A robotic arm posture planning method and system

Abstract

The present invention relates to the field of posture planning technology, and in particular to a method and system for posture planning of a robotic arm. The method comprises the following steps: obtaining a product welding structure diagram and basic setting parameters of a robotic arm welding gun, analyzing the morphological differences between welding point sequences, and obtaining morphological difference data of the welding point sequences; secondly, based on the basic setting parameters of the robotic arm welding gun, simulating and analyzing the morphological deviation of the welding pool, and performing strength loss posture association mining to obtain strength loss posture association data; finally, performing joint angle spatial limit matching based on the strength loss posture association feature data, generating joint angle spatial limit matching data, and performing posture planning and design of the robotic arm welding gun based on this data to obtain final posture planning data. The present invention makes posture planning technology more accurate by optimizing the posture planning technology.

Description

Technical Field

[0001] The present invention relates to the technical field of posture planning, and in particular to a method and system for posture planning of a robotic arm. Background Art

[0002] The posture planning of the robot arm is one of the key technologies to achieve efficient and precise operation. Welding is a process widely used in the manufacturing industry, and it has very high requirements for welding accuracy and weld quality. The posture planning of the robot arm is particularly important in this process. During the welding process, the posture of the robot arm directly affects the formation of the welding pool, the welding quality and the final strength of the product. With the diversification of product structures and the continuous improvement of welding process requirements, traditional methods have gradually shown their limitations in dealing with complex welding tasks. Specifically, during the welding process, the position, angle and posture of the robot arm need to be constantly adjusted to cope with the morphological differences of the welding points, changes in the molten pool and other environmental factors. This requires the robot arm to have more flexible and precise posture planning capabilities. However, a traditional robot arm posture planning method has the problem of inaccurate analysis of the strength loss caused by deviation posture welding, and thus cannot accurately perform robot arm welding posture planning. Summary of the Invention

[0003] Based on this, it is necessary to provide a robotic arm posture planning method and system to solve at least one of the above technical problems.

[0004] To achieve the above object, a robot arm posture planning method is provided, the method comprising the following steps:

[0005] Step S1: Obtain a product welding structure diagram and basic setting parameters of a robotic arm welding gun; perform morphological difference analysis between welding point sequences based on the product welding structure diagram, thereby obtaining morphological difference data between welding point sequences;

[0006] Step S2: performing a welding pool morphology deviation simulation analysis on the morphology difference data between the welding point sequences according to the basic setting parameters of the robotic arm welding gun to obtain welding pool morphology deviation data; performing strength loss posture association mining based on the welding pool morphology deviation data to obtain strength loss posture association data;

[0007] Step S3: Perform joint angle spatial limit matching based on the strength loss posture correlation feature data to generate joint angle spatial limit matching data; perform robot arm welding gun posture planning and design based on the joint angle spatial limit matching data to obtain robot arm welding gun posture planning data.

[0008] Preferably, step S1 includes the following steps:

[0009] Step S11: Obtain product welding structure diagram and basic setting parameters of the robotic arm welding gun;

[0010] Step S12: vectorizing the product welding structure diagram to generate a product welding structure vector diagram;

[0011] Step S13: extracting the welding requirement form from the product welding structure vector diagram to obtain product welding requirement form data;

[0012] Step S14: performing morphological difference analysis between welding point sequences on the product welding requirement morphological data based on the product welding structure vector diagram, thereby obtaining morphological difference data between welding point sequences.

[0013] Preferably, step S2 includes the following steps:

[0014] Step S21: extracting the welding gun setting posture from the basic setting parameters of the robot arm welding gun;

[0015] Step S22: performing a welding pool morphology deviation simulation analysis on the morphology difference data between the welding point sequences according to the basic setting parameters of the robot arm welding gun and the setting posture of the welding gun to obtain welding pool morphology deviation data;

[0016] Step S23: performing weld spot strength loss tolerance fitting on the weld pool morphology deviation data to generate weld spot strength loss tolerance data;

[0017] Step S24: performing strength loss posture association mining on the welding gun setting posture based on the weld point strength loss tolerance data to obtain strength loss posture association data.

[0018] Preferably, step S22 includes the following steps:

[0019] Step S221: extracting the welding setting voltage, current and speed from the basic setting parameters of the robot arm welding gun; calculating the point-to-point vector angle between the welding points according to the set posture of the welding gun, thereby obtaining the vector angle between the welding point and the welding gun;

[0020] Step S222: performing vector angle change analysis on the morphological difference data between the welding points in sequence based on the morphological difference data between the welding points in sequence, and obtaining posture change vector angle data between the welding points and the welding gun;

[0021] Step S223: simulating and analyzing the instantaneous temperature field of the weld pool during the multi-frequency welding process based on the posture change vector angle data according to the welding set voltage, current, and speed to obtain the instantaneous temperature field of the weld pool during the multi-frequency welding process;

[0022] Step S224: performing anisotropic fluid solidification behavior imbalance analysis on the instantaneous temperature field of the molten pool at multiple frequencies of the welding spot, thereby obtaining anisotropic fluid solidification behavior imbalance data;

[0023] Step S225: performing interfacial tension vibration regression analysis on the unbalanced solidification behavior data of the anisotropic fluids to obtain interfacial tension vibration regression data;

[0024] Step S226: performing a welding pool morphology deviation simulation analysis on the morphology difference data between welding spot sequences according to the anisotropic fluid solidification behavior imbalance data and the interfacial tension vibration regression data to obtain welding pool morphology deviation data.

[0025] Preferably, step S224 includes the following steps:

[0026] The heat flux density non-uniformity vector is identified for the instantaneous temperature field of the molten pool of multiple-frequency solder joints to obtain the heat flux density non-uniformity vector;

[0027] Based on the heat flux density inhomogeneity vector, differential coupling of the time series change of the solder joint surface tension is performed to generate differential coupling data of the time series change of tension;

[0028] The local variance of the molten metal viscosity is calculated based on the differential coupling data of the heat flux density inhomogeneity vector and the tension time series variation, and the local variance of the molten metal viscosity is obtained;

[0029] Based on the heat flux density inhomogeneity vector, the differential coupling data of the tension time series change and the local variance of the molten metal viscosity, the heat energy release ratio difference of the solidification latent heat between different solder joint orientations is deduced to generate the heat energy release ratio difference between different solder joint orientations;

[0030] The imbalance analysis of solidification behavior of anisotropic fluid is performed based on the difference in heat energy release ratios between different solder joint orientations, thereby obtaining the imbalance data of solidification behavior of anisotropic fluid.

[0031] Preferably, step S23 includes the following steps:

[0032] Step S231: performing a molten pool morphology deviation network division process on the welding molten pool morphology deviation data to obtain molten pool morphology deviation network division data;

[0033] Step S232: performing a melt pool geometric deviation feature analysis on the melt pool morphology deviation network division data, thereby obtaining melt pool geometric deviation feature data;

[0034] Step S233: simulating the weld point collision force offset distribution based on the molten pool geometric deviation characteristic data to generate collision force offset distribution data;

[0035] Step S234: performing stress distribution strain rate component decomposition on the collision force offset distribution data to obtain strain rate component decomposition data;

[0036] Step S235 : performing weld strength loss tolerance fitting based on the strain rate component decomposition data to generate weld strength loss tolerance data.

[0037] Preferably, step S3 includes the following steps:

[0038] Step S31: performing feature learning on the intensity loss posture associated data to obtain intensity loss posture associated feature data;

[0039] Step S32: performing joint angle space limit matching on the welding gun set posture according to the strength loss posture correlation feature data to generate joint angle space limit matching data;

[0040] Step S33: performing singular configuration optimization adjustment on the joint angle space limit matching data, thereby obtaining joint angle space limit optimization data;

[0041] Step S34: performing robot arm welding gun posture planning and design through joint angle spatial limit optimization data to obtain robot arm welding gun posture planning data.

[0042] Preferably, step S32 includes the following steps:

[0043] Step S321: performing joint verticality matching on the welding gun set posture according to the strength loss posture correlation feature data to obtain joint verticality matching data;

[0044] Step S322: performing adaptive adjustment control of the thrust angle / drag angle of the robot arm welding gun according to the joint verticality matching data and the strength loss posture correlation feature data to obtain thrust angle / drag angle adaptive control data;

[0045] Step S323: performing feed welding speed matching on the push angle / drag angle adaptive control data, thereby obtaining feed welding speed matching data;

[0046] Step S324: performing joint angle space limit matching on the welding gun setting posture based on the joint verticality matching data, the push angle / drag angle adaptive control data and the feed welding speed matching data to generate joint angle space limit matching data.

[0047] Preferably, the present invention further provides a robotic arm posture planning system for executing the robotic arm posture planning method described above, the robotic arm posture planning system comprising:

[0048] The morphological difference analysis module is used to obtain the product welding structure diagram and the basic setting parameters of the robot arm welding gun; based on the product welding structure diagram, the morphological difference analysis between the welding point sequences is performed to obtain the morphological difference data between the welding point sequences;

[0049] The strength loss posture association mining module is used to simulate and analyze the morphological difference data between the welding points according to the basic setting parameters of the robotic arm welding gun to obtain the welding pool morphological deviation data; based on the welding pool morphological deviation data, the strength loss posture association mining is performed to obtain the strength loss posture association data;

[0050] The robot arm welding gun posture planning module is used to perform joint angle space limit matching based on the strength loss posture correlation feature data to generate joint angle space limit matching data; the robot arm welding gun posture planning and design is performed through the joint angle space limit matching data to obtain the robot arm welding gun posture planning data.

