a preset frequency. The intersection point data includes the intersection point coordinates, reflection time, and reflection intensity (e.g., per millimeter or per millisecond).

[0063] It should be noted that the reflection time and reflection intensity are acquired by a small-line laser sensor;

[0064] The small-line laser sensor is equipped with a high-precision receiver (such as a photodiode array or CCD array), which can receive the reflected laser light signal. The receiver obtains the precise intersection coordinates by calculating the position of the laser reflection spot.

[0065] Establish a correspondence between the intersection point coordinates and the corresponding reflection time and reflection intensity, create an intersection point data set, and filter out the intersection point coordinates belonging to the weld area based on the intersection point data set, and mark them as weld point coordinates;

[0066] The intersection data set is represented as follows: ;

[0067] In the formula Indicates the coordinates of the intersection point. Indicates and The corresponding reflection intensity and reflection time, Indicates the intensity of reflection. Indicates the reflection time.

[0068] It should be noted that the surface characteristics of the weld area usually differ from those of the surrounding area, and the reflection intensity is an important basis for distinguishing the weld area from the non-weld area. By analyzing the reflection intensity, the intersection of the weld area can be identified; the reflection intensity of the base material area is stable, while the reflection intensity of the weld area (smoothness, roughness, oxide layer) will change;

[0069] During welding, the temperature in the weld area fluctuates significantly, which may cause a "time drift" phenomenon in the reflected signal. By analyzing the time difference of the reflected signal, the weld area and the substrate area can be distinguished.

[0070] Methods for selecting the coordinates of intersection points belonging to the weld area based on reflection time and reflection intensity, and marking them as weld point coordinates, include:

[0071] Multiple sets of reflection intensities and reflection times in the substrate area were obtained experimentally.

[0072] Based on multiple sets of reflection intensity and reflection time, the standard intensity range value of the substrate reflection intensity and the standard time range value of the reflection time are obtained.

[0073] It should be noted that methods for obtaining the standard intensity range value of substrate reflection intensity based on multiple sets of reflection intensities include:

[0074] Calculate the mean μ and standard deviation σ of multiple sets of reflection intensities;

[0075] It should be noted that the formulas for calculating the mean and standard deviation are:

[0076] ;

[0077] ;

[0078] In the formula, For the first Group reflection intensity data, This represents the total number of data sets.

[0079] The standard strength range is set based on the mean and standard deviation. The method for setting the standard range is: standard range = [μ−2σ,μ+2σ].

[0080] Similarly, based on the above method, the standard time range value of the substrate reflection time is obtained based on multiple sets of reflection times, which will not be elaborated here.

[0081] Obtain intersection data within the intersection data set where neither the reflection intensity nor the reflection time is within the standard intensity range or the standard time range. Obtain the intersection coordinates within the intersection data set and mark these intersection coordinates as the welding point coordinates.

[0082] Obtain intersection data points within the intersection data set where both reflection intensity and reflection time are within the standard intensity and time ranges, and remove such intersection data points from the intersection data set.

[0083] Obtain intersection data from the intersection data set where either the reflection intensity or reflection time falls within the standard intensity or time range. Input the intersection data and the preset path coordinate range into the pre-built first machine learning model. The output is whether it is a weld area. If the output is a weld area, mark the coordinates in the intersection data as the welding point coordinates. Otherwise, if the output is not a weld area, remove the intersection data from the intersection data set.

[0084] The method for constructing the first machine learning model includes:

[0085] The first machine learning model structure is initialized. The first machine learning model structure adopts a multi-layer feedforward network structure of MLP type, with four input layers, two hidden layers, and one output layer. The four input layers correspond to the intersection coordinates, reflection intensity, reflection time, and preset path coordinate range, respectively. The hidden layers include a first hidden layer and a second hidden layer. The first hidden layer has 128 nodes and uses the ReLU function as the activation function. The second hidden layer has 64 nodes and uses the ReLU function as the activation function. The output layer is the first output layer, which outputs whether the area is a weld seam or not.

