Automobile tail lamp welding system and method

Abstract

The invention discloses a welding method, belongs to the technical field of welding systems, and particularly relates to an automobile tail lamp welding system and method.The automobile tail lamp welding system comprises a switch unit, a heating unit, a welding unit, a curing unit, a detection unit and a main control unit; the detection unit comprises a pre-welding detection module, a welding detection module and a post-welding detection module; full-process online monitoring and closed-loop control of the welding quality are achieved, so that the stability of the automobile welding quality is improved, the advanced welding seam online detection technology, the big data analysis and intelligent decision technology and the like are integrated, and full-process accurate monitoring of the welding quality can be achieved; the system can automatically and efficiently recognize various welding defects and intelligently adjust welding process parameters based on a big data analysis result, defects are completely eradicated from the source, and the welding quality is continuously optimized. By means of the system, welding defects can be effectively avoided, and welding consistency and stability are remarkably improved.

Description

Technical Field

[0001] The invention belongs to the technical field, and in particular relates to a welding system and method for automobile taillights. Background Art

[0002] Hot plate welding of automobile taillights is a welding technology used by automobile taillight welding systems and methods to connect plastic parts, and is particularly suitable for repairing automobile taillight housings and other plastic components. Hot plate welding is a method of using a heating plate to heat the plastic contact surface to a molten state and then press the two parts together. It is usually used to connect two thermoplastic plastic parts together. This technology has high precision and firmness, and is often used in the repair and production of automobile taillights, body plastic parts, etc.

[0003] The basic principle of hot plate welding is to heat the connecting surfaces of two plastic parts to a molten state through a heating plate (usually a metal plate made of aluminum or steel), then quickly press the parts together under pressure, and form a strong connection after cooling.

[0004] In the existing technology, automobile welding quality has a great impact on the quality and safety of the whole vehicle. However, traditional welding quality inspection relies on manual sampling and offline inspection, which is inefficient and has a large detection blind spot. It is impossible to detect and correct welding defects in time, which affects the stability of automobile welding quality. Summary of the invention

[0005] Purpose of the invention: To provide a system and method for welding automobile taillights to solve the above-mentioned problems.

[0006] Technical solution:

[0007] In the first aspect, a car taillight welding system includes: a switch unit, including an equipment gas source switch and an equipment power switch, after turning on the switch, a text operation screen is entered on the touch screen to set parameters such as welding temperature, heating time, and curing time;

[0008] A heating unit is used to preheat the upper and lower hot molds to a set temperature and to heat them;

[0009] A welding unit, used for welding workpieces;

[0010] A curing unit is used to cure the workpiece after welding;

[0011] Detection unit, used for real-time detection of each working stage of the product;

[0012] The main control unit is used to connect the switch unit, the switch unit, the detection unit, the welding unit and the curing unit to control the work of each unit.

[0013] In a further embodiment, the detection unit comprises:

[0014] The pre-welding inspection module is used to detect whether there are defects in the welding place to ensure the cleanliness of the welding place. It collects pictures visually and analyzes them, classifies the defects, and sends the classification results to the welding module.

[0015] The welding detection module is used to simulate the physical phenomena of welding products. It includes various equipment and workpiece models required by the welding workshop. The model can be used for motion simulation and physical behavior simulation based on real-time data. The model motion simulation simulates the rotation and extension motion of the welding equipment, the movement of the workpiece, and the movement of other equipment. It also simulates the collision behavior of objects and the generation behavior of welds during the welding process.

[0016] The post-weld inspection module is used to inspect the quality, residual stress and weld spatter of the area after welding, and connect the welding module to optimize the next welding.