[0051] The beneficial effect of the present invention is that, by obtaining the product welding structure diagram and the basic setting parameters of the robotic arm welding gun, the morphological differences between the welding point sequences can be accurately analyzed, and the morphological difference data between the welding point sequences can be obtained. This provides basic data support for the subsequent optimization of the welding process. Through the analysis of the morphological differences, the requirements of each welding point can be better understood, providing an important basis for parameter adjustment and optimization during the welding process, which helps to improve the accuracy and efficiency of the welding process. According to the basic setting parameters of the robotic arm welding gun, a simulation analysis of the morphological deviation of the welding pool is performed on the morphological difference data between the welding point sequences, and the morphological deviation data of the welding pool can be obtained. Through this simulation analysis, not only can the molten pool deviation that occurs during the welding process be foreseen, but the correlation between strength loss and welding posture can also be further mined to obtain strength loss posture correlation data. This correlation data can help optimize the welding process, reduce welding defects, and improve welding quality. After obtaining the strength loss posture correlation data, the joint angle space limit matching is performed, and the joint angle space limit matching data is generated to ensure that the robot arm can operate within a suitable spatial range, avoiding equipment damage or welding defects caused by excessive movement. Based on these matching data, the robot arm welding gun posture planning design is performed to optimize the robot arm movement trajectory during the welding process, making the posture planning more accurate and able to maintain efficient and stable operating performance in complex welding tasks. Therefore, the present invention is an optimization process made to a traditional robot arm posture planning method, which solves the problem that a traditional robot arm posture planning method has an inaccurate analysis of the strength loss caused by deviation posture welding, thereby failing to accurately perform the robot arm welding posture planning, improves the accuracy of the strength loss analysis caused by deviation posture welding, and improves the accuracy of the robot arm welding posture planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A schematic flow chart of the steps of a robot arm posture planning method;

[0053] Figure 2 for Figure 1Detailed implementation steps of step S2 in FIG.

[0054] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG. DETAILED DESCRIPTION

[0055] See also Figures 1 to 3 , a robot arm posture planning method, the method comprising the following steps:

[0056] Step S1: Obtain a product welding structure diagram and basic setting parameters of a robotic arm welding gun; perform morphological difference analysis between welding point sequences based on the product welding structure diagram, thereby obtaining morphological difference data between welding point sequences;

[0057] Step S2: performing a welding pool morphology deviation simulation analysis on the morphology difference data between the welding point sequences according to the basic setting parameters of the robotic arm welding gun to obtain welding pool morphology deviation data; performing strength loss posture association mining based on the welding pool morphology deviation data to obtain strength loss posture association data;

[0058] Step S3: Perform joint angle spatial limit matching based on the strength loss posture correlation feature data to generate joint angle spatial limit matching data; perform robot arm welding gun posture planning and design based on the joint angle spatial limit matching data to obtain robot arm welding gun posture planning data.

[0059] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a robot arm posture planning method of the present invention. In this example, the robot arm posture planning method includes the following steps:

[0060] Step S1: Obtain a product welding structure diagram and basic setting parameters of a robotic arm welding gun; perform morphological difference analysis between welding point sequences based on the product welding structure diagram, thereby obtaining morphological difference data between welding point sequences;

[0061] In an embodiment of the present invention, a three-dimensional welding structure diagram of a target product is first acquired. A high-precision industrial CT scanning device is used to perform non-destructive testing on the entire product. The scanned tomographic images of the product are reconstructed layer by layer using an image segmentation algorithm. A complete product welding structure diagram is obtained through three-dimensional reconstruction. The three-dimensional welding structure diagram is converted into a format using CAD software and output as a structure file in a standard STEP or IGES format. Subsequently, a geometric feature extraction algorithm based on vector analysis is called to extract the spatial coordinates, welding path direction, and weld thickness parameters of all welding points in the structure diagram point by point. A point-to-point sequence relationship mapping is performed on all welding points using a sorted topological relationship matrix. Then, an angle change analysis in Euclidean space is performed on the spatial vector variation between the welding point sequences. The minimum spatial turning radius between different welding points and the length difference between adjacent weld points are vector-scaled and normalized. The minimum welding spacing threshold and weld thickness change rate specified in the welding process are combined to perform segmentation error statistics. Finally, a serialization output module is used to generate morphological difference data between welding point sequences, including a welding point position sequence, an angle change matrix, and a length difference vector group.

[0062] Step S2: performing a welding pool morphology deviation simulation analysis on the morphology difference data between the welding point sequences according to the basic setting parameters of the robotic arm welding gun to obtain welding pool morphology deviation data; performing strength loss posture association mining based on the welding pool morphology deviation data to obtain strength loss posture association data;

[0063] In an embodiment of the present invention, first, according to the morphological difference data between the welding spots obtained in step S1, combined with the basic setting parameters of the robotic arm welding gun, the process parameters such as voltage, current, wire feeding speed, robotic arm welding gun posture and gas shielding flow in the setting are extracted, and a three-dimensional instantaneous heat source intensity field model based on Gaussian distribution is established through the heat source modeling module, and the heat source input parameters are grid discretized using the finite volume method. Then, the spatial distribution data of the welding spots are imported, and the point-to-point path motion trajectory between the welding spots is simulated. The multi-step linear interpolation algorithm is called to realize the trajectory continuity simulation. The instantaneous temperature distribution under different welding paths is numerically integrated step by step by the transient heat conduction equation to obtain the instantaneous temperature gradient field of each welding spot area. Then, the temperature rise effect is iteratively calculated in combination with the specific heat capacity, thermal conductivity, density and other data in the material thermophysical parameter database to form the welding pool morphology deviation data. Then, based on the deviation data, the strength loss posture correlation mining is performed. First, the temperature gradient field is reconstructed by multi-scale local stress field, and Von The Mises yield criterion is used to quantify the strength loss index of the thermal stress at each unit grid point. A three-dimensional vector angle fitting is performed between the welding posture angle change and the local stress loss distribution. Based on the Pearson correlation coefficient, a point-by-point correlation analysis is performed on each posture angle variable and the strength loss. The posture angle change interval with an absolute value of the correlation coefficient greater than 0.85 is selected to form the strength loss posture correlation data.

[0064] Step S3: Perform joint angle spatial limit matching based on the strength loss posture correlation feature data to generate joint angle spatial limit matching data; perform robot arm welding gun posture planning and design based on the joint angle spatial limit matching data to obtain robot arm welding gun posture planning data.

[0065] In an embodiment of the present invention, the strength loss posture association data obtained in step S2 is first input into the joint angle space limit matching module, and the spatial position and posture angle of the end effector of the manipulator are solved by the inverse kinematics solution algorithm. The relationship between the joint angular velocity and the end linear velocity is solved by a method based on the singular value decomposition of the Jacobian matrix. Combined with the actual hardware joint limit angle constraint range of the manipulator, joint angle limit detection is performed to identify posture angle solutions with angle out-of-bounds risks. The nonlinear constraint optimization algorithm is used to perform angle fallback and re-solve infeasible solutions to ensure that all solutions meet the kinematic constraints of the manipulator. Subsequently, the minimum strength loss interval is screened according to the linear fitting slope of the posture angle change and the strength loss index to generate joint angle space limit matching data that meets the minimum strength loss. On this basis, posture trajectory smoothing processing is performed, and the joint angle change curve is smoothed and fitted by the quintic polynomial interpolation method to eliminate transient angular velocity fluctuations, forming a continuous, smooth, and executable posture angle trajectory planning path. Collision detection and reachability verification are performed through the trajectory feasibility simulation platform, and finally the posture planning data of the manipulator welding gun is output.

[0066] Step S1 includes the following steps:

[0067] Step S11: Obtain product welding structure diagram and basic setting parameters of the robotic arm welding gun;

[0068] Step S12: vectorizing the product welding structure diagram to generate a product welding structure vector diagram;

[0069] Step S13: extracting the welding requirement form from the product welding structure vector diagram to obtain product welding requirement form data;

[0070] Step S14: performing morphological difference analysis between welding point sequences on the product welding requirement morphological data based on the product welding structure vector diagram, thereby obtaining morphological difference data between welding point sequences.