[0086] After initializing the structure of the first machine learning model, the first machine learning model is trained using the initial data and whether the corresponding area is a weld seam. The initial data and whether the corresponding area is a weld seam are obtained from the database. The database records the data of each precise weld seam positioning and tracking based on the small-line laser sensor for model optimization. The optimizer is Adam, the loss function is the MSE loss function, the batch size is set to 32, and the number of iterations is 200. Training ends when the loss function converges, indicating that training is complete.

[0087] It should be noted that the Adam optimizer is one of the five major optimizers commonly used in machine learning, and its full name is Adaptive Moment Estimation.

[0088] The training method for the first machine learning model includes:

[0089] The initial data is converted into a corresponding set of feature vectors.

[0090] The initial data is used as input to the machine learning model. The machine learning model outputs whether each set of initial data corresponds to a weld area and uses whether the actual corresponding area of ​​each set of initial data is a weld area as the prediction target. The numerical labels for whether it is a weld area are set to 1 and 0. The training objective is to minimize the loss function value of the machine learning model. Training stops when the loss function value of the machine learning model is less than or equal to the preset target loss value.

[0091] The loss function value of the first machine learning model is the mean squared error;

[0092] Mean squared error is one of the commonly used loss functions. This is achieved by using the loss function formula... = By training the model with minimization as the objective, the machine learning model can better fit the data, thereby improving the model's performance and accuracy.

[0093] In the loss function Here, x represents the loss function value of the first machine learning model, and e represents the feature vector group number; e represents the number of feature vector groups. Let x be the numerical label output by the first machine learning model corresponding to the x-th feature vector. denoted as the actual numerical label corresponding to the x-th feature vector group.

[0094] It is particularly important to note that, since this is strategic data training, whether the initial data corresponds to the optimal decision control group of the initial sample of the weld area is obtained by making the optimal decision manually after manual experimentation and recording it. At present, the initial stage of decision-making training data all requires manual decision-making. Machine learning is based on human decision-making, and logical judgment and classification are obtained. The most typical example is the manually labeled data involved in the big data feeding of GPT.

[0095] In this embodiment, by comparing the reflection intensity and reflection time standard range of the substrate region obtained from the experiment, the weld region and the substrate region can be effectively distinguished. The weld region usually has different optical characteristics (such as surface roughness, temperature influence, etc.), which are significantly different from the reflection signal characteristics of the substrate region. Through this screening method, the coordinates of the weld point can be accurately identified from a large amount of data, thereby improving the accuracy of weld positioning. By pre-screening the intersection data according to the intensity and time of the reflection signal, the amount of data that needs further analysis can be greatly reduced. Data that clearly belongs to the substrate region will be directly removed, reducing the amount of computation and complexity of subsequent processing and improving processing speed.

[0096] In summary, by using the standard reflection intensity and reflection time range obtained from experiments, combined with the adaptability of the machine learning model, the system can dynamically adjust the recognition standard according to different welding conditions and material characteristics. As the welding environment changes (such as different materials, different welding parameters, etc.), the system can optimize the recognition process by adjusting the standard range and the output results of the machine learning model.

[0097] The coordinates of the welding points collected in real time are compared with the preset path to determine whether to generate a welding path adjustment command.

[0098] It should be noted that the preset path is a pre-established path for the weld seam based on welding process requirements and path planning.

[0099] Example 2

[0100] Please see Figures 1-3 This embodiment further discloses, based on the above embodiments, that:

[0101] Methods for determining whether a welding path adjustment command has been generated include:

[0102] Obtain the width and centerline of the preset path, wherein the width of the preset path is the actual width of the weld.

[0103] It should be noted that the width of the preset path is usually determined based on welding process requirements, welding materials, welding equipment, etc.

[0104] Obtain the vertical distance from the welding point coordinates to the path centerline, and record it as the coordinate distance;

[0105] The methods for obtaining the coordinate distance include:

[0106] The centerline is formed by a series of discrete points P. j (A) j B j The path is composed of line segments described by the discrete points, with each pair of consecutive discrete points defining a path segment.

