Whole welding process quality regulation and control method of welding robot

Abstract

The invention discloses a whole welding process quality regulation and control method of a welding robot, and relates to the technical field related to welding. The method at least comprises the steps that S1, a pre-welding parameter drawing technology is designed, welding numerical simulation analysis is conducted to predict optimal welding parameters, the optimal welding parameters are compared with a thermal cycle curve of a test, the accuracy of a model is verified, the predicted parameters are input into a welding robot, and the welding prediction effect is achieved; and S2, designing a welding molten pool regulation and control technology, collecting a molten pool image through an industrial camera, processing the collected molten pool image, then building a molten pool monitoring platform, and regulating and controlling the quality of the molten pool during welding through the platform. According to the method, a proper welding quality guarantee strategy of the welding robot is provided, and the welding quality of the welding robot is guaranteed in the whole process of pre-welding, during-welding and after-welding through the cooperative design of pre-welding parameter drawing-up based on the numerical simulation technology, during-welding molten pool monitoring based on camera recognition and after-welding quality study and judgment based on deep learning.

Description

Technical Field

[0001] The invention relates to the technical field related to welding, and in particular to a whole-process quality control method for a welding robot. Background Art

[0002] As an indispensable manufacturing process in the engineering field, welding plays a key role in various fields such as bridges, buildings, ships, aerospace, etc.

[0003] In recent years, with the intelligent development of my country's welding technology, traditional manual welding operations have gradually been replaced by intelligent welding robots.

[0004] While welding robots improve welding efficiency and reduce labor costs, how to ensure their welding quality has also become a key issue of concern.

[0005] Quality problems and risks in the welding process may have a serious impact on engineering projects, not only resulting in additional costs, but also jeopardizing the safety and reliability of the project.

[0006] The welding robot consists of a control module, an execution module, and a vision module to achieve welding automation.

[0007] However, before welding, no clear parameters were set for the required welding effect, and the welding results could not be predicted; during welding, the molten pool was not monitored and information processed, and the molten pool quality could not be controlled; after welding, the welding quality was not tested, and an alarm could not be issued for defective welds to remind workers to perform post-weld processing.

[0008] Therefore, in order to ensure the welding quality of the welding robot, it is urgent to design a quality control method for the entire welding process of the welding robot. Summary of the invention

[0009] The purpose of the present invention is to provide a whole-process quality control method for a welding robot, so as to ensure the welding quality of the welding robot throughout the whole process and solve the technical problems raised in the background technology.

[0010] To achieve the above object, the present invention provides the following technical solution: a method for controlling the quality of the entire welding process of a welding robot, comprising at least the following steps:

[0011] S1: Design pre-welding parameter formulation technology, perform welding numerical simulation analysis to predict the optimal welding parameters, and compare them with the experimental thermal cycle curve to verify the accuracy of the model. Input the predicted parameters into the welding robot to achieve welding prediction effect;

[0012] S2: Design the molten pool control technology during welding, collect the molten pool image through the industrial camera, process the collected molten pool image, and then build a molten pool monitoring platform to monitor the molten pool information in real time, and use the platform to control the molten pool quality during welding;

[0013] S3: Design post-weld quality assessment technology, collect weld surface defect data sets, and train through deep learning. When the welding robot detects weld surface defects, it can record the location and type of the defects and issue an alarm. The surface defects include at least cracks, pores and weld nodules.

[0014] Furthermore, the S1 at least comprises the following steps:

[0015] S1.1: The thermal cycle curve of the welded plate in the test is measured by a thermocouple sensor. This method directly contacts the welded parts, and the measurement point recording is more convenient, and the measurement results are more accurate;

[0016] S1.2: Carry out numerical simulation analysis based on welding temperature field theory;

[0017] S1.3: Based on the analysis results of the numerical simulation to predict the welding thermal behavior, the welding process is optimized to achieve better welding results. Finally, the proposed optimal welding parameters are input into the welding robot.