[0017] In a second aspect, a method for welding a car taillight is implemented by using a car taillight welding system, and the welding method process includes:

[0018] Process 1. Switch unit Turn on the gas source switch of the equipment, turn on the power switch of the equipment, enter the text operation screen on the touch screen, and set the welding temperature, heating time, curing time and other parameters;

[0019] Process 2: Turn on the heating switch of the heating unit to preheat the upper and lower heating molds to the set temperature;

[0020] Process 3: When the warning light is green, press the reset button, the tooling and hot mold return to the initial position, and the equipment enters the automatic mode and waits for welding;

[0021] Process 4: Place the workpiece to be welded on the upper die, press the upper die clamping switch on the left, and use the vacuum suction cup to absorb the upper workpiece; install the lower workpiece in the same way; after installation, press the start buttons on both sides at the same time, and the welding unit starts automatic welding;

[0022] Process 5: The safety grating detects the safety of personnel in the welding area, the temperature sensor detects whether the temperature of the upper and lower hot molds meets the set requirements, the pressure sensor detects the vacuum adsorption of the upper and lower workpieces, and the position sensor detects the position of the upper and lower molds and the upper and lower hot molds. The welding action starts after it meets the program setting requirements;

[0023] Process 6: The hot plate movement cylinder drives the hot plate to move the upper and lower hot molds installed on it to the welding position, and the cylinder sensor detects whether the upper and lower hot molds are in place;

[0024] Process 7: The upper and lower mold moving cylinders push the upper and lower molds from the starting position to the heating position. After the heating position sensor detects that the upper and lower molds are in place, the timing starts according to the set heating time and heating is performed;

[0025] Process 8: After the heating time is over, the upper and lower mold moving cylinders pull the upper and lower molds back to the starting positions. After the starting position sensor detects that the molds are in place, the hot plate moving cylinders return the upper and lower hot molds to the hot mold starting positions.

[0026] Process 9: After the cylinder sensor detects that the hot plate cylinder is reset, the moving cylinders of the upper and lower molds push the upper and lower molds to the welding position. After the welding position sensor detects that the upper and lower molds are in place, pressure holding and curing begins according to the set curing time;

[0027] Process 10: After the curing time is over, the vacuum adsorption of the upper and lower workpieces is released, and the upper and lower molds are pulled back to the starting position by the moving cylinders. After the starting position sensor detects that the upper and lower molds are reset to the origin, the equipment program enters the standby state;

[0028] Process 11: After welding is completed, take out the workpiece.

[0029] In a further embodiment, during welding, the detection unit performs the following detection on the workpiece:

[0030] Pre-welding defect detection is used to detect the quality of the product welding before welding;

[0031] Post-weld quality inspection is used to inspect the quality and residual stress of the area after welding.

[0032] In a further embodiment, the pre-weld defect detection collects images visually and merges the weld seam and defect points obtained by preliminary extraction as the entire weld seam. The specific steps are as follows:

[0033] Step 1, read the original image, and obtain the preliminary weld image G1 = (x, y) by dynamic threshold;

[0034] Step 2: Obtain the circumscribed rectangular image G of G1 = (x, y) ′ =(x,y), and use the dynamic threshold to extract the defect point G2 =(x,y);

[0035] Step 3: If there are defects, take the union of G1 = (x, y) and G2 = (x, y) as the whole weld G3 = (x, y); if there are no defects, take G1 = (x, y) as the whole weld G3 = (x, y);

[0036] Step 4: Morphology and stroke processing to remove edge interference and obtain accurate weld area R ′ =(x,y);

[0037] Step 5: In the weld area R ′ =(x,y) and use the global threshold to extract the welding position.

[0038] In a further embodiment, in step 5, the weld is first segmented in the entire area using a dynamic threshold method, and then the sensitive information is segmented again in the weld area using a global threshold method to extract the welding position area. The specific steps are as follows:

[0039] Step 51, first filter the entire image, then make a difference between the original image and the filtered image, and determine the extraction area according to the difference;

[0040] Step 52, after the image is smoothed, the grayscale values ​​of the original image and the smoothed image are compared. If the grayscale value of the original image is greater than the grayscale value of the smoothed image, it is equivalent to selecting a brighter area in the image, which corresponds to the weld area in the image in the text.

[0041] Step 53: Obtain the minimum circumscribed rectangle image of the preliminary weld image, and use the dynamic threshold to extract defect points again, and merge the preliminary weld and the burst point according to the collection formula as the entire weld area.