[0107] Obtain welding coordinates ( ， ), calculate the welding coordinates to the current path point P j (A) j B j A vector of ) is denoted as: ;

[0108] Calculate path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 A vector of ) is denoted as: ;

[0109] Calculate vectors exist The projection size is calculated using the following formula:

[0110] ;

[0111] Specifically, by calculating the size of the projection, we can know the degree of offset of the welding point in the path direction, thereby determining whether the point deviates from the path, as well as the direction and degree of deviation.

[0112] Calculate the coordinates of the projection point based on the projection size. ;

[0113] in, ;

[0114] ;

[0115] Calculate whether the projection point lies on path segment P. j (A) j B j ) to P j+1 (A)j+1 B j+1 )superior;

[0116] When the projection point is on path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 When the coordinates are on the welding coordinates ( ) ， ) to path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 The distance is the perpendicular distance from the intersection point to the projection point, and the calculation formula is:

[0117]

[0118] When the projection point is not on path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 When on, welding coordinates ( ， The perpendicular distance from the path centerline is the welding coordinate ( ， ) to endpoint P j (A) j B j ) or P j+1 (A) j+1 B j+1 The Euclidean distance between the intersection points is calculated, and the minimum distance is selected as the perpendicular distance from the intersection point to the path.

[0119] The calculation formula is as follows:

[0120]

[0121] For each welding coordinate ( , ), calculate the distance to each path segment, and select the smallest distance as the perpendicular distance from the intersection point to the path.

[0122] Calculate whether the projection point lies on path segment P. j (A) j B j ) to P j+1 (A) j+1B j+1 The method is as follows:

[0123] When 0≤ When ≤1, the projection point lies on path segment P. j (A) j B j ) to P j+1 (A) j+1 B j+1 If the projection point is on the path segment P, then the projection point is not on the path segment P. j (A) j B j ) to P j+1 (A) j+1 B j+1 )superior.

[0124] Based on the coordinate distance and path width, determine whether the coordinates of the welding point are within the width of the path. If the coordinate distance is greater than half of the path width, it is considered that the coordinates of the welding point deviate from the preset path. The coordinates of the welding point are marked as offset coordinates, an offset coordinate set is established, and the offset of each offset coordinate in the offset coordinate set is calculated to establish an offset set.

[0125] The offset is equal to the coordinate distance minus half the path width;

[0126] Extract feature data from the offset set and generate offset coefficients based on the feature data;

[0127] The method for extracting feature data from the offset set and generating offset coefficients based on the feature data includes:

[0128] Collect the number of offset coordinates in the offset set, the maximum value, the average value, and the standard deviation of the offset set;

[0129] The offset coefficient is generated by weighting the number of offset coordinates, the maximum value, the average value, and the standard deviation of the offset set.

[0130] It's important to note that the number of offset coordinates reflects the distribution of offset points. A large number indicates that the offset is a widespread phenomenon, rather than an anomaly at individual points. If there are few offset points, it may be a random error, requiring no adjustment to the welding path. The maximum offset value measures the severity of the weld offset. If the offset in a certain area is particularly large, it indicates a significant welding problem in that area, potentially requiring rapid adjustment of the welding path. The average offset value measures the severity of the weld offset, judging whether the welding path is shifted overall, rather than focusing on individual extreme points. A large average value indicates that the entire weld is shifted in a certain direction, requiring overall path adjustment. The standard deviation of the offset measures the volatility of the offset. A large standard deviation indicates large fluctuations in weld offset and poor welding path stability, potentially requiring adjustments to welding parameters such as feed rate or current. Therefore, by combining multiple features, the overall trend and local offset of the welding trajectory can be more accurately assessed, leading to more reasonable adjustments to the welding path. The weighted calculation combines the number of offset points, maximum offset, average offset, and offset volatility, making welding path adjustments more precise, efficient, and stable, thereby improving welding quality and reducing the need for human intervention.