[0018] Furthermore, the S1.1 at least includes the following steps:

[0019] First, before the test, select appropriate measuring points on the plate to be welded and clean them to prevent the surface defects of the specimen from interfering with the test results;

[0020] Then, a welding test is carried out, and the temperature change of the measuring point during welding is recorded with a thermocouple sensor, and the temperature measurement is terminated when the welded parts cool to room temperature;

[0021] Finally, the test data were integrated to draw the thermal cycle curve during the welding test, and the cross-sectional morphology of the welding pool was measured by image processing method.

[0022] Furthermore, the S1.2 at least includes the following steps:

[0023] First, a proportional model of the welded part is established based on the test;

[0024] Secondly, pre-processing settings related to the test conditions are performed, wherein the pre-processing settings at least include material properties, boundary conditions and ambient temperature, and restore the welding environment of the test;

[0025] Then, a welding subroutine is written using Fortran language, wherein the welding subroutine at least includes a heat source model, a welding path and welding parameters, and a thermodynamic simulation is performed;

[0026] Finally, the thermal cycle curves of the corresponding calculation nodes at the positions corresponding to the test are extracted and fitted with the thermal cycle curves measured in the test to verify the correctness of the simulation model.

[0027] Furthermore, the S2 at least includes the following steps:

[0028] S2.1: Monitor the liquid flow of the molten pool during welding through an industrial camera and process the collected images;

[0029] S2.2: Obtain accurate molten pool information, calculate values ​​according to specific conditions, and ensure welding stability;

[0030] S2.3: Develop an intelligent welding interface to read the above molten pool information and monitor the molten pool conditions during welding in real time. The size of the molten pool, the overlap between the center of the molten pool and the center line of the weld, and the fusion of the parent material are all important monitoring objects. When the molten pool size is not good during welding, the overlap between the center of the molten pool and the center line of the weld is too large, and the fusion of the parent material is poor, the monitoring system can make a judgment on the molten pool quality, and then send feedback adjustment instructions to the welding robot control terminal through the system to adjust the trajectory and welding parameters of the welding robot. The welding parameters include at least speed, current and voltage, thereby regulating the molten pool quality.

[0031] Furthermore, the S2.1 at least includes the following steps:

[0032] The camera is installed on the welding gun so that the camera moves synchronously with the welding gun to extract the molten pool information in real time;

[0033] In order to avoid interference caused by strong arc light during welding, it is also necessary to install a neutral density filter and a filter on the camera lens;

[0034] The neutral density filter adopts an OD neutral density film;

[0035] The filter is a 650nm filter, which can effectively reduce the impact of arc light on image acquisition during welding.

[0036] Furthermore, the S2.2 at least includes the following steps:

[0037] Firstly, the image processing algorithm is used to reduce the noise of the collected image and enhance the melt pool information to be collected;

[0038] Secondly, the edge extraction based on directional wavelet and the method based on active contour model are used to extract the edge features of the molten pool;

[0039] Finally, the molten pool size is calculated, the overlap between the molten pool center and the weld centerline and the fusion of the parent material are monitored to ensure the stability of the welding process.

[0040] Furthermore, the S3 at least includes the following steps:

[0041] S3.1: Detect weld surface defects based on deep learning methods and judge welding quality;

[0042] S3.2: Compare and use different target detection methods to select the best model for weld surface defect detection, then optimize the model to improve the detection accuracy of the model, and finally carry out sufficient model training times to detect weld surface defects and judge the weld quality, so that the welding robot has the function of weld surface defect detection. When it detects surface defects, it can use the defect detection function of the welding robot to record the location and type of the defect in the monitoring system, and issue an alarm to remind workers to carry out post-weld grinding, repair welding, etc. in time to ensure welding quality.

