CN-122023511-A - Welding point location method and system based on machine vision

CN122023511ACN 122023511 ACN122023511 ACN 122023511ACN-122023511-A

Abstract

The invention discloses a welding point location method and a system based on machine vision, and belongs to the technical field of welding. The method comprises the steps of constructing working condition sample sets under different interference states by establishing an interference factor sample library of a workpiece sample library and a multi-type interference factor combination, dynamically evaluating the attitude error coefficients of all components of the welding robot arm based on a Z-Score model, judging the normality of positioning parameters, quantifying the relevance of different interference states through a working condition suitability matrix, constructing attitude feature points, matching the current welding scene interference states, screening optimal positioning attitude parameter sample clusters and guiding positioning. The invention solves the problems that the traditional welding positioning ignores the influence of multi-interference combination, the attitude error evaluation is inaccurate, the optimal parameter screening lacks quantitative basis and the like, realizes the high-precision point location positioning under the complex interference working condition, improves the anti-interference capability and stability of the welding positioning, is suitable for various welding scenes such as automobile parts, steel structures and the like, and provides technical support for automatic welding quality control.

Inventors

ZHAO DE
HOU XINLI
QIU HANCHENG
Zhang Lingcai
ZHANG YUEPENG
ZHOU JIANFEI

Assignees

南京绥德自动焊接装备有限公司

Dates

Publication Date: 20260512
Application Date: 20251212

Claims (10)