[0071] In an embodiment of the present invention, a three-dimensional laser scanner is first used to perform a full-surface scan of the product to be welded, and the product shape data is imported into a CAD environment through a high-density point cloud data acquisition device. The point cloud data is reconstructed using industrial CAD software, and the surface segments are fitted point by point using a NURBS surface fitting algorithm. The scanned data is converted into a product welding structure drawing file in a standard STEP format, and then the basic setting parameters of the robotic arm welding gun are extracted from the welding process document provided by the process department. The basic setting parameters include the definition parameters of the welding gun end coordinate system, the preset range of the attitude angle, the setting range of the wire feeding speed, the arc voltage setting range, and the welding current setting range. The setting range of the welding gun attitude angle is 45° to 75°, the wire feeding speed is 10 to 15 m / min, the arc voltage is 20 to 30 V, and the welding current is 150 to 250 A. All parameters are input into the parameter database through the numerical input method. Based on the product welding structure drawing file obtained in step S11, the CAD graphic data vectorization processing module is called, and a vectorization algorithm based on contour tracking is used to extract the paths of all weld paths in the drawing. Each weld line segment is subjected to segmented curve fitting processing, and the path continuity is reconstructed using a spline curve interpolation algorithm. Subsequently, the three-dimensional spatial coordinate point of each weld point is projected into the global workpiece coordinate system using a spatial coordinate system mapping method, and a weld point sequence index table is established. The attribute parameters such as weld thickness and width are linearly encoded by attribute vectors, and a structural attribute labeling method is used to assign a unique identification code to each weld line segment. Finally, a product welding structure vector diagram containing weld path point coordinates, path direction vector, and weld size attributes is output. Based on the product welding structure vector diagram generated in step S12, a feature extraction algorithm based on morphological filtering is used to automatically identify the welding requirement areas in the structure vector diagram. First, the welding path vector is subjected to regional clustering processing, and the continuous weld areas are classified using the density peak clustering algorithm. After classification, each welding area is numerically encoded with three dimensions of path length, weld width, and weld thickness. Subsequently, attribute feature screening is performed on all welding areas, and non-process specified welding areas are eliminated. The threshold discrimination method is used to set the welding thickness threshold to be greater than 2mm for screening. The areas that meet the welding requirements are screened for angle attribute matching in combination with the weld groove angle database. The welding point is resampled for each welding area that passes the screening, and the linear uniform sampling method is used to regenerate the weld point coordinate sequence with a point spacing of 1mm. Finally, the product welding requirement morphological data containing the welding area identification, weld point coordinate sequence, welding direction vector, and weld size characteristics is output.Using the product welding requirement morphological data obtained in step S13 as input and combining it with the global coordinate system information in the product welding structure vector diagram, a sequential relationship modeling process is performed on all weld point position pairs. First, the Dijkstra path algorithm based on the shortest path priority is called to calculate the point-to-point path length of the weld point sequence and establish the minimum motion distance matrix between weld points. Then, the spatial vector angle analysis method is used to measure the angle between the path vectors of adjacent weld points and extract the three-dimensional spatial angle change between the weld point sequences. Then, the length change, spatial rotation angle, and weld width change rate in the weld point sequence are numerically counted. The data of different physical quantity dimensions are dimensionlessly processed using a standardization method. The normalized data is subjected to the K-means clustering algorithm to group and classify the morphological differences and identify the morphological mutation point sequence. Finally, the morphological difference data between the weld point sequences including the weld point sequence number, the angle change between the point pairs, the path length change, and the morphological mutation point identifier are output.

[0072] Step S2 includes the following steps:

[0073] Step S21: extracting the welding gun setting posture from the basic setting parameters of the robot arm welding gun;

[0074] Step S22: performing a welding pool morphology deviation simulation analysis on the morphology difference data between the welding point sequences according to the basic setting parameters of the robot arm welding gun and the setting posture of the welding gun to obtain welding pool morphology deviation data;

[0075] Step S23: performing weld spot strength loss tolerance fitting on the weld pool morphology deviation data to generate weld spot strength loss tolerance data;

[0076] Step S24: performing strength loss posture association mining on the welding gun setting posture based on the weld point strength loss tolerance data to obtain strength loss posture association data.

[0077] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0078] Step S21: extracting the welding gun setting posture from the basic setting parameters of the robot arm welding gun;

[0079] In an embodiment of the present invention, the welding gun setting posture parameters corresponding to the current product process are extracted from the basic setting parameter database of the robotic arm welding gun obtained in step S1, specifically including the spatial posture matrix of the welding gun end, the gun body deflection angle, the tilt angle, the thrust angle, the drag angle and the angular velocity limit threshold. The thrust angle range in the setting posture parameters is set to 10° to 20°, and the drag angle range is set to 5° to 15°. The end posture coordinate system is jointly described by a three-dimensional position vector and a rotation matrix. For different spatial azimuth changes during the welding process, the Euler angle description method is used to solve the rotation matrix into a posture angle triplet. The inverse kinematics equation group is further used to perform real-time analysis of the angular position of each joint of the robotic arm to obtain the welding gun setting posture sequence data corresponding to the spatial position of the target weld point.

[0080] Step S22: performing a welding pool morphology deviation simulation analysis on the morphology difference data between the welding point sequences according to the basic setting parameters of the robot arm welding gun and the setting posture of the welding gun to obtain welding pool morphology deviation data;

[0081] In the embodiment of the present invention, the welding gun setting posture data obtained in step S21 and the morphological difference data between the welding point sequences output in step S1 are first read. Based on the morphological difference characteristics such as the change in the space vector angle and the change in the path length, the spatial posture reconstruction is performed in combination with the thrust angle and drag angle parameters in the welding gun setting posture. A high-density posture transition trajectory point set is generated between the welding points through a three-dimensional linear interpolation method. Then, the welding process heat input parameters are assigned to each posture point. The heat input parameters include a welding current of 180A, a welding voltage of 25V, a wire feeding speed of 12m / min, and a gas shielding flow rate of 20L / min. A transient heat source volume model is constructed based on these process parameters, and the three-dimensional finite volume method is used to discretize the welding path segment grid. The mesh size is 0.5 mm³, and the temperature gradient distribution and heat flux density vector distribution are calculated for each unit node. The temperature field in the welding interval is solved by multi-step time iteration based on the thermal conductivity, specific heat capacity and density data in the material thermophysical parameter library. The explicit Euler method is used for time step control, and the time step is set to 1 ms. The volume integration of the molten pool temperature distribution under different postures is performed, and the isothermal surface of the molten pool geometry is extracted. The isothermal surface temperature threshold is set to 1450 ° C to form a point cloud of the molten pool boundary geometry. Finally, the spatial error of the molten pool boundary change during all posture angle changes is quantified to generate welding pool morphology deviation data including posture angle sequence, molten pool boundary coordinate change, and temperature gradient peak difference.

[0082] Step S23: performing weld spot strength loss tolerance fitting on the weld pool morphology deviation data to generate weld spot strength loss tolerance data;

[0083] In an embodiment of the present invention, based on the welding pool morphology deviation data obtained in step S22, the molten pool morphology deviation interval is first discretized, and the deviation range is divided into three dimensions: length, width, and depth. Each dimension is divided into 0.1 mm intervals. Subsequently, the multiple regression analysis method is used to fit and model the molten pool deviation size and weld strength loss historical database. The strength loss data in the historical database comes from the standardized shear strength experiment. In the experiment, an Instron universal material testing machine is used to perform shear strength tests on different welding deviation samples. The loading rate is set to 5 mm / min, and the maximum load is recorded as the final strength data. During the fitting process, the least squares method is selected to perform linear regression on the molten pool deviation size and shear strength loss, and the residual sum of squares is minimized. The determination coefficient R² of the fitting curve is calculated, and the R² threshold is set to be not less than 0.85 to ensure fitting accuracy. Finally, the weld strength loss tolerance data including the length direction tolerance interval, the width direction tolerance interval, the depth direction tolerance interval and the corresponding strength loss prediction value interval are output.

[0084] Step S24: performing strength loss posture association mining on the welding gun setting posture based on the weld point strength loss tolerance data to obtain strength loss posture association data.

[0085] In an embodiment of the present invention, the weld strength loss tolerance data obtained in step S23 is read, and combined with the welding gun posture setting sequence data in step S21, a multivariate correlation analysis method is used to perform a variable-by-variable relationship analysis on the posture angle parameter and the strength loss tolerance, and the Pearson correlation coefficient is calculated for the posture parameter variables such as the push angle, drag angle, and gun body deflection angle in turn. The calculation process normalizes the numerical distribution within each group of posture parameter change intervals and the strength loss tolerance interval, and the variable sample size is set to 300 groups of data points. The posture variables with an absolute value of the correlation coefficient exceeding 0.8 are screened for significance, and then the screened posture variables are subjected to principal component analysis, and the first principal component is extracted as the representative eigenvector of the posture angle change. Cluster analysis is performed on different posture intervals based on the eigenvector direction, and the K-means clustering method is used to divide the posture change samples into high-risk, medium-risk, and low-risk areas for strength loss. Finally, the output includes the strength loss posture association data described by the posture angle change interval, the strength loss risk level, and the principal component eigenvector.

[0086] Step S22 includes the following steps:

[0087] Step S221: extracting the welding setting voltage, current and speed from the basic setting parameters of the robot arm welding gun; calculating the point-to-point vector angle between the welding points according to the set posture of the welding gun, thereby obtaining the vector angle between the welding point and the welding gun;

[0088] Step S222: performing vector angle change analysis on the morphological difference data between the welding points in sequence based on the morphological difference data between the welding points in sequence, and obtaining posture change vector angle data between the welding points and the welding gun;

[0089] Step S223: simulating and analyzing the instantaneous temperature field of the weld pool during the multi-frequency welding process based on the posture change vector angle data according to the welding set voltage, current, and speed to obtain the instantaneous temperature field of the weld pool during the multi-frequency welding process;

[0090] Step S224: performing anisotropic fluid solidification behavior imbalance analysis on the instantaneous temperature field of the molten pool at multiple frequencies of the welding spot, thereby obtaining anisotropic fluid solidification behavior imbalance data;

[0091] Step S225: performing interfacial tension vibration regression analysis on the unbalanced solidification behavior data of the anisotropic fluids to obtain interfacial tension vibration regression data;

[0092] Step S226: performing a welding pool morphology deviation simulation analysis on the morphology difference data between welding spot sequences according to the anisotropic fluid solidification behavior imbalance data and the interfacial tension vibration regression data to obtain welding pool morphology deviation data.