[0043] Furthermore, the S3.1 at least includes the following steps:

[0044] Firstly, common weld surface defect images are collected and a weld surface defect detection dataset is created;

[0045] Then, in order to improve the generalization ability of the model and the detection effect of the model, the data set is subjected to data enhancement operations, wherein the data enhancement operations at least include flipping, rotating, shifting, cropping, deforming and scaling;

[0046] Finally, build a deep learning training model.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention proposes a suitable welding quality assurance strategy for welding robots. Through the coordinated design of pre-welding parameter formulation based on numerical simulation technology, molten pool monitoring during welding based on camera recognition, and post-welding quality assessment based on deep learning, the welding quality of the welding robot is guaranteed throughout the entire process before, during and after welding. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0050] Figure 1 This is a schematic diagram for introducing the welding robot module of the present invention;

[0051] Figure 2 It is a schematic diagram of the technical solution of the present invention;

[0052] Figure 3This is a schematic diagram of the welding experiment of the present invention;

[0053] Figure 4 It is a numerical simulation schematic diagram of the present invention;

[0054] Figure 5 This is a schematic diagram of the molten pool of the present invention;

[0055] Figure 6 It is a schematic diagram of the molten pool image processing of the present invention;

[0056] Figure 7 This is a schematic diagram of the molten pool quality control of the welding robot of the present invention;

[0057] Figure 8 A schematic diagram for making a weld surface defect data set of the present invention;

[0058] Fig. 9 It is a schematic diagram of weld surface defect detection according to the present invention. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0060] The welding quality of the welding robot is very critical. The present invention is based on the following technical solutions to achieve quality control of the welding robot throughout the welding process. Mainly: pre-welding parameter formulation based on numerical simulation technology, molten pool monitoring during welding based on camera recognition, and post-welding quality assessment based on deep learning. Among them, in terms of pre-welding parameter formulation, the optimal welding parameters such as current and voltage are input into the welding robot through the welding machine; in terms of molten pool monitoring during welding, the molten pool quality during welding is monitored and the molten pool quality is controlled through the welding robot vision module and the monitoring system is built, and the welding trajectory and welding speed are controlled through the welding robot execution module; in terms of post-welding quality assessment, the weld surface defects are detected through the welding robot vision module, and the welding robot alarm module is used to remind workers to perform post-welding grinding and repair welding on defective welds, such as Figure 1 .

[0061] Among them, the technical solutions for pre-welding parameter formulation, welding pool monitoring and post-welding quality assessment are as follows: Figure 2 .

[0062] 1. In terms of pre-welding parameter formulation, welding process parameters, such as welding speed, current, voltage, etc., affect the state of the molten pool during welding, and thus affect the quality of welding. In order to explore the intrinsic relationship between welding parameters and molten pool morphology, the present invention uses numerical simulation technology to analyze and predict the optimal welding process parameters, formulates the welding parameters before welding, and inputs the formulated welding parameters into the welding robot before welding to achieve the prediction results.

[0063] 2. In terms of molten pool control during welding, molten pool information, such as the degree of overlap between the midpoint of the molten pool and the center line of the weld, the size of the molten pool, and the flow pattern, can more intuitively reflect the stability of the welding process and the fusion of the parent material. The present invention monitors the molten pool characteristics during welding based on visual sensing technology, and builds a visual monitoring system linked to the welding robot control terminal. The system is embedded with two panels. The first panel can monitor the molten pool in real time and process the molten pool information. When the system detects that the molten pool is unqualified, the system can be used to control the welding parameters, operating speed, and trajectory planning of the welding robot. The second panel is used for post-welding surface defect detection.

[0064] 3. In terms of post-weld quality assessment, surface defects of welded parts, such as common defects such as pores, weld bumps and cracks, not only affect the aesthetics of welded parts, but may also reduce the strength and durability of welded parts, and even cause welded parts to break or fail during use. The present invention is based on deep learning technology, constructs and trains a neural network model, and detects weld surface defects. When weld surface defects are detected, the location and type of weld surface defects can be recorded in the second section of the above-mentioned monitoring system, and an alarm is issued to remind workers to deal with bad welds in a timely manner.