1. The welding point location method based on machine vision is characterized by comprising the following steps of: Step S1, establishing a relevant sample library of interference factors of a workpiece and a welding process, wherein the relevant sample library comprises a workpiece sample library for recording a contour to be welded and an interference factor sample library for recording the influence state of the interference factors, and constructing a working condition sample set under the influence state of different interference factors aiming at the positioning posture parameters of a welding robot arm initialized by the contour to be welded; s2, based on a Z-Score model, evaluating the attitude error coefficients of all components of the welding robot arm, and judging the normality of the positioning attitude parameters of all the components; S3, constructing a working condition suitability matrix based on the positioning attitude parameter normality judgment result, and evaluating interference correlation under different interference states; And S4, constructing attitude feature points of the welding robot arm, matching the interference factor influence state of the current welding scene, screening out an optimal positioning attitude parameter sample cluster, and guiding the positioning operation of the point positions of the welding robot arm.
2. The machine vision-based welding point location method according to claim 1, wherein the specific implementation process of step S1 includes: Establishing a workpiece sample library, wherein the workpiece sample library records the contours to be welded of the workpieces, and marking any ith contour to be welded as The contours to be welded are identified through a machine vision technology; establishing an interference factor sample library in the welding process, and initializing and adapting to the contour to be welded under the influence of the interference factor The interference factor influence state is a single type interference factor or a combination of at least two types of interference factors, and the interference factor influence state is recorded into the interference factor sample library; the influence state of the x-th interference factor is recorded as , wherein, Represents the E-th interference factor, E represents the total number of interference factors, and the state is affected by the interference factors Next, the outline to be welded is adapted in an initialized manner Is marked as the y-th positioning attitude parameter sample cluster , wherein, The positioning attitude parameter of the R-th component of the welding robot arm is represented, and R represents the total number of components of the welding robot arm; constructed in the influence state of interference factors Lower initialization adaptation to the profile to be welded Is recorded as a working condition sample set Y represents the state of influence of the disturbance factor Lower initialization adaptation to the profile to be welded The total number of the positioning attitude parameter sample clusters.
3. The machine vision-based welding point location method according to claim 2, wherein the specific implementation process of step S2 includes: Based on the Z-Score model, the state of influence of interference factors on the same component of the welding robot arm is evaluated Lower component attitude error coefficient In which, in the process, Is the average value of the positioning gesture parameters of the same component under different positioning gesture parameter sample clusters, and , Is the standard deviation of the positioning posture parameters of the same component under different positioning posture parameter sample clusters, and , Representing a sample cluster derived from a localization attitude parameter Is a positioning attitude parameter; Presetting a Z-Score threshold of a component attitude error coefficient, if the component attitude error coefficient is Greater than or equal to the Z-Score threshold, then determining that the sample cluster is derived from the positioning attitude parameter Is a positioning attitude parameter of (a) If the component attitude error coefficient is abnormal If the position and orientation parameter sample cluster is smaller than the Z-Score threshold, determining that the position and orientation parameter sample cluster is sourced Is a positioning attitude parameter of (a) Normal.
4. The machine vision based welding point location method according to claim 3, wherein the implementation process of step S3 includes: based on the working condition sample set, constructing a working condition suitability matrix with a row index as a positioning posture parameter sample cluster serial number and a column index as a welding robot arm assembly serial number, if the working condition suitability matrix is from the positioning posture parameter sample cluster Is a positioning attitude parameter of (a) If the judgment is abnormal, recording a numerical value 1 at the matrix position of the y-th row and the r-th column in the working condition suitability matrix, and if the numerical value is derived from a positioning attitude parameter sample cluster Is a positioning attitude parameter of (a) If the judgment is normal, recording a numerical value 0 at the matrix position of the y-th row and the r-th column in the working condition suitability matrix; Will influence the state in the presence of interference factors The lower part is to be welded with the outline The corresponding working condition suitability matrix generated during the initialization adaptation is recorded as ; Evaluating contours to be welded Interference correlation in different interference factor influence states In which, in the process, Representing the state of influence of the f-th interference factor, and f +.x, Is expressed in the influence state of interference factors Lower initialization adaptation to the profile to be welded Is a sample set of the working conditions of (a), Is expressed in the influence state of interference factors The lower part is to be welded with the outline The corresponding working condition suitability matrix is generated when initialization adaptation is carried out, F represents the influence state of interference factors Lower initialization adaptation to the profile to be welded Is a total number of the set of positioning pose parameter sample clusters, For summing functions for matrix The sum is carried out on the values of all elements in the matrix, and T is the transposed symbol of the matrix.
5. The machine vision based welding point location method according to claim 4, wherein the implementation process of step S4 includes: based on interference correlation, constructing a welding robot arm to execute a contour to be welded During welding, the attitude characteristic points under the influence of different interference factors Joint domain combining with abscissa independent variables as interference factor influence states The ordinate dependent variable is the interference correlation Will be As a first parameter of the joint definition field, will As a second parameter of the joint definition field, and the first parameter and the second parameter are satisfied after exchanging positions I.e. ; Acquiring a currently executed contour to be welded To generate a state to be matched of the interference factor And carrying out matching screening of the joint definition domain: Definition of variables And (2) and As either the first parameter or the second parameter, , A sample library for interference factors; Evaluating a variable To be matched with the state to be matched Degree of matching between If (if) Then screen the variables If (if) Then the variables are not filtered In which, in the process, Representing variables To be matched with the state to be matched The number of types of interference factors included in the result of the inter-intersection operation, Representing variables To be matched with the state to be matched The number of types of interference factors contained in the result of the union operation, Sample library for representing interference factors The interference factors contained in the system affect the total number of states; combining the selected variables pairwise to obtain a combined definition domain combination screening set And selecting the joint definition domain combination when the interference correlation degree is maximum In which, in the process, The method comprises the steps of selecting a joint definition domain combination for maximum interference correlation as a feedback indication function; Selected joint domain combinations based For the same component of the welding robot arm, comparing the influence states of interference factors Component attitude error coefficient under condition and influence state under interference factor The size of the lower component attitude error coefficient is calibrated, and the positioning attitude parameter corresponding to the minimum component attitude error coefficient is used as the optimal positioning attitude parameter of the component; after the calibration of the optimal positioning attitude parameters of each component of the welding robot arm is completed, the profile to be welded which is currently executed is formed The optimal positioning attitude parameter sample cluster of the welding robot arm is used for guiding the welding robot arm to execute the contour to be welded And positioning the point position during welding.
6. The welding point location system based on machine vision for executing the welding point location method based on machine vision according to any one of claims 1 to 5, wherein the system comprises a sample library construction module, an attitude error evaluation module, a working condition suitability analysis module and an optimal location parameter screening module; The sample library construction module is used for constructing a workpiece sample library and an interference factor sample library and constructing working condition sample sets under different interference factor influence states; The attitude error evaluation module is used for evaluating the attitude error coefficients of all components of the welding robot arm based on the Z-Score model and judging the normality of positioning attitude parameters; the working condition suitability analysis module is used for constructing a working condition suitability matrix based on the normal judgment result of the positioning attitude parameters and evaluating the interference correlation degree under the condition that different interference factors influence the state; the optimal positioning parameter screening module is used for constructing gesture feature points of the welding robot arm, matching the current interference factor influence state, screening an optimal positioning gesture parameter sample cluster and guiding point location positioning operation.
7. The machine vision-based weld spot location system of claim 6, wherein the sample library construction module comprises a workpiece sample library construction unit, an interference factor sample library construction unit, and a working condition sample set construction unit; The workpiece sample library establishing unit is used for identifying the outline of the workpiece to be welded through a machine vision technology, and establishing and storing outline information to be welded; The interference factor sample library establishing unit is used for recording interference factors of different types and interference factor influence states formed by combination, initializing positioning gesture parameters of the welding robot arm which are matched with the profile to be welded, and forming a positioning gesture parameter sample cluster; The working condition sample set construction unit is used for combining the profile to be welded, the influence state of the interference factors and the corresponding positioning posture parameter sample clusters to construct a corresponding working condition sample set.
8. The machine vision based weld spot positioning system of claim 6, wherein the posing error assessment module comprises a parameter mean calculation unit, a parameter standard deviation calculation unit, and a posing parameter normality determination unit; The parameter average value calculation unit is used for calculating the positioning attitude parameter average value of the same welding robot arm assembly under different positioning attitude parameter sample clusters; The parameter standard deviation calculation unit is used for calculating the standard deviation of the positioning posture parameters of the same welding robot arm assembly under different positioning posture parameter sample clusters; The attitude parameter normality judging unit is used for evaluating the attitude error coefficient of the component based on the Z-Score model and the calculated mean value and standard deviation and judging whether the positioning attitude parameter is normal or not.
9. The machine vision-based welding point location system of claim 6, wherein the condition suitability analysis module comprises a condition suitability matrix construction unit and an interference correlation evaluation unit; the working condition suitability matrix construction unit is used for constructing a working condition suitability matrix according to the normal judgment result of the positioning attitude parameters by taking the positioning attitude parameter sample cluster serial number and the welding robot arm assembly serial number as indexes; the interference correlation evaluation unit evaluates the interference correlation of the profile to be welded under the influence states of different interference factors based on the working condition suitability matrix.
10. The welding point location system based on machine vision according to claim 6, wherein the optimal location parameter screening module comprises a gesture feature point construction unit, an interference state matching unit and an optimal parameter calibration unit; The attitude feature point construction unit is used for constructing attitude feature points of the welding robot arm based on different interference factor influence states and corresponding interference correlation degrees; the interference state matching unit is used for acquiring interference factors of the current welding scene, generating an interference state to be matched, and screening the interference factor influence state matched with the state to be matched; The optimal parameter calibration unit is used for comparing the attitude error coefficients of the components in the screened interference factor influence state, calibrating the optimal positioning attitude parameters of each component, forming an optimal positioning attitude parameter sample cluster and guiding the positioning of the point positions of the welding robot arm.