[0131] Optionally, the formula for calculating the offset coefficient is:

[0132] ;

[0133] In the formula, This is the offset coefficient. The number of offset coordinates, The maximum value of the offset set, This is the average value. Standard deviation, , , and These are the weighting coefficients, and , , and All are greater than 0.

[0134] It should be noted that before generating the offset coefficients, the number of offset coordinates, the maximum value, the average value, and the standard deviation of the offset set are weighted and calculated. This process requires standardization to eliminate dimensional differences. The specific method for standardization is to calculate the mean and standard deviation of each parameter and convert the original data into Z-score standardized values.

[0135] , , and The value of the weighting coefficient is a specific numerical value obtained by quantifying each data point to facilitate subsequent comparison. The value of the weighting coefficient depends on the number of comprehensive parameters and the weighting coefficient initially set by those skilled in the art for each set of comprehensive parameters.

[0136] The offset coefficient is compared and analyzed with the preset offset coefficient threshold to determine whether to generate a welding path adjustment command.

[0137] The method for comparing and analyzing the offset coefficient with a preset offset coefficient threshold to determine whether to generate a welding path adjustment command includes:

[0138] When the offset coefficient is greater than or equal to the preset offset coefficient threshold, a welding path adjustment command is generated.

[0139] If the offset coefficient is less than the preset offset coefficient threshold, no welding path adjustment command will be generated.

[0140] When a welding path adjustment command is generated, the obtained offset coefficients, offset amounts, offset coordinates, and preset paths are input into the pre-built second machine learning model, which outputs welding path adjustment parameters and sends these parameters to the welding equipment to control its movement.

[0141] The construction method of the second machine learning model includes:

[0142] Each set of offset coefficients, offset amounts, offset coordinates, and preset paths are converted into the first feature vector.

[0143] The set of all first feature vectors is used as the input of the second machine learning model. The second machine learning model takes the welding path adjustment parameters predicted for each set of offset coefficients and offsets, offset coordinates and preset paths as the output, takes the actual welding path adjustment parameters corresponding to each set of offset coefficients and offsets, offset coordinates and preset paths as the prediction target, and takes minimizing the sum of the first prediction accuracies of all predicted welding path adjustment parameters as the training objective.

[0144] The formula for calculating the first prediction accuracy is as follows: ,in, Each set of offset coefficients, offset amounts, offset coordinates, and the first feature vector corresponding to the preset path is numbered. For the highest prediction accuracy, For the first The group offset coefficient and offset amount, offset coordinates, and the predicted welding path adjustment parameters corresponding to the preset path. For the first The parameters for adjusting the group offset coefficients, offset amounts, offset coordinates, and actual welding paths corresponding to the preset paths are set; the second machine learning model is trained until the sum of the first prediction accuracies converges and training stops; the second machine learning model is a regression model or a random forest.

[0145] In this embodiment, the weld seam positioning and tracking method, through a small-line laser sensor combined with an intelligent processing algorithm, can achieve high-precision positioning of the weld seam even in harsh welding environments, thereby improving welding quality and consistency.

[0146] By precisely projecting data using a small-line laser sensor, the system can acquire intersection data of the welding area in real time, ensuring accurate weld positioning. Whether due to dynamic changes during the welding process or minor deviations in equipment or workpiece position, the system can accurately track the weld path through high-frequency data acquisition and real-time analysis. Combined with intelligent algorithms, it can quickly assess the error between the intersection point and the preset path, thereby adjusting the welding position in real time. This precision directly improves welding quality and reduces the occurrence of defective welds.

[0147] Example 3

[0148] Please see Figure 4 This implementation discloses a weld seam positioning and tracking system and a weld seam positioning and tracking method. The system includes:

[0149] The data acquisition module projects a laser line onto the welding area on the workpiece surface using a small-line laser sensor, and acquires multiple intersection point data at a preset frequency. The intersection point data includes the intersection point coordinates, reflection time, and reflection intensity.