[0065] The present invention proposes a suitable welding quality assurance strategy for the welding robot through the above technical means, which ensures the welding quality of the welding robot from the whole process before, during and after welding.

[0066] The details are as follows:

[0067] A method for controlling the quality of the entire welding process of a welding robot comprises at least the following steps:

[0068] S1: Design pre-welding parameter formulation technology, perform welding numerical simulation analysis to predict the optimal welding parameters, and compare them with the experimental thermal cycle curve to verify the accuracy of the model. Input the predicted parameters into the welding robot to achieve welding prediction effect;

[0069] S2: Design the molten pool control technology during welding, collect the molten pool image through the industrial camera, process the collected molten pool image, and then build a molten pool monitoring platform to monitor the molten pool information in real time, and use the platform to control the molten pool quality during welding;

[0070] S3: Design post-weld quality assessment technology, collect weld surface defect data sets, and train through deep learning. When the welding robot detects weld surface defects, it can record the location and type of the defects and issue an alarm. Surface defects include at least cracks, pores, and weld nodules.

[0071] S1 includes at least the following steps:

[0072] S1.1: The thermal cycle curve of the welded plate in the test is measured by a thermocouple sensor. This method directly contacts the welded parts, and the measurement point recording is more convenient, and the measurement results are more accurate;

[0073] S1.2: Carry out numerical simulation analysis based on welding temperature field theory;

[0074] S1.3: Based on the analysis results of the numerical simulation prediction of welding thermal behavior, the welding process is optimized. For example, by adjusting the welding speed, current, voltage and other parameters, a combination that makes the welding process more stable, higher quality and more efficient can be found, and the welding thermal cycle characteristics can be changed to achieve better welding results. Finally, the proposed optimal welding parameters are input into the welding robot.

[0075] S1.1 includes at least the following steps:

[0076] First, before the test, select appropriate measuring points on the plate to be welded and clean them to prevent the surface defects of the specimen from interfering with the test results;

[0077] Then, a welding test is carried out, and the temperature change of the measuring point during welding is recorded with a thermocouple sensor, and the temperature measurement is terminated when the welded parts cool to room temperature;

[0078] Finally, the test data were integrated to draw the thermal cycle curve during the welding test, and the cross-sectional morphology of the welding pool was measured by image processing methods, such as Figure 3 .

[0079] S1.2 includes at least the following steps:

[0080] First, a proportional model of the welded part is established based on the test;

[0081] Secondly, perform pre-processing settings related to the test conditions. The pre-processing settings at least include material properties, boundary conditions and ambient temperature to restore the welding environment of the test.

[0082] Then, a welding subroutine is written using Fortran language, the welding subroutine at least includes a heat source model, a welding path and welding parameters, and a thermodynamic simulation is performed;

[0083] Finally, the thermal cycle curves of the corresponding calculation nodes at the positions corresponding to the test are extracted and fitted with the thermal cycle curves measured in the test to verify the correctness of the simulation model. Figure 4 .

[0084] S2 includes at least the following steps:

[0085] S2.1: Monitor the liquid flow of the molten pool during welding through an industrial camera and process the collected images;

[0086] S2.2: Obtain accurate molten pool information, calculate values ​​according to specific conditions, and ensure welding stability;

[0087] S2.3: Develop an intelligent welding interface to read the above molten pool information and monitor the molten pool conditions in real time during welding. The size of the molten pool, the overlap between the center of the molten pool and the center line of the weld, and the fusion of the parent material are all important monitoring objects. When the molten pool size is not good during welding, the overlap between the center of the molten pool and the center line of the weld is too large, and the fusion of the parent material is poor, the monitoring system can make a judgment on the molten pool quality. After that, the system sends feedback adjustment instructions to the welding robot control terminal to adjust the trajectory and welding parameters of the welding robot. The welding parameters include at least speed, current and voltage, and then regulate the molten pool quality, such as Figure 7 In summary, the welding robot vision module is used to build a monitoring system to monitor the quality of the molten pool during welding and to control the quality of the molten pool, and the welding robot execution module is used to control the welding trajectory and welding speed.