[0093] In an embodiment of the present invention, a process parameter reading module is first called to extract key process parameters such as welding setting voltage, current and speed from the basic setting parameter database of the robotic arm welding gun, where the voltage is 24V, the current is 200A, and the welding speed is set to 12mm / s. Subsequently, based on the welding gun setting posture data obtained in step S21, the spatial coordinates of all welding points are read in sequence, and the three-dimensional vector angle calculation method is used to calculate the point-to-point vector angle between the spatial position vectors of adjacent welding points and the normal vector of the welding gun end. The specific calculation process uses the vector dot product method to solve the cosine value of the angle, and then the angle value is obtained by the inverse cosine function. During the calculation process, the coordinate units are unified in millimeters, and the spatial distance and posture angle difference between each pair of welding points are synchronously recorded, and finally the vector angle data between the welding point and the welding gun is generated, including the welding point number, spatial position coordinates, and point-to-point vector angle value. Read the vector angle data between the weld point and the welding gun obtained in step S221 and the morphological difference data between the weld point sequences obtained in step S1. First, perform sequence synchronization processing on the two sets of data to ensure that the angle data and the weld point sequence are strictly one-to-one corresponding. Use the central difference method to calculate the vector angle change rate of the angle data of adjacent weld points. During the calculation process, the angle difference of each pair of consecutive weld points is divided by the actual spatial distance between the two points to obtain the angle change rate per unit length. Then, perform statistical feature analysis on the angle change rate of the entire weld point sequence to extract the maximum angle change rate, minimum angle change rate, and angle difference between the two weld points. The three statistics of angle change rate and average angle change rate are standardized to make the numerical range between 0 and 1. The angle change rate curve is fitted by segment fitting based on the change trend fitting method of piecewise linear regression, and the inflection point of the fitting curve is identified. The trend turning point position of the angle change interval is determined by the first-order derivative mutation point detection method, and the trend inflection point is marked in segments. Finally, the posture change vector angle data containing multiple parameters such as welding point number, point-to-point angle change rate, trend inflection point position, and piecewise fitting slope value are output.Based on the posture change vector angle data obtained in step S222, combined with the process parameters such as voltage, current and welding speed extracted in step S221, the instantaneous temperature field of the weld pool in the multi-frequency welding process of different posture change intervals is simulated and analyzed. First, a three-dimensional Gaussian distribution heat source model is established using the transient heat source input method, and the heat source parameters are numerically set, where the heat source radius is set to 2mm, the heat source power density is calculated based on the input current and voltage, the time step is set to 0.5ms, and the side length of the space grid division unit is set to 0.2mm. The welding path in each angle change section is discretized by linear interpolation, and the discrete path points are discretized. Point-by-point heat input simulation is performed, and the temperature field control equation is numerically solved using the finite volume method. The boundary conditions are set as natural convection boundaries, and the initial temperature is set to room temperature 25°C. The implicit time integration method is used in the heat conduction solution to improve the calculation stability. The temperature value of each grid unit node is recorded at each moment, and the three-dimensional matrix data of the instantaneous temperature field of the multi-frequency solder joint molten pool is output. The maximum isothermal volume, temperature peak position, and temperature gradient direction vector of the molten pool at each time step are extracted and serialized and stored time by time, ultimately forming the multi-frequency solder joint molten pool instantaneous temperature field data containing time dimension, spatial coordinates, temperature gradient vector, and multi-frequency isothermal volume sequence.

[0094] First, read the three-dimensional matrix data of the instantaneous temperature field of the multi-frequency weld pool generated in step S223, use the temperature gradient tensor field solution method to calculate the temperature gradient vector distribution at different time steps, use the spatial gradient discretization method to numerically approximate the first-order partial derivative of the temperature gradient of each grid unit, perform spatial second-order derivative analysis on the temperature gradient change rate to extract the local heat flow direction change rate, combine the latent heat release interval data of the metal liquid and solid transition temperature zones in the material thermal property parameter table, set the liquid-solid transition critical temperature range to 1400℃ to 1450℃, perform time series energy flux integration on the unit nodes in this temperature zone, use the cumulative energy method to compare the heat flux density vectors in different directions, and analyze the solidification. The velocity of the solid interface movement is numerically calculated along different spatial directions using the differential method, and the velocity vector field at the interface front is extracted. Based on the velocity vector field distribution, vector variance analysis is performed on the degree of solidification velocity heterogeneity in different orientations. The anisotropic fluid dynamics criterion is used to perform a numerical correlation coefficient analysis on the correlation between the velocity variance distribution and the directional change of the heat flux density. The regions with high differences in solidification velocity and directional deviations of the heat flux density gradient are clustered, and the density-based spatial clustering algorithm DBSCAN is used to divide the anisotropic solidification zone into independent imbalance feature zones. The final output is the imbalance data of the anisotropic fluid solidification behavior, which includes the solidification velocity variance, the directional deviation angle of the heat flux density, the position index of the anisotropic zone, and the imbalance block identifier. Based on the imbalance data of solidification behavior of anisotropic fluids obtained in step S224, the solidification front velocity time series data and the interface temperature gradient change series data in each imbalance characteristic area are extracted. First, the solidification front velocity time series is subjected to fast Fourier transform, and the spectrum of the vibration main frequency is extracted. The frequency analysis range is set to 0 to 200 Hz. Then, the local characteristics of the vibration frequency time are analyzed by wavelet transform. The Morlet wavelet basis function is selected for multi-scale decomposition. The frequency amplitude change trend at different time scales is fitted. The frequency amplitude curve at each scale is multi-scaled using the least squares method. For the interface temperature gradient change sequence, the sliding average filter method is used to remove high-frequency noise interference. The filtered gradient change sequence is processed by first-order difference to extract the change rate. The interface vibration frequency data and the gradient change rate data are subjected to Pearson correlation analysis. The correlation coefficient is calculated to determine the degree of dependence of the vibration frequency on the temperature gradient change. The nonlinear least squares method is used to perform quadratic polynomial regression modeling on the high-correlation segment data. The output includes the interface tension vibration regression data including the main peak value of the vibration frequency, the amplitude change rate, the interface temperature gradient change rate, the regression curve coefficient and the sum of squares of the fitting residuals.The imbalance data of solidification behavior of anisotropic fluids obtained in step S224 and the interfacial tension vibration regression data output in step S225 are read, and the morphological difference data between the welding points are used as the spatial geometry input variable. The point-to-point angle change, path length change and imbalance block position in the morphological difference data are paired. The solidification rate variance, heat flow direction deviation angle, vibration frequency main peak value and gradient change rate are used as explanatory variables using the multivariate regression analysis method, and the morphological difference spatial deviation is used as the dependent variable. A multivariate linear regression equation group is constructed, and the gradient descent method is used to solve the parameters. The iteration step size is set to 0.01, the maximum number of iterations is 10,000 times, and the residual tolerance threshold is set to. During the calculation process, the input variables are Z-score standardized to avoid the influence of multivariate dimensions, the regression residuals are tested for normality, and the P value of the residual distribution is calculated using the Shapiro-Wilk test method to ensure the statistical reliability of the fitting results. Finally, the deviation prediction results after fitting are denormalized and restored, and the spatial deviation of each weld point sequence segment is numerically restored in a three-dimensional coordinate system to generate welding pool morphology deviation data containing multiple parameters such as point pair angle deviation, path length deviation, and spatial position deviation vector.

[0095] Step S224 includes the following steps:

[0096] The heat flux density non-uniformity vector is identified for the instantaneous temperature field of the molten pool of multiple-frequency solder joints to obtain the heat flux density non-uniformity vector;

[0097] Based on the heat flux density inhomogeneity vector, differential coupling of the time series change of the solder joint surface tension is performed to generate differential coupling data of the time series change of tension;

[0098] The local variance of the molten metal viscosity is calculated based on the differential coupling data of the heat flux density inhomogeneity vector and the tension time series variation, and the local variance of the molten metal viscosity is obtained;

[0099] Based on the heat flux density inhomogeneity vector, the differential coupling data of the tension time series change and the local variance of the molten metal viscosity, the heat energy release ratio difference of the solidification latent heat between different solder joint orientations is deduced to generate the heat energy release ratio difference between different solder joint orientations;

[0100] The imbalance analysis of solidification behavior of anisotropic fluid is performed based on the difference in heat energy release ratios between different solder joint orientations, thereby obtaining the imbalance data of solidification behavior of anisotropic fluid.

[0101] In the embodiment of the present invention, in the first sub-step of step S224, the three-dimensional matrix data of the instantaneous temperature field of the multi-frequency weld pool generated in step S223 is read, and the heat flux density vector analytical method based on the control volume method is used to numerically calculate the temperature gradient of each grid unit in the temperature field. The temperature gradient of each grid unit is discretized by the first-order partial derivative in three-dimensional space using the central difference format. According to Fourier's heat conduction law, the thermal conductivity parameter of the material is used, and the thermal conductivity value of the material is set to 45W / (m·K). The heat flux density of each grid unit in the X, Y, and Z directions is calculated. Point-by-point vectorized calculation is performed, and then the statistical variance method is used to perform spatial variance analysis on the heat flux density vector of each grid unit in the entire temperature field. The grid units with spatial variance more than 1.5 times the average level are marked as heat flux density inhomogeneity areas. The angular deviation distribution analysis of the heat flux vector direction in the heat flux density inhomogeneity area is performed, and the local area that deviates from the overall heat flux direction by more than 30° is extracted. Finally, the heat flux density inhomogeneity vector data containing the grid unit position index, heat flux density vector component, direction deviation angle and absolute value of spatial variance are output to complete the heat flux density inhomogeneity vector identification.