Description

Welding point location method and system based on machine vision Technical Field The invention relates to the technical field of welding, in particular to a welding point location method and a system based on machine vision. Background In the field of automatic welding, the positioning accuracy of welding points directly determines the quality of welding seams, the assembly accuracy of workpieces and the production efficiency, and is one of core management and control links in industrial manufacturing. Along with the progress of the welding technology to refinement and scale, the requirements on positioning precision, anti-interference capability and suitability are increasingly improved, but the conventional welding point positioning technology still has a plurality of key problems, and the high-quality development of automatic welding is severely restricted. The traditional welding positioning method depends on preset fixed parameters or single sensor data, and the complexity of interference factors in the welding process is not fully considered. In a welding scene, interference factors such as splashing, arc light and the like often exist in a combined form, and the prior art is mainly modeled only aiming at a single interference factor, so that the comprehensive influence of multi-interference coupling on positioning attitude parameters is ignored, the suitability of a positioning model is poor, and positioning deviation is easy to occur under a complex working condition. The existing attitude error assessment mostly adopts fixed threshold judgment, dynamic differences of parameter distribution under different components and different interference states are not considered, the subjectivity of threshold setting is strong, normal fluctuation and abnormal deviation of parameters are difficult to accurately distinguish, the error assessment precision is low, and reliable basis cannot be provided for positioning parameter optimization. Meanwhile, the traditional method lacks quantitative analysis on the relevance of different interference states, and when the working conditions change, the adaptive positioning parameters cannot be matched quickly, and the debugging is needed again, so that the production efficiency is reduced. In the prior art, manual experience or simple comparison is relied on, a quantitative mapping relation between an interference state and positioning parameters is not established, the screened parameters are difficult to adapt to actual requirements of current working conditions, and particularly, the problems of insufficient positioning precision, unstable welding quality and the like are easy to occur in a multi-interference combined scene. In addition, the traditional system does not form closed loop logic of 'sample construction-error evaluation-working condition adaptation-parameter screening', hysteresis exists in data processing and parameter adjustment, and when the interference state is suddenly changed, the system is difficult to respond in time, so that a large number of unqualified products are generated, and the production cost is increased. The problems cause the defects of weak anti-interference capability, low positioning precision, poor suitability, delayed response and the like of the existing welding point location technology, and cannot meet the refined requirements of modern industry on automatic welding under complex working conditions. Therefore, developing a welding point location method and system capable of covering multi-interference combination, accurate error evaluation, quantitative parameter screening and efficient and stable location becomes an urgent need in the current automatic welding field. Disclosure of Invention The invention aims to provide a welding point location method and a system based on machine vision, which are used for solving the problems in the background technology. In order to solve the technical problems, the invention provides the following technical scheme: The welding point location system based on machine vision comprises a sample library construction module, an attitude error evaluation module, a working condition suitability analysis module and an optimal location parameter screening module; The sample library construction module is used for constructing a workpiece sample library and an interference factor sample library and constructing working condition sample sets under different interference factor influence states; The attitude error evaluation module is used for evaluating the attitude error coefficients of all components of the welding robot arm based on the Z-Score model and judging the normality of positioning attitude parameters; the working condition suitability analysis module is used for constructing a working condition suitability matrix based on the normal judgment result of the positioning attitude parameters and evaluating the interference correlation degree under the condition that different interference fa