[0088] S2.1 includes at least the following steps:

[0089] The camera is installed on the welding gun so that the camera moves synchronously with the welding gun to extract the molten pool information in real time;

[0090] In order to avoid interference caused by strong arc light during welding, it is also necessary to install a neutral density filter and a filter on the camera lens;

[0091] The neutral density filter uses OD neutral density film;

[0092] The filter is a 650nm filter, which can effectively reduce the impact of arc light on image acquisition during welding. Figure 5 .

[0093] S2.2 includes at least the following steps:

[0094] Firstly, the image processing algorithm is used to reduce the noise of the collected image and enhance the melt pool information to be collected;

[0095] Secondly, the edge extraction based on directional wavelet and the method based on active contour model are used to extract the edge features of the molten pool;

[0096] Finally, calculate the size of the molten pool, monitor the overlap between the center of the molten pool and the center line of the weld, and the fusion of the parent material to ensure the stability of the welding process, such as Figure 6 .

[0097] S3 includes at least the following steps:

[0098] S3.1: Detect weld surface defects based on deep learning methods and judge welding quality;

[0099] S3.2: Compare different target detection methods, such as the 2-stage method represented by R-CNN, Fast R-CNN, and Faster R-CNN, and the 2-stage method represented by YOLO and SSD, select the best model for weld surface defect detection, then optimize the model to improve the detection accuracy of the model, and finally conduct sufficient model training times to detect weld surface defects and judge the weld quality, such as Fig. 9 , so that the welding robot has the function of detecting weld surface defects. When it detects surface defects, it can use the defect detection function of the welding robot to record the location and type of the defects in the monitoring system, and issue an alarm to remind workers to carry out post-welding grinding, repair welding, etc. in time to ensure welding quality.

[0100] S3.1 includes at least the following steps:

[0101] First, common weld surface defect images are collected and weld surface defect detection datasets are prepared, such as Figure 8 ;

[0102] Then, in order to improve the generalization ability of the model and the detection effect of the model, data enhancement operations are performed on the data set. The data enhancement operations at least include flipping, rotating, shifting, cropping, deforming and scaling;

[0103] Finally, build a deep learning training model.

[0104] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for controlling the quality of the entire welding process of a welding robot, characterized in that: At least the following steps are included: S1: Design pre-welding parameter formulation technology, perform welding numerical simulation analysis to predict the optimal welding parameters, and compare them with the experimental thermal cycle curve to verify the accuracy of the model. Input the predicted parameters into the welding robot to achieve welding prediction effect; S2: Design the molten pool control technology during welding, collect the molten pool image through the industrial camera, process the collected molten pool image, and then build a molten pool monitoring platform to monitor the molten pool information in real time, and use the platform to control the molten pool quality during welding; S3: Design post-weld quality assessment technology, collect weld surface defect data sets, and train through deep learning. When the welding robot detects weld surface defects, it can record the location and type of the defects and issue an alarm. The surface defects include at least cracks, pores and weld nodules.

2. The whole-process quality control method of a welding robot according to claim 1 is characterized in that: The S1 at least comprises the following steps: S1.1: The thermal cycle curve of the welded plate in the test is measured by a thermocouple sensor. This method directly contacts the welded parts, and the measurement point recording is more convenient, and the measurement results are more accurate; S1.2: Carry out numerical simulation analysis based on welding temperature field theory; S1.3: Based on the analysis results of the numerical simulation to predict the welding thermal behavior, the welding process is optimized to achieve better welding results. Finally, the proposed optimal welding parameters are input into the welding robot.

3. The whole-process quality control method of a welding robot according to claim 2 is characterized in that: S1.1 at least includes the following steps: First, before the test, select appropriate measuring points on the plate to be welded and clean them to prevent the surface defects of the specimen from interfering with the test results; Then, a welding test is carried out, and the temperature change of the measuring point during welding is recorded with a thermocouple sensor, and the temperature measurement is terminated when the welded parts cool to room temperature; Finally, the test data were integrated to draw the thermal cycle curve during the welding test, and the cross-sectional morphology of the welding pool was measured by image processing method.