[0042] Step 54, extract the welding position, establish a feature model of the welding position and calculate the feature vector of the candidate area of ​​the welding position, use principal component analysis to reduce the dimension of the feature vector, and use support vector machine to classify the welding position and texture to obtain the specific welding position.

[0043] In a further embodiment, the quality inspection in the post-weld quality inspection constructs a welding defect recognition model and judges the quality problems of the weld, and adopts a welding quality inspection model of multi-modal tensor fusion;

[0044] First, each modal subnet extracts different modal features, and then multimodal feature fusion is realized by tensor fusion. Next, the attention weight on the tensor is calculated through the tensor attention module, and finally the welding quality prediction value is obtained by the feature re-extraction structure.

[0045] In a further embodiment, for welding single-mode data, welding pool shape, pool temperature, and current and voltage characteristics are extracted respectively through three subnets;

[0046] The shape and temperature features of the welding pool are extracted by using two single-side feature extraction networks with the same structure to obtain the shape feature vectors of the front and back sides of the welding pool. Then, the shape feature vectors of the welding pool are obtained by splicing the feature channels and passing through the fully connected layer.

[0047] Extraction of welding current and voltage features: A set of welding current and voltage sequences are divided into 10 time steps and sent into LSTM. The output feature vector is reduced in dimension through full connection. After obtaining two feature vectors, feature concatenation and full connection are also used to obtain the welding current and voltage features.

[0048] In a further embodiment, the residual stress detection in the post-weld quality detection first collects sound during the welding process, and then filters the collected sound signals.

[0049] Beneficial effects: The present invention realizes full-process online monitoring and closed-loop control of welding quality to improve the stability of automobile welding quality. It integrates advanced online weld detection technology, big data analysis and intelligent decision-making technology, etc., and can realize full-process accurate monitoring of welding quality. The system can automatically and efficiently identify various types of welding defects, and intelligently adjust welding process parameters based on big data analysis results to eliminate defects from the source and continuously optimize welding quality. The application of this system can not only effectively avoid welding defects, but also significantly improve welding consistency and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a system schematic diagram of the present invention.

[0051] Figure 2 It is a pre-welding detection flow chart of the present invention.

[0052] Figure 3 It is a flow chart of post-weld quality inspection of the present invention. DETAILED DESCRIPTION

[0053] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Embodiment 1:

[0055] A vehicle taillight welding system and method, such as Figure 1 As shown, including:

[0056] The switch unit includes the equipment gas source switch and the equipment power switch. After turning on the switch, the text operation screen is entered on the touch screen to set parameters such as welding temperature, heating time, and curing time;

[0057] A heating unit is used to preheat the upper and lower hot molds to a set temperature and to heat them;

[0058] A welding unit, used for welding workpieces;

[0059] A curing unit is used to cure the workpiece after welding;

[0060] Detection unit, used for real-time detection of each working stage of the product;

[0061] The main control unit is used to connect the switch unit, the switch unit, the detection unit, the welding unit and the curing unit to control the work of each unit.

[0062] In one embodiment, the detection unit comprises:

[0063] The pre-welding inspection module is used to detect whether there are defects in the welding place to ensure the cleanliness of the welding place. It collects pictures visually and analyzes them, classifies the defects, and sends the classification results to the welding module.

[0064] The welding detection module is used to simulate the physical phenomena of welding products. It includes various equipment and workpiece models required by the welding workshop. The model can be used for motion simulation and physical behavior simulation based on real-time data. The model motion simulation simulates the rotation and extension motion of the welding equipment, the movement of the workpiece, and the movement of other equipment. It also simulates the collision behavior of objects and the generation behavior of welds during the welding process.

[0065] The post-weld inspection module is used to inspect the quality, residual stress and weld spatter of the area after welding, and connect the welding module to optimize the next welding.