[0102] In the second step of step S224, based on the heat flux density non-uniformity vector data obtained in the first step, the time series heat flux density change data of each non-uniform heat flux area is extracted, and the heat flux density time series is differentiated using the first-order time difference method. The differentiated data is subjected to a sliding average filter, and the filter window width is set to 5ms. The filtered heat flux density differential sequence and the surface temperature gradient change sequence in the corresponding area are synchronously normalized. Subsequently, based on the surface tension gradient equation, the normalized heat flux density differential sequence is linearly weighted, and the weight coefficient is set according to the surface tension temperature coefficient of the metal material, and the setting range is -0.4 to -0.6. The tension gradient changes within the time step are accumulated and summed by the numerical integration method, and finally the differential coupling data of the tension time series changes in each inhomogeneous area are obtained. The data format includes the time step sequence, the tension change rate per unit time, the corresponding heat flux density change rate, and the cumulative tension change, completing the generation of the differential coupling data of the tension time series changes.

[0103] In the third sub-step of step S224, the heat flux density heterogeneity vector data obtained in the first sub-step and the tension time-series variation differential coupling data obtained in the second sub-step are read, and the data in the same spatial unit are paired one-to-one. The local viscosity gradient sensitivity analysis method is used to perform bivariate correlation modeling based on the heat flux density vector direction change rate and the tension change rate. The local heat flux change rate and the tension change rate are linearly fitted using the least squares method. The fitting residual variance is used as the local viscosity variation heterogeneity indicator. The residual variance of all non-homogeneous area units is statistically processed. The viscosity variation fluctuation coefficient in the local area is calculated using the coefficient of variation method. The area with a fluctuation coefficient exceeding 0.2 is marked as a viscosity fluctuation area. The time series variance of the viscosity change rate in the fluctuation area is calculated. Finally, the local variance data of the molten metal viscosity including the grid unit position, the heat flux density change rate, the tension change rate, the fitting residual variance and the local viscosity fluctuation coefficient are output to complete the local variance calculation of the molten metal viscosity.

[0104] In the process of deducing the difference in the solidification latent heat release ratio between different solder joint orientations based on the heat flux density inhomogeneity vector, the tension time series variation differential coupling data and the local variance of the molten metal viscosity, the heat flux density vector component of each grid unit in the heat flux density inhomogeneity vector data is first read, and the unit tension change rate in the corresponding time step in the tension time series variation differential coupling data and the viscosity change variance value of each unit in the molten metal viscosity local variance data are combined to perform intra-unit aggregation processing on the data at the same solder joint orientation. The multivariate weighted integration method is used to calculate the energy conversion coefficient of the local heat flux density vector modulus, tension change rate and viscosity variance in a single solder joint orientation. The heat flux density weight factor is set to 0.5, the tension change rate weight factor is set to 0.3, and the viscosity variance weight factor is set to 0.2. The weight factors are calculated based on the previous The calibration data of the thermophysical property experiment was determined, and the polymerization energy conversion coefficient in different solder joint orientations was standardized. The normalization interval was set to 0 to 1. The normalized energy conversion coefficient was multiplied by the theoretical value of solidification latent heat of the corresponding orientation. The theoretical value of latent heat was set to 270 kJ / kg. The solidification latent heat release per unit volume of solder joints in different orientations was accumulated and integrated. The step integration method with a time step of 1 ms was used to ratio the cumulative latent heat release of all solder joint orientation units. The orientation with the largest latent heat release was selected as the reference orientation, and the latent heat release of other orientations was normalized and ratioed. Finally, the heat release ratio difference data between solder joints with different orientations was generated, including the orientation number, heat energy release per unit volume, latent heat ratio difference, and cumulative energy curve slope, completing the heat release ratio difference deduction process between solder joints with different orientations.In the process of analyzing the imbalance of solidification behavior of anisotropic fluids based on the difference in heat energy release ratios between different weld orientations, the heat energy release ratio difference data between different weld orientations are first read, the standard deviation analysis of the difference sequence is performed, the mean and standard deviation of the latent heat release ratios between all weld orientations are calculated, and the imbalance prediction is performed on the weld orientation pairs whose standard deviation exceeds the set threshold. The threshold value is 0.15. The significance of the difference in heat energy release ratios between different orientations is tested by performing a single-factor variance analysis on the heat energy release ratio difference between the orientations. The significance level α is set to 0.05, and the solidification velocity gradient analysis is further performed on the significant orientation pairs. The solidification interface velocity vector of the corresponding orientation unit is extracted, and the solidification velocity vector is calculated using the Euler angle method. The angle change rate between the heat flux vector and the heat flux density vector is measured, and the imbalance zone is marked for areas with an angle change rate greater than 25° / mm. A multivariate covariance analysis is performed on the temperature gradient change rate, viscosity fluctuation coefficient, and tension change rate within the imbalance block. A multivariate collaborative imbalance index system among the heat flux, surface tension, viscosity, and latent heat release ratio difference is established. The principal component analysis method is used to rank the weights of each influencing factor, extract the main imbalance factors, reconstruct the spatial position of the imbalance zone, and output the imbalance data of the solidification behavior of anisotropic fluids, including the spatial coordinates of the imbalance block, solidification velocity gradient, heat energy release difference, viscosity fluctuation coefficient, tension change rate, and principal component score, to complete the imbalance analysis of the solidification behavior of anisotropic fluids.

[0105] Step S23 includes the following steps:

[0106] Step S231: performing a molten pool morphology deviation network division process on the welding molten pool morphology deviation data to obtain molten pool morphology deviation network division data;

[0107] Step S232: performing a melt pool geometric deviation feature analysis on the melt pool morphology deviation network division data, thereby obtaining melt pool geometric deviation feature data;

[0108] Step S233: simulating the weld point collision force offset distribution based on the molten pool geometric deviation characteristic data to generate collision force offset distribution data;

[0109] Step S234: performing stress distribution strain rate component decomposition on the collision force offset distribution data to obtain strain rate component decomposition data;

[0110] Step S235 : performing weld strength loss tolerance fitting based on the strain rate component decomposition data to generate weld strength loss tolerance data.

[0111] In an embodiment of the present invention, the welding pool morphology deviation data generated in step S22 is spatially topologically divided and processed, and a voxel space grid division technology based on the region growing method is adopted. First, the entire welding pool area is uniformly voxelized according to the three-dimensional grid coordinate system, and the voxel size is set to 0.1 mm³. The three-dimensional gradient calculation is performed on the molten pool morphology deviation value within the voxel unit. The threshold value of the spatial gradient modulus greater than twice the average gradient modulus is used as the abnormal deviation area judgment criterion. The abnormal deviation areas of adjacent voxel units are connected by the depth-first search algorithm. Finally, the welding pool morphology deviation data is divided into multiple deviation network sub-areas with spatial adjacency. Each sub-area outputs the deviation network number, the number of voxel units, the average deviation intensity and the maximum deviation gradient direction vector to generate the molten pool morphology deviation network division data. In step S232, the geometric feature analysis of the molten pool morphology deviation network division data is performed, and the ellipsoid parameters of the spatial distribution of each deviation network sub-area are calculated based on the three-dimensional minimum circumscribed ellipsoid fitting algorithm. The angle between the major axis, minor axis, and principal axis direction cosines and the spatial main direction vector is extracted, and the three-axis spatial discreteness index of the deviation network area is calculated. The deviation sub-area with a major axis length greater than 3 mm and a major-minor axis ratio greater than 2 is marked as a high anisotropy deviation area. At the same time, the deviation intensity distribution in the deviation area is subjected to a third-order statistical moment analysis, and the deviation kurtosis and skewness coefficient are calculated. The area with a deviation kurtosis greater than 3 is subjected to high peak state identification. Finally, the molten pool geometric deviation feature data including spatial morphology ellipsoid parameters, deviation discreteness, deviation kurtosis, skewness coefficient and principal axis direction angle is output.

[0112] Based on the geometric deviation characteristic data of the molten pool, the collision force offset distribution of the weld spot is simulated. The finite volume method is used to initialize the residual stress distribution field in the weld spot area. The initial residual stress data comes from the previous heat conduction calculation results. The equivalent volume force perturbation is applied to each unit in the geometric deviation characteristic block of the molten pool. The magnitude of the perturbation force is linearly amplified according to the deviation intensity, and the amplification factor is set to 5N / mm³ deviation unit. The quasi-static mechanical response simulation of the weld spot area under the perturbation force is performed, and the displacement vectors of the nodes in the weld spot area are calculated. The displacement vectors are reconstructed by three-dimensional interpolation to obtain the relative displacement distribution between the nodes. The displacement vector gradient is extracted, and the offset peak is identified in the area where the gradient is greater than the threshold. The threshold is set to 0.1mm. The collision force offset distribution data including spatial coordinates, node displacement vectors, offset gradients and force concentration factors are output for the entire weld spot area. The stress distribution strain rate component decomposition of the collision force offset distribution data is performed. First, the force vector and displacement vector of each node are read, and the differential derivative of the force-displacement relationship between the nodes is calculated to obtain the strain rate tensor of each node. The strain rate tensor is decomposed by eigenvalue, and the principal strain rate components are extracted. The principal strain rate components are projected in the spatial direction and decomposed into radial, tangential and axial components using the spherical coordinate system. At the same time, each component is normalized with the normalization interval set to -1 to 1. The Pearson correlation coefficient is calculated for the coupling relationship between the components to determine the coupling strength between components in different directions. Finally, the strain rate component decomposition data including the node coordinates, principal strain rate values, component values ​​in each direction and component coupling coefficients are output. Based on the strain rate component decomposition data, the weld strength loss tolerance fitting is performed. The multivariate linear regression analysis method is used, with the radial strain rate, tangential strain rate and axial strain rate of each node as independent variables, and the weld shear strength loss rate obtained in the previous experiment as the dependent variable. The regression coefficient fitting is performed. The number of regression sample points is set to 1000, covering the entire strain rate component variation range. The regression model is tested for residual variance, and a residual variance less than 0.05MPa² is considered a passing standard for fitting. The regression model is used to predict the strength loss rate of all weld area nodes. The predicted strength loss rate is combined with the weld design strength for a tolerance bandwidth analysis. The strength safety margin threshold is set to 15%, and the strength loss rate prediction value and tolerance interval of each weld position are output to finally generate the weld strength loss tolerance data.