4. The whole-process quality control method of a welding robot according to claim 2 is characterized in that: S1.2 at least The following steps are involved: First, a proportional model of the welded part is established based on the test; Secondly, pre-processing settings related to the test conditions are performed, wherein the pre-processing settings at least include material properties, boundary conditions and ambient temperature, and restore the welding environment of the test; Then, a welding subroutine is written using Fortran language, wherein the welding subroutine at least includes a heat source model, a welding path and welding parameters, and a thermodynamic simulation is performed; Finally, the thermal cycle curves of the corresponding calculation nodes at the positions corresponding to the test are extracted and fitted with the thermal cycle curves measured in the test to verify the correctness of the simulation model.

5. The whole-process quality control method of a welding robot according to claim 1 is characterized in that: The S2 at least comprises the following steps: S2.1: Monitor the liquid flow of the molten pool during welding through an industrial camera and process the collected images; S2.2: Obtain accurate molten pool information, calculate values ​​according to specific conditions, and ensure welding stability; S2.3: Develop an intelligent welding interface to read the above molten pool information and monitor the molten pool conditions during welding in real time. The size of the molten pool, the overlap between the center of the molten pool and the center line of the weld, and the fusion of the parent material are all important monitoring objects. When the molten pool size is not good during welding, the overlap between the center of the molten pool and the center line of the weld is too large, and the fusion of the parent material is poor, the monitoring system can make a judgment on the molten pool quality, and then send feedback adjustment instructions to the welding robot control terminal through the system to adjust the trajectory and welding parameters of the welding robot. The welding parameters include at least speed, current and voltage, thereby regulating the molten pool quality.

6. The whole-process quality control method of a welding robot according to claim 5 is characterized in that: The S2.1 at least comprises the following steps: The camera is installed on the welding gun so that the camera moves synchronously with the welding gun to extract the molten pool information in real time; In order to avoid interference caused by strong arc light during welding, it is also necessary to install a neutral density filter and a filter on the camera lens; The neutral density filter adopts an OD neutral density film; The filter is a 650nm filter, which can effectively reduce the impact of arc light on image acquisition during welding.

7. The whole-process quality control method of a welding robot according to claim 1 is characterized in that: The S2.2 at least comprises the following steps: Firstly, the image processing algorithm is used to reduce the noise of the collected image and enhance the melt pool information to be collected; Secondly, the edge extraction based on directional wavelet and the method based on active contour model are used to extract the edge features of the molten pool; Finally, the molten pool size is calculated, the overlap between the molten pool center and the weld centerline and the fusion of the parent material are monitored to ensure the stability of the welding process.

8. The whole-process quality control method of a welding robot according to claim 1 is characterized in that: The S3 at least includes the following steps: S3.1: Detect weld surface defects based on deep learning methods and judge welding quality; S3.2: Compare and use different target detection methods to select the best model for weld surface defect detection, then optimize the model to improve the detection accuracy of the model, and finally carry out sufficient model training times to detect weld surface defects and judge the weld quality, so that the welding robot has the function of weld surface defect detection. When it detects surface defects, it can use the defect detection function of the welding robot to record the location and type of the defect in the monitoring system, and issue an alarm to remind workers to carry out post-weld grinding, repair welding, etc. in time to ensure welding quality.

9. The whole-process quality control method of a welding robot according to claim 8, characterized in that: S3.1 at least The following steps are involved: Firstly, common weld surface defect images are collected and a weld surface defect detection dataset is created; Then, in order to improve the generalization ability of the model and the detection effect of the model, the data set is subjected to data enhancement operations, wherein the data enhancement operations at least include flipping, rotating, shifting, cropping, deforming and scaling; Finally, build a deep learning training model.

Images