[0066] Embodiment 2:

[0067] A method for welding a car taillight is implemented by using a car taillight welding system. The welding method process includes:

[0068] Process 1. Switch unit Turn on the gas source switch of the equipment, turn on the power switch of the equipment, enter the text operation screen on the touch screen, and set the welding temperature, heating time, curing time and other parameters;

[0069] Process 2: Turn on the heating switch of the heating unit to preheat the upper and lower heating molds to the set temperature;

[0070] Process 3: When the warning light is green, press the reset button, the tooling and hot mold return to the initial position, and the equipment enters the automatic mode and waits for welding;

[0071] Process 4: Place the workpiece to be welded on the upper die, press the upper die clamping switch on the left, and use the vacuum suction cup to absorb the upper workpiece; install the lower workpiece in the same way; after installation, press the start buttons on both sides at the same time, and the welding unit starts automatic welding;

[0072] Process 5: The safety grating detects the safety of personnel in the welding area, the temperature sensor detects whether the temperature of the upper and lower hot molds meets the set requirements, the pressure sensor detects the vacuum adsorption of the upper and lower workpieces, and the position sensor detects the position of the upper and lower molds and the upper and lower hot molds. The welding action starts after it meets the program setting requirements;

[0073] Process 6: The hot plate movement cylinder drives the hot plate to move the upper and lower hot molds installed on it to the welding position, and the cylinder sensor detects whether the upper and lower hot molds are in place;

[0074] Process 7: The upper and lower mold moving cylinders push the upper and lower molds from the starting position to the heating position. After the heating position sensor detects that the upper and lower molds are in place, the timing starts according to the set heating time and heating is performed;

[0075] Process 8: After the heating time is over, the upper and lower mold moving cylinders pull the upper and lower molds back to the starting positions. After the starting position sensor detects that the molds are in place, the hot plate moving cylinders return the upper and lower hot molds to the hot mold starting positions.

[0076] Process 9: After the cylinder sensor detects that the hot plate cylinder is reset, the moving cylinders of the upper and lower molds push the upper and lower molds to the welding position. After the welding position sensor detects that the upper and lower molds are in place, pressure holding and curing begins according to the set curing time;

[0077] Process 10: After the curing time is over, the vacuum adsorption of the upper and lower workpieces is released, and the upper and lower molds are pulled back to the starting position by the moving cylinders. After the starting position sensor detects that the upper and lower molds are reset to the origin, the equipment program enters the standby state;

[0078] Process 11: After welding is completed, take out the workpiece.

[0079] In one embodiment, during welding, the detection unit performs the following detection on the workpiece:

[0080] Pre-welding defect detection is used to detect the quality of the product welding before welding;

[0081] Post-weld quality inspection is used to inspect the quality and residual stress of the area after welding.

[0082] In one embodiment, Figure 2 As shown, the pre-welding defect detection collects images visually and merges the weld seam and defect points obtained by preliminary extraction as the entire weld seam. The specific steps are as follows:

[0083] Step 1, read the original image, and obtain the preliminary weld image G1 = (x, y) by dynamic threshold;

[0084] Step 2: Obtain the circumscribed rectangular image G of G1 = (x, y) ′=(x,y), and use the dynamic threshold to extract the defect point G2 =(x,y);

[0085] Step 3: If there are defects, take the union of G1 = (x, y) and G2 = (x, y) as the whole weld G3 = (x, y); if there are no defects, take G1 = (x, y) as the whole weld G3 = (x, y);

[0086] Step 4: Morphology and stroke processing to remove edge interference and obtain accurate weld area R ′ =(x,y);

[0087] Step 5: In the weld area R ′ =(x,y) and use the global threshold to extract the welding position.

[0088] In one embodiment, in step 5, the weld is first segmented in the entire area using a dynamic threshold method, and then the sensitive information is segmented again in the weld area using a global threshold method to extract the welding position area. The specific steps are as follows:

[0089] Step 51, first filter the entire image, then make a difference between the original image and the filtered image, and determine the extraction area according to the difference;

[0090] Step 52, after the image is smoothed, the grayscale values ​​of the original image and the smoothed image are compared. If the grayscale value of the original image is greater than the grayscale value of the smoothed image, it is equivalent to selecting a brighter area in the image, which corresponds to the weld area in the image in the text.

[0091] Step 53: Obtain the minimum circumscribed rectangle image of the preliminary weld image, and use the dynamic threshold to extract defect points again, and merge the preliminary weld and the burst points according to the collection formula as the entire weld area.