[0113] Step S3 includes the following steps:

[0114] Step S31: performing feature learning on the intensity loss posture associated data to obtain intensity loss posture associated feature data;

[0115] Step S32: performing joint angle space limit matching on the welding gun set posture according to the strength loss posture correlation feature data to generate joint angle space limit matching data;

[0116] Step S33: performing singular configuration optimization adjustment on the joint angle space limit matching data, thereby obtaining joint angle space limit optimization data;

[0117] Step S34: performing robot arm welding gun posture planning and design through joint angle spatial limit optimization data to obtain robot arm welding gun posture planning data.

[0118] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0119] Step S31: performing feature learning on the intensity loss posture associated data to obtain intensity loss posture associated feature data;

[0120] In an embodiment of the present invention, feature learning processing is performed on the strength loss posture association data obtained in step S24. First, a multidimensional feature matrix is ​​constructed for the posture angle sequence, weld point spatial coordinate sequence, and strength loss rate data in the strength loss posture association data. The mean normalization method is used to perform interval standardization on all numerical features, and the normalization interval is set to 0 to 1. The normalized feature matrix is ​​subjected to principal component analysis processing, and the first five principal component eigenvectors with a cumulative contribution rate greater than 85% are extracted. Variance analysis is performed on each principal component feature, and low-variance feature dimensions with a variance less than 0.01 are eliminated. The maximum correlation coefficient discriminant analysis is performed on the remaining feature dimensions, and the Pearson correlation coefficient of each posture angle to the strength loss rate is calculated. Feature variables with an absolute value of the correlation coefficient greater than 0.7 are screened as strongly correlated posture feature variables. Finally, the strength loss posture association feature data including the principal component eigenvectors, variance analysis results, correlation coefficient matrix, and strongly correlated feature variable index sequence are output.

[0121] Step S32: performing joint angle space limit matching on the welding gun set posture according to the strength loss posture correlation feature data to generate joint angle space limit matching data;

[0122] In an embodiment of the present invention, the joint angle space limit matching processing is performed on the set posture of the robotic arm welding gun according to the strength loss posture correlation feature data obtained in step S31. First, the kinematic link parameters of the robotic arm are obtained, including the number of joints, the joint limit angle range, the connecting rod length parameters and the end effector installation azimuth angle. An analytical algorithm based on inverse kinematics is used to perform feasible solution inverse calculation on each strongly correlated feature posture. The posture target is iterated forward and inversely using the DH parameter table of the six-axis robotic arm. For each joint angle sequence obtained by solution, it is determined whether the limit angle range constraints of each joint are met. The limit angle range error tolerance is set to plus or minus 0.5°. The solved postures that exceed the limit interval are eliminated, and the posture solution uniqueness is judged for the solved solutions that meet the limit interval. Redundant solutions are removed and the remaining solution sets are sorted by minimizing the sum of squares of the target posture angle errors. Finally, the joint angle space limit matching data including the feasible solution joint angle sequence, the corresponding spatial target posture, the joint limit constraint verification results and the sum of squares of the target posture angle errors are output.

[0123] Step S33: performing singular configuration optimization adjustment on the joint angle space limit matching data, thereby obtaining joint angle space limit optimization data;

[0124] In an embodiment of the present invention, the joint angle space limit matching data generated in step S32 is subjected to singular configuration optimization adjustment. First, the Jacobian matrix of each feasible solution is derived, and the numerical differentiation method is used to calculate the differential increment of each component of the Jacobian matrix. The step size is set to 0.01°. The obtained Jacobian matrix is ​​subjected to singular value decomposition, and the minimum singular value and condition number are extracted. The posture solution with the minimum singular value less than 0.001 or the condition number greater than 1000 is determined to be a singular configuration existence solution. The singular configuration existence solution is optimized using a small perturbation configuration correction method. The specific operation is to apply a Gaussian distribution small perturbation to the joint angle sequence in the main sensitive direction, with the perturbation mean set to 0 and the standard deviation set to 0.2°. After the perturbation, the Jacobian matrix of the new solution is recalculated until the minimum singular value is greater than 0.005 and the condition number is less than 500. Finally, the joint angle space limit optimization data including the optimized joint angle sequence, the corresponding singular value, the condition number and the number of corrected iterations is output.

[0125] Step S34: performing robot arm welding gun posture planning and design through joint angle spatial limit optimization data to obtain robot arm welding gun posture planning data.

[0126] In an embodiment of the present invention, the joint angle space limit optimization data obtained in step S33 is used to plan and design the posture of the robotic arm welding gun. First, the trajectory of the robotic arm end effector is fitted with cubic spline interpolation according to the optimized joint angle sequence, and the interpolation point spacing is set to 2 mm. The velocity profile is planned for the interpolation trajectory, and a trapezoidal velocity curve control strategy is adopted. The maximum speed is set to 50 mm / s, the acceleration limit is set to 500 mm / s², and the velocity profile is sampled in discrete time steps, and the time step is set to 10 ms. The joint angle change within each time step is differentially calculated to generate a real-time joint angular velocity instruction sequence, and the joint angular velocity instruction is trajectory smoothed. The angular velocity sequence is smoothed using a five-point weighted sliding average method, and finally the robotic arm welding gun posture planning data including the time series, joint angle sequence, joint angular velocity sequence, and end effector trajectory coordinate point sequence is output.

[0127] Step S32 includes the following steps:

[0128] Step S321: performing joint verticality matching on the welding gun set posture according to the strength loss posture correlation feature data to obtain joint verticality matching data;

[0129] Step S322: performing adaptive adjustment control of the thrust angle / drag angle of the robot arm welding gun according to the joint verticality matching data and the strength loss posture correlation feature data to obtain thrust angle / drag angle adaptive control data;

[0130] Step S323: performing feed welding speed matching on the push angle / drag angle adaptive control data, thereby obtaining feed welding speed matching data;

[0131] Step S324: performing joint angle space limit matching on the welding gun setting posture based on the joint verticality matching data, the push angle / drag angle adaptive control data and the feed welding speed matching data to generate joint angle space limit matching data.

[0132] In an embodiment of the present invention, the joint verticality matching is performed on the set posture of the robotic arm welding gun according to the strength loss posture association feature data obtained in step S31. First, the posture offset angle feature and the spatial direction vector feature in the strength loss posture association feature data are extracted, and the posture normal vector of each set posture is extracted by the normal vector angle projection method. The angle between the joint end normal vector and the product welding surface normal vector is calculated. The angle calculation adopts the spatial vector dot product angle formula, and the joint verticality deviation tolerance is set to plus or minus 3°. All posture angle sequences with angles less than the tolerance are screened, and the joint limit angle interval verification is performed on each qualified posture angle sequence. The posture sequence with out-of-bounds joint angles is eliminated, and finally the joint verticality matching data including the joint angle sequence, normal angle data and joint angle interval verification results are output. According to the joint verticality matching data obtained in step S321 and the strength loss posture associated feature data obtained in step S31, the propulsion angle and the drag angle of the robotic arm welding gun are adaptively adjusted and controlled. First, the posture angle sequence and the normal angle sequence in the joint verticality matching data are extracted, and the posture angle and strength loss rate features in the strength loss posture associated feature data are extracted. The welding path tangent direction vector is calculated for each posture angle sequence. The propulsion angle and the drag angle are reversely mapped for each posture point using the point-to-tangent direction angle method. The propulsion angle adjustment range is set to plus or minus 15°, and the drag angle adjustment range is set to plus or minus 10°. The linear weight distribution method is used to interpolate and optimize between the posture point and the strength loss rate, and the interpolation step size is set to 1°. The spatial posture positive solution is calculated for each interpolated posture angle sequence, and the path continuity error is judged at the same time. The error threshold is set to 2mm, and the path error exceeding the limit sequence is eliminated. Finally, the propulsion angle and drag angle adaptive control data including the propulsion angle adjustment amount, the drag angle adjustment amount, the interpolation step size, and the path continuity error value are output.

[0133] The push angle and drag angle adaptive control data obtained in step S322 are processed for feed welding speed matching. First, the push angle adjustment amount, drag angle adjustment amount and posture angle sequence are extracted, and the heat input demand coefficient is reversely calculated for each posture point. The arc energy input is calculated according to the welding heat input formula. The voltage is set to 24V and the current is 150A. According to the proportional relationship between the calculated heat input demand and the weld thickness coefficient, the target feed speed of each posture point is reversely corrected using a linear regression equation. The correction coefficient depends on the gradient of the push angle and drag angle. When the gradient change exceeds 5° per section, the feed speed adjustment range is not less than 10%. The speed value range is set to 100 to 300 mm / min. The corrected speed sequence is subjected to secondary smoothing. The smoothing adopts the moving average method with a window width of 5 data points. Finally, the feed welding speed matching data including the posture point sequence, the corresponding feed speed, the heat input coefficient and the speed adjustment range are output. Based on the joint verticality matching data obtained in step S321, the thrust angle and drag angle adaptive control data obtained in step S322, and the feed welding speed matching data generated in step S323, the joint angle space limit matching processing is performed on the posture of the robot arm welding gun. First, the posture angle sequence, thrust angle and drag angle adjustment amount, and feed speed obtained in each step are fused with multivariable posture features, and the joint angle space multidimensional interval reconstruction algorithm is used to remap the posture sequence. During the remapping process, the joint limit angle interval, joint velocity limit, joint acceleration limit and posture angle change rate constraint are considered. Each fusion solution is verified by inverse kinematics forward solution, and the multi-objective function minimization iteration is performed based on the posture target point position error, direction error and welding speed error. The iterative algorithm adopts the gradient descent method, and the step size is set to 0.05°. The solutions that do not meet the error constraint conditions are discarded, and the solutions that meet the conditions are deduplicated with redundant solutions. Finally, the output contains the joint angle sequence, position error data, direction error data, speed error data and joint angle space limit matching data of each iterative step size adjustment amount.