[0150] The coordinate acquisition module filters out the coordinates of intersection points belonging to the weld area based on reflection time and reflection intensity, and marks them as weld point coordinates.

[0151] The data analysis module compares the real-time collected welding point coordinates with the preset path to determine whether to generate a welding path adjustment command.

[0152] When a welding path adjustment command is generated, the path adjustment module inputs the obtained offset coefficients, offset amounts, offset coordinates, and preset paths into the pre-built second machine learning model, outputs welding path adjustment parameters, and sends the welding path adjustment parameters to the welding equipment to control the movement of the welding equipment.

[0153] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A weld seam positioning tracking method, characterized in that, The method comprises the following steps: Step S01, projecting a laser line onto a welding area on a workpiece surface through a small-line laser sensor, and collecting a plurality of intersection data at a preset frequency, the intersection data comprising intersection coordinates, reflection time and reflection intensity; Step S02, screening intersection coordinates belonging to a weld area based on the reflection time and the reflection intensity, and marking the intersection coordinates as welding point coordinates; Step S03, comparing the welding point coordinates collected in real time with a preset path, and determining whether to generate a welding path adjustment instruction; The method for determining whether to generate a welding path adjustment instruction comprises: obtaining the width and the center line of the preset path, wherein the width of the preset path is the actual width of the weld; obtaining the perpendicular distance from the welding point coordinates to the path center line, denoted as coordinate distance; determining whether the welding point coordinates are within the width range of the path according to the coordinate distance and the path width, and when the coordinate distance is greater than half of the path width, considering that the welding point coordinates deviate from the preset path, marking the welding point coordinates as offset coordinates, establishing an offset coordinate set, and calculating the offset amount of each offset coordinate in the offset coordinate set to establish an offset amount set; extracting feature data of the offset amount set, and generating an offset coefficient based on the feature data; comparing and analyzing the offset coefficient with a preset offset coefficient threshold value, and determining whether to generate a welding path adjustment instruction; Step S04, when the welding path adjustment instruction is generated, inputting the obtained offset coefficient and offset amount, offset coordinates and preset path into a pre-constructed second machine learning model, outputting a welding path adjustment parameter, and sending the welding path adjustment parameter to a welding device to control the welding device to move.

2. The weld seam positioning tracking method according to claim 1, characterized in that, In the step S02, the intersection coordinates and the corresponding reflection time and reflection intensity are associated, an intersection data set is created, the intersection coordinates belonging to the weld area are screened based on the intersection data set, and the intersection coordinates are marked as welding point coordinates.

3. The weld seam positioning tracking method according to claim 2, characterized in that, The method for screening the intersection coordinates belonging to the weld area based on the reflection time and the reflection intensity, and marking the intersection coordinates as welding point coordinates comprises: obtaining a plurality of groups of reflection intensity and reflection time of the base material area by experiment; based on the plurality of groups of reflection intensity and reflection time, obtaining a standard intensity range value of the base material reflection intensity and a standard time range value of the reflection time; obtaining intersection data in which the reflection intensity and the reflection time are not within the standard intensity range value and the standard time range value from the intersection data set, obtaining intersection coordinates in the intersection data, and marking the intersection coordinates as welding point coordinates; obtaining intersection data in which the reflection intensity and the reflection time are within the standard intensity range value and the standard time range value from the intersection data set, and removing the intersection data from the intersection data set; obtaining intersection data in which one of the reflection intensity and the reflection time is within the standard intensity range value or the standard time range value from the intersection data set, inputting the intersection data and a preset path coordinate range into a pre-constructed first machine learning model, and outputting whether it is a weld area; when the output is a weld area, marking the coordinates in the intersection data as welding point coordinates, and otherwise, when the output is not a weld area, removing the intersection data from the intersection data set.

4. The weld seam positioning tracking method according to claim 3, characterized in that, The method for obtaining the standard intensity range value of the substrate reflection intensity based on the multiple sets of reflection intensity comprises: calculating the mean value μ and the standard deviation σ of the multiple sets of reflection intensity; setting the standard intensity range according to the mean value and the standard deviation, and the setting method is: standard range = [μ−2σ, μ+2σ].