[0092] Step 54, extract the welding position, establish a feature model of the welding position and calculate the feature vector of the candidate area of ​​the welding position, use principal component analysis to reduce the dimension of the feature vector, and use support vector machine to classify the welding position and texture to obtain the specific welding position.

[0093] In one embodiment, Figure 3 As shown, the quality inspection in the post-weld quality inspection builds a welding defect recognition model and judges the quality problems of the weld, and adopts a welding quality inspection model of multi-modal tensor fusion;

[0094] First, each modal subnet extracts different modal features, and then multimodal feature fusion is realized by tensor fusion. Next, the attention weight on the tensor is calculated through the tensor attention module, and finally the welding quality prediction value is obtained by the feature re-extraction structure.

[0095] In one embodiment, for welding single-mode data, the welding pool shape, the welding pool temperature, and the current and voltage characteristics are extracted respectively through three subnets;

[0096] The shape and temperature features of the welding pool are extracted by using two single-side feature extraction networks with the same structure to obtain the shape feature vectors of the front and back sides of the welding pool. Then, the shape feature vectors of the welding pool are obtained by splicing the feature channels and passing through the fully connected layer.

[0097] Extraction of welding current and voltage features: A set of welding current and voltage sequences are divided into 10 time steps and sent into LSTM. The output feature vector is reduced in dimension through full connection. After obtaining two feature vectors, feature concatenation and full connection are also used to obtain the welding current and voltage features.

[0098] In one embodiment, the residual stress detection in the post-weld quality detection first collects sound during the welding process, filters the collected sound signals through a filter, and finally extracts the sound signal features and inputs them into the detection model to determine the residual stress.

[0099] In one embodiment, the hot plate welding process:

[0100] Preparation: Clean the surface to be welded, remove surface oil, dust and other impurities, and ensure that the contact surface is clean. Align the two plastic parts to be welded and fix them in the welding machine fixture.

[0101] Heating stage: Place the heating plate between the workpieces to be connected. The contact surface is heated to the melting temperature by the heating plate. Different types of products have different melting temperatures, and the hot plate welding machine usually automatically adjusts the temperature according to the material requirements.

[0102] Welding stage:

[0103] After heating is completed, quickly remove the heating plate and immediately weld it through the welding equipment

[0104] Cooling and curing stage:

[0105] Once the weld is complete, pressure is continued until the weld cools and solidifies. The cooling process is critical because it determines the strength and tightness of the weld joint.

[0106] Inspection and repair:

[0107] After welding is completed, check the appearance and strength of the weld area to ensure there are no bubbles, cracks or looseness.

[0108] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. An automobile taillight welding system, characterized in that: The welding system comprises: The switch unit includes the equipment gas source switch and the equipment power switch. After turning on the switch, the text operation screen is entered on the touch screen to set parameters such as welding temperature, heating time, and curing time; A heating unit is used to preheat the upper and lower hot molds to a set temperature and to heat them; A welding unit, used for welding workpieces; A curing unit is used to cure the workpiece after welding; Detection unit, used for real-time detection of each working stage of the product; The main control unit is used to connect the switch unit, the switch unit, the detection unit, the welding unit and the curing unit to control the work of each unit.

2. The automobile taillight welding system according to claim 1, characterized in that: The detection unit comprises: The pre-welding inspection module is used to detect whether there are defects in the welding place to ensure the cleanliness of the welding place. It collects pictures visually and analyzes them, classifies the defects, and sends the classification results to the welding module. The welding detection module is used to simulate the physical phenomena of welding products. It includes various equipment and workpiece models required by the welding workshop. The model can be used for motion simulation and physical behavior simulation based on real-time data. The model motion simulation simulates the rotation and extension motion of the welding equipment, the movement of the workpiece, and the movement of other equipment. It also simulates the collision behavior of objects and the generation behavior of welds during the welding process. The post-weld inspection module is used to inspect the quality, residual stress and weld spatter of the area after welding, and connect the welding module to optimize the next welding.