[0134] The present invention also provides a robot arm posture planning system for executing the robot arm posture planning method described above, the robot arm posture planning system comprising:

[0135] The morphological difference analysis module is used to obtain the product welding structure diagram and the basic setting parameters of the robot arm welding gun; based on the product welding structure diagram, the morphological difference analysis between the welding point sequences is performed to obtain the morphological difference data between the welding point sequences;

[0136] The strength loss posture association mining module is used to simulate and analyze the morphological difference data between the welding points according to the basic setting parameters of the robotic arm welding gun to obtain the welding pool morphological deviation data; based on the welding pool morphological deviation data, the strength loss posture association mining is performed to obtain the strength loss posture association data;

[0137] The robot arm welding gun posture planning module is used to perform joint angle space limit matching based on the strength loss posture correlation feature data to generate joint angle space limit matching data; the robot arm welding gun posture planning and design is performed through the joint angle space limit matching data to obtain the robot arm welding gun posture planning data.

[0138] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A robot arm posture planning method, characterized in that: The following steps are involved: Step S1: Obtain product welding structure diagram and basic setting parameters of robotic arm welding gun; Based on the product welding structure diagram, a morphological difference analysis is performed between welding point sequences, so as to obtain morphological difference data between welding point sequences; wherein the morphological difference data between welding point sequences refers to the point-by-point extraction of the spatial coordinates, welding path direction, and weld thickness parameters of all welding points in the structure diagram based on a geometric feature extraction algorithm of vector analysis, and the point-to-point sequence relationship mapping of all welding points using a sorting topological relationship matrix, and then the angle change analysis in Euclidean space is performed on the spatial vector change between the welding point sequences, and the minimum spatial turning radius between different welding points and the length difference between adjacent points of the weld are vector-scaled and normalized, and the segmented error statistics are performed in combination with the minimum welding spacing threshold and the weld thickness change rate specified in the welding process, and finally, the morphological difference data between the welding point sequences including the welding point position sequence, the angle change matrix, and the length difference vector group are formed through a serialization output module; Step S2: performing a welding pool morphology deviation simulation analysis on the morphology difference data between the welding point sequences according to the basic setting parameters of the robotic arm welding gun to obtain welding pool morphology deviation data; performing strength loss posture association mining based on the welding pool morphology deviation data to obtain strength loss posture association data; wherein the strength loss posture association mining refers to performing strength loss posture association mining based on the deviation data, reconstructing the multi-scale local stress field of the temperature gradient field, quantifying the strength loss index of the thermal stress of each unit grid point using the Von Mises yield criterion, performing three-dimensional vector angle fitting on the welding posture angle change and the local stress loss distribution, performing point-by-point correlation analysis on each posture angle variable and the strength loss amount based on the Pearson correlation coefficient, selecting the posture angle change interval with the absolute value of the correlation coefficient greater than 0.85, and forming the strength loss posture association data; Step S3 is specifically as follows: Step S31: performing feature learning on the intensity loss posture associated data to obtain intensity loss posture associated feature data; Step S32: performing joint angle space limit matching on the welding gun set posture according to the strength loss posture correlation feature data to generate joint angle space limit matching data; wherein the joint angle space limit matching refers to solving the relationship between the joint angular velocity and the end linear velocity, performing joint angle limit detection based on the actual hardware joint limit angle constraint range of the robot arm, identifying posture angle solutions with the risk of angle crossing the limit, and using a nonlinear constraint optimization algorithm to perform angle fallback and re-solve infeasible solutions to ensure that all solutions meet the kinematic constraints of the robot arm; Step S33: performing singular configuration optimization adjustment on the joint angle space limit matching data, thereby obtaining joint angle space limit optimization data; Step S34: performing robot arm welding gun posture planning and design through joint angle spatial limit optimization data to obtain robot arm welding gun posture planning data.

2. The robot arm posture planning method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Obtain product welding structure diagram and basic setting parameters of the robotic arm welding gun; Step S12: vectorizing the product welding structure diagram to generate a product welding structure vector diagram; Step S13: extracting the welding requirement form from the product welding structure vector diagram to obtain product welding requirement form data; Step S14: performing morphological difference analysis between welding point sequences on the product welding requirement morphological data based on the product welding structure vector diagram, thereby obtaining morphological difference data between welding point sequences.

3. The robot arm posture planning method according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: extracting the welding gun setting posture from the basic setting parameters of the robot arm welding gun; Step S22: performing a welding pool morphology deviation simulation analysis on the morphology difference data between the welding point sequences according to the basic setting parameters of the robot arm welding gun and the setting posture of the welding gun to obtain welding pool morphology deviation data; Step S23: performing weld spot strength loss tolerance fitting on the weld pool morphology deviation data to generate weld spot strength loss tolerance data; Step S24: performing strength loss posture association mining on the welding gun setting posture based on the weld point strength loss tolerance data to obtain strength loss posture association data.

4. The robot arm posture planning method according to claim 3, characterized in that: Step S22 includes the following steps: Step S221: Extracting the welding setting voltage, current, and speed from the basic setting parameters of the robotic arm welding gun; calculating the point-to-point vector angle between welding points based on the set welding gun posture, thereby obtaining the vector angle between the welding point and the welding gun; wherein the vector angle between welding points refers to the spatial position vector between adjacent welding points; Step S222: Based on the morphological difference data between the weld points, a vector angle change analysis is performed on the morphological difference data between the weld points to obtain the posture change vector angle data between the weld points and the welding gun; wherein the vector angle change analysis refers to: using the central difference method to calculate the vector angle change rate of the angle data of adjacent weld points, and during the calculation process, the angle difference of each pair of consecutive weld points is divided by the actual spatial distance between the two points to obtain the angle change rate per unit length, and then the angle change rate of the entire weld point sequence is statistically analyzed to extract the maximum angle change rate. , minimum angle change rate and average angle change rate, and standardize the statistical results to make the numerical range normalized to between 0 and 1. Use the change trend fitting method based on piecewise linear regression to fit the angle change rate curve in sections, identify the inflection points of the fitting curve, and use the first-order derivative mutation point detection method to determine the trend turning point position of the angle change interval, mark the trend inflection point in sections, and finally output the posture change vector angle data including the welding point number, point-to-point angle change rate, trend inflection point position, and piecewise fitting slope value; Step S223: performing simulation and analysis of the instantaneous temperature field of the weld pool during the multi-frequency welding process on the posture change vector angle data according to the welding set voltage, current, and speed to obtain the instantaneous temperature field of the weld pool during the multi-frequency welding process; wherein the simulation and analysis of the instantaneous temperature field of the weld pool during the multi-frequency welding process refers to discretizing the welding path within each angle change section by linear interpolation, performing point-by-point heat input simulation on the discrete path points, and numerically solving the temperature field control equation using the finite volume method, with the boundary condition set to a natural convection boundary and the initial temperature set to a room temperature of 25°C. An implicit time integration method is used in the heat conduction solution process; Step S224: performing an anisotropic fluid solidification behavior imbalance analysis on the instantaneous temperature field of the multi-frequency weld pool, thereby obtaining anisotropic fluid solidification behavior imbalance data; wherein the anisotropic fluid solidification behavior imbalance analysis refers to performing a spatial second-order derivative analysis on the temperature gradient change rate to extract the local heat flow direction change rate, combining the latent heat release interval data of the metal liquid and solid transition temperature zones in the material thermophysical parameter table, setting the liquid-solid transition critical temperature range to 1400°C to 1450°C, performing time-series energy flux integration on the unit nodes in the temperature zone, using the cumulative energy method to compare the heat flux density vectors in different directions, performing a differential method to numerically calculate the solidification interface movement speed along different spatial directions, extracting the interface front velocity vector field, performing a vector variance analysis on the degree of solidification velocity inhomogeneity in different directions based on the velocity vector field distribution, performing a numerical correlation coefficient analysis on the correlation between the velocity variance distribution and the heat flux density directional change using the anisotropic fluid dynamics criterion, and performing clustering processing on regions with high solidification velocity differences and heat flux density gradient direction deviations; Step S225: performing interfacial tension vibration regression analysis on the imbalance data of solidification behavior of the anisotropic fluid to obtain interfacial tension vibration regression data; wherein the interfacial tension vibration regression analysis refers to extracting the solidification front velocity time series data and the interface temperature gradient change series data in each imbalance characteristic region, first performing a fast Fourier transform on the solidification front velocity time series, performing spectrum extraction on the main vibration frequency, and setting the frequency analysis range to 0 to 200 Hz, then using a wavelet transform to analyze the local temporal characteristics of the vibration frequency, selecting a Morlet wavelet basis function for multi-scale decomposition, fitting the frequency amplitude change trends at different time scales, using the least squares method to perform polynomial regression on the frequency amplitude curves at each scale, using a sliding average filter method to remove high-frequency noise interference for the interface temperature gradient change series, performing first-order difference processing on the filtered gradient change series to extract the change rate, performing a Pearson correlation analysis on the interface vibration frequency data and the gradient change rate data, and calculating the correlation coefficient to determine the degree of dependence of the vibration frequency on the temperature gradient change; Step S226: performing a welding pool morphology deviation simulation analysis on the morphology difference data between welding spot sequences according to the anisotropic fluid solidification behavior imbalance data and the interfacial tension vibration regression data to obtain welding pool morphology deviation data.