5. The weld seam positioning tracking method according to claim 1, characterized in that, The offset is equal to the coordinate distance minus half of the path width.

6. The weld seam positioning tracking method according to claim 1, characterized in that, The method for extracting the feature data of the offset set and generating the offset coefficient based on the feature data comprises: collecting the number of offset coordinates in the offset set, the maximum value, the average value and the standard deviation of the offset set; performing a weighted operation based on the number of offset coordinates, the maximum value, the average value and the standard deviation of the offset set to generate the offset coefficient.

7. The weld seam positioning tracking method according to claim 1, characterized in that, The method for comparing and analyzing the offset coefficient with the preset offset coefficient threshold value to determine whether to generate the welding path adjustment instruction comprises: when the offset coefficient is greater than or equal to the preset offset coefficient threshold value, a welding path adjustment instruction is generated; when the offset coefficient is less than the preset offset coefficient threshold value, no welding path adjustment instruction is generated.

8. The weld seam positioning tracking method according to claim 5, characterized in that, The method for obtaining the coordinate distance comprises: The centerline is composed of a series of discrete points P j (A j , B j ) and each two continuous discrete points define a path line segment; Obtain the welding coordinate (x ， ), calculate the vector of the welding coordinate to the current path point P j (A j , B j ), denoted as: ； The path segment P j (A j , B j ) to P j+1 (A j+1 , B j+1 ) is denoted by: ； Computing the vector On the projection size ​ According to the projection size, the coordinates of the projection point are calculated ; Compute whether the projected point is on the path segment P j (A j , B j ) to P j+1 (A j+1 , B j+1 ) on; When the projection point is on path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 When the coordinates are on the welding coordinates ( ) ， ) to path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 The distance is the perpendicular distance from the intersection point to the projection point; When the projection point is not on path segment P j (A) j B j ) to P j+1 (A) j+1 B j+1 When on, welding coordinates ( ， The perpendicular distance from the path centerline is the welding coordinate ( ， ) to endpoint P j (A) j B j ) or P j+1 (A) j+1 B j+1 The Euclidean distance between the intersection points is used to determine the perpendicular distance from the intersection point to the path.

9. A weld seam positioning and tracking system characterized by, implementing the welding seam positioning and tracking method according to any one of claims 1-8, and the system comprises: a data acquisition module that projects a laser line onto a welding area on the surface of a workpiece through a small-line laser sensor to collect multiple intersection data at a preset frequency, and the intersection data comprises intersection coordinates, reflection time and reflection intensity; a coordinate acquisition module that filters out intersection coordinates belonging to the welding seam area based on the reflection time and the reflection intensity, and marks the intersection coordinates as welding point coordinates; a data analysis module that compares the real-time collected welding point coordinates with a preset path to determine whether to generate a welding path adjustment instruction, and the method for determining whether to generate a welding path adjustment instruction comprises: obtaining the width and the center line of the preset path, and the width of the preset path is the actual width of the welding seam; obtaining the perpendicular distance from the welding point coordinates to the path center line, which is recorded as the coordinate distance; determining whether the welding point coordinates are within the width range of the path according to the coordinate distance and the path width, and when the coordinate distance is greater than half of the path width, it is considered that the welding point coordinates deviate from the preset path, the welding point coordinates are marked as offset coordinates, an offset coordinate set is established, and the offset of each offset coordinate in the offset coordinate set is calculated to establish an offset set; extracting the feature data of the offset set and generating an offset coefficient based on the feature data; comparing and analyzing the offset coefficient with the preset offset coefficient threshold value to determine whether to generate a welding path adjustment instruction; a path adjustment module that inputs the obtained offset coefficient and offset, offset coordinates and preset path into a pre-constructed second machine learning model to output welding path adjustment parameters when a welding path adjustment instruction is generated, sends the welding path adjustment parameters to a welding device, and controls the welding device to move.

Images