5. The robot arm posture planning method according to claim 4, characterized in that: Step S224 includes the following steps: The heat flux density non-uniformity vector is identified for the instantaneous temperature field of the multi-frequency weld pool to obtain the heat flux density non-uniformity vector. The heat flux density non-uniformity vector identification includes: numerically calculating the temperature gradient of each grid cell in the temperature field, discretizing the temperature gradient of each grid cell in three-dimensional space by the first-order partial derivative using the central difference scheme, and performing point-by-point vector calculation of the heat flux density of each grid cell in the X, Y, and Z directions according to Fourier's heat conduction law and the material thermal conductivity parameter, with the material thermal conductivity value set to 45 W / (m·K). Subsequently, a statistical variance method is used to perform spatial variance analysis on the heat flux density vectors of each grid cell in the entire temperature field, and marking grid cells whose spatial variance exceeds the average level by more than 1.5 times as a heat flux density non-uniformity area. Based on the heat flux density non-uniformity vector, differential coupling of the time series change of the solder joint surface tension is performed to generate differential coupling data of the time series change of the tension; wherein the differential coupling of the time series change of the solder joint surface tension refers to extracting the time series heat flux density change data of each non-uniform heat flux area, using the first-order time difference method to perform differential operation on the heat flux density time series, performing sliding average filtering on the differentiated data, and setting the filter window width to 5ms. The filtered heat flux density differential sequence and the surface temperature gradient change sequence in the corresponding area are synchronously normalized, and then based on the surface tension gradient equation, the normalized heat flux density differential sequence is linearly weighted, and the weight coefficient is set according to the surface tension temperature coefficient of the metal material, and the setting range is -0.4 to -0.6mN / (m・℃). The tension gradient change within the time step is accumulated and summed by the numerical integration method; performing local variance calculation of the molten metal viscosity based on differential coupling data of the heat flux density heterogeneity vector and the tension time series variation to obtain the local variance of the molten metal viscosity; wherein the local variance calculation of the molten metal viscosity refers to performing bivariate correlation modeling based on the heat flux density vector directional change rate and the tension change rate, performing linear fitting of the local heat flux change rate and the tension change rate using the least squares method, using the fitted residual variance as an indicator of local viscosity variation heterogeneity, performing statistical processing on the residual variance of all heterogeneous area units, calculating the viscosity variation fluctuation coefficient in the local area using the coefficient of variation method, marking the area with a fluctuation coefficient exceeding 0.2 as a viscosity fluctuation area, and performing time series variance calculation on the viscosity change rate in the fluctuation area; Based on the heat flux density inhomogeneity vector, the tension time series variation differential coupling data and the local variance of the molten metal viscosity, the solidification latent heat heat energy release ratio difference between the different solder joint orientations is deduced to generate the heat energy release ratio difference between the different solder joint orientations; wherein the deduction of the solidification latent heat heat energy release ratio difference between the different solder joint orientations refers to: reading the heat flux density vector component of each grid unit in the heat flux density inhomogeneity vector data, combining the unit tension change rate within the corresponding time step in the tension time series variation differential coupling data and the viscosity change variance value of each unit in the molten metal viscosity local variance data, performing intra-unit aggregation processing on the data under the same solder joint orientation, and using a multivariate weighted integration method to calculate the local heat flux density vector modulus and tension change within a single solder joint orientation. The energy conversion coefficient is calculated based on the rate and viscosity variance. The heat flux weight factor is set to 0.5, the tension change rate weight factor is set to 0.3, and the viscosity variance weight factor is set to 0.

2. The polymerization energy conversion coefficient in different solder joint orientations is standardized, and the normalization interval is set to 0 to 1. The normalized energy conversion coefficient is multiplied by the theoretical value of the solidification latent heat of the corresponding orientation. The theoretical value of the latent heat is set to 270 kJ / kg. The solidification latent heat release per unit volume of solder joints in different orientations is accumulated and integrated. The step integration method with a time step of 1 ms is used to ratio the cumulative latent heat release of all solder joint orientation units. The orientation with the largest latent heat release is selected as the reference orientation, and the latent heat release of other orientations is normalized and ratioed. The imbalance analysis of solidification behavior of anisotropic fluid is performed based on the difference in heat energy release ratios between different solder joint orientations, thereby obtaining the imbalance data of solidification behavior of anisotropic fluid.

6. The robot arm posture planning method according to claim 3, characterized in that: Step S23 includes the following steps: Step S231: performing a meshing process on the weld pool morphology deviation data to obtain meshing data of the weld pool morphology deviation; wherein the meshing process of the weld pool morphology deviation is specifically as follows: first, uniformly voxelizing the entire weld pool region according to a three-dimensional grid coordinate system is performed using a voxel space meshing technique based on a region growing method, with the voxel size set to 0.1 mm³; performing a three-dimensional gradient calculation on the weld pool morphology deviation value within the voxel unit; using a threshold value of a spatial gradient modulus greater than twice the average gradient modulus as a criterion for determining an abnormal deviation area; and determining connected regions of abnormal deviation areas of adjacent voxel units using a depth-first search algorithm; Step S232: performing a melt pool geometric deviation feature analysis on the melt pool morphology deviation grid division data, thereby obtaining melt pool geometric deviation feature data; wherein the melt pool geometric deviation feature analysis comprises: calculating ellipsoid parameters for the spatial distribution of each deviation grid sub-region using a three-dimensional minimum circumscribed ellipsoid fitting algorithm, extracting the angle between the major axis, minor axis, and main axis direction cosine and the spatial main direction vector, calculating the three-axis spatial discreteness index of the deviation grid region, marking the deviation sub-region with a major axis length greater than 3 mm and a major-minor axis ratio greater than 2 as a high anisotropy deviation region, and performing a third-order statistical moment analysis on the deviation intensity distribution within the deviation region, calculating the deviation kurtosis and skew coefficient, and performing high peak state identification on the region with a deviation kurtosis greater than 3; Step S233: simulating the weld point collision force offset distribution based on the molten pool geometric deviation characteristic data to generate collision force offset distribution data; Step S234: performing stress distribution strain rate component decomposition on the collision force offset distribution data to obtain strain rate component decomposition data; wherein the stress distribution strain rate component decomposition on the collision force offset distribution data includes: reading the force vector and displacement vector of each node, performing differential derivative calculation on the force-displacement relationship between the nodes to obtain the strain rate tensor of each node, performing eigenvalue decomposition on the strain rate tensor, extracting the principal strain rate components, performing spatial direction projection on the principal strain rate components, decomposing the strain rate components into radial components, tangential components, and axial components using a spherical coordinate system, and normalizing each component with the normalization interval set to -1 to 1, calculating the Pearson correlation coefficient of the coupling relationship between the components, and determining the coupling strength between components in different directions; Step S235 : performing weld strength loss tolerance fitting based on the strain rate component decomposition data to generate weld strength loss tolerance data.

7. The robot arm posture planning method according to claim 1, characterized in that: Step S32 includes the following steps: Step S321: performing joint verticality matching on the set posture of the welding gun according to the strength loss posture association feature data to obtain joint verticality matching data; wherein the joint verticality matching comprises: extracting the posture offset angle feature and the spatial direction vector feature in the strength loss posture association feature data, extracting the posture normal vector for each set posture using the normal vector angle projection method, calculating the angle between the joint end normal vector and the product welding surface normal vector, calculating the angle using the spatial vector dot product angle formula, setting the joint verticality deviation tolerance to plus or minus 3°, screening all posture angle sequences with angles less than the tolerance, and performing joint limit angle interval verification on each qualified posture angle sequence to eliminate posture sequences with joint angles exceeding the limit; Step S322: performing adaptive adjustment control of the thrust angle / drag angle of the robot arm welding gun according to the joint verticality matching data and the strength loss posture correlation feature data to obtain thrust angle / drag angle adaptive control data; Step S323: performing feed welding speed matching on the push angle / drag angle adaptive control data, thereby obtaining feed welding speed matching data; Step S324: performing joint angle space limit matching on the welding gun setting posture based on the joint verticality matching data, the push angle / drag angle adaptive control data and the feed welding speed matching data to generate joint angle space limit matching data.

8. A robotic arm posture planning system, characterized in that: For executing the robot arm posture planning method according to claim 1, the robot arm posture planning system comprises: The morphological difference analysis module is used to obtain the product welding structure diagram and the basic setting parameters of the robot arm welding gun; based on the product welding structure diagram, the morphological difference analysis between the welding point sequences is performed to obtain the morphological difference data between the welding point sequences; The strength loss posture association mining module is used to simulate and analyze the morphological difference data between the welding points according to the basic setting parameters of the robotic arm welding gun to obtain the welding pool morphological deviation data; based on the welding pool morphological deviation data, the strength loss posture association mining is performed to obtain the strength loss posture association data; The robot arm welding gun posture planning module is used to perform joint angle space limit matching based on the strength loss posture correlation feature data to generate joint angle space limit matching data; the robot arm welding gun posture planning and design is performed through the joint angle space limit matching data to obtain the robot arm welding gun posture planning data.

Images