3. A method for welding automobile taillights, characterized in that: The method is implemented by using a car taillight welding system as described in any one of claims 1 to 2, wherein the welding method process comprises: Process 1. Switch unit Turn on the gas source switch of the equipment, turn on the power switch of the equipment, enter the text operation screen on the touch screen, and set the welding temperature, heating time, curing time and other parameters; Process 2: Turn on the heating switch of the heating unit to preheat the upper and lower heating molds to the set temperature; Process 3: When the warning light is green, press the reset button, the tooling and hot mold return to the initial position, and the equipment enters the automatic mode and waits for welding; Process 4: Place the workpiece to be welded on the upper die, press the upper die clamping switch on the left, and use the vacuum suction cup to absorb the upper workpiece; install the lower workpiece in the same way; after installation, press the start buttons on both sides at the same time, and the welding unit starts automatic welding; Process 5: The safety grating detects the safety of personnel in the welding area, the temperature sensor detects whether the temperature of the upper and lower hot molds meets the set requirements, the pressure sensor detects the vacuum adsorption of the upper and lower workpieces, and the position sensor detects the position of the upper and lower molds and the upper and lower hot molds. The welding action starts after it meets the program setting requirements; Process 6: The hot plate movement cylinder drives the hot plate to move the upper and lower hot molds installed on it to the welding position, and the cylinder sensor detects whether the upper and lower hot molds are in place; Process 7: The upper and lower mold moving cylinders push the upper and lower molds from the starting position to the heating position. After the heating position sensor detects that the upper and lower molds are in place, the timing starts according to the set heating time and heating is performed; Process 8: After the heating time is over, the upper and lower mold moving cylinders pull the upper and lower molds back to the starting positions. After the starting position sensor detects that the molds are in place, the hot plate moving cylinders return the upper and lower hot molds to the hot mold starting positions. Process 9: After the cylinder sensor detects that the hot plate cylinder is reset, the moving cylinders of the upper and lower molds push the upper and lower molds to the welding position. After the welding position sensor detects that the upper and lower molds are in place, pressure holding and curing begins according to the set curing time; Process 10: After the curing time is over, the vacuum adsorption of the upper and lower workpieces is released, and the upper and lower molds are pulled back to the starting position by the moving cylinders. After the starting position sensor detects that the upper and lower molds are reset to the origin, the equipment program enters the standby state; Process 11: After welding is completed, take out the workpiece.

4. The automobile taillight welding method according to claim 3, characterized in that: During welding, the inspection unit performs the following inspections on the workpiece: Pre-welding defect detection is used to detect the quality of the product welding before welding; Post-weld quality inspection is used to inspect the quality and residual stress of the area after welding.

5. The automobile taillight welding method according to claim 4, characterized in that: The pre-weld defect detection collects images visually and merges the weld seam and defect points obtained by preliminary extraction as the entire weld seam. The specific steps are as follows: Step 1, read the original image, and obtain the preliminary weld image G1 = (x, y) by dynamic threshold; Step 2: Obtain the circumscribed rectangular image G of G1 = (x, y) ′ =(x,y), and use the dynamic threshold to extract the defect point G2 =(x,y); Step 3: If there are defects, take the union of G1 = (x, y) and G2 = (x, y) as the whole weld G3 = (x, y); if there are no defects, take G1 = (x, y) as the whole weld G3 = (x, y); Step 4: Morphology and stroke processing to remove edge interference and obtain accurate weld area R ′ =(x,y); Step 5: In the weld area R ′ =(x,y) and use the global threshold to extract the welding position.

6. A method for welding automobile taillights according to claim 5, characterized in that: In step 5, the weld is first segmented using the dynamic threshold method in the entire area, and then the sensitive information is segmented again using the global threshold method in the weld area to extract the welding position area. The specific steps are as follows: Step 51, first filter the entire image, then make a difference between the original image and the filtered image, and determine the extraction area according to the difference; Step 52, after the image is smoothed, the grayscale values ​​of the original image and the smoothed image are compared. If the grayscale value of the original image is greater than the grayscale value of the smoothed image, it is equivalent to selecting a brighter area in the image, which corresponds to the weld area in the image in the text. Step 53: Obtain the minimum circumscribed rectangle image of the preliminary weld image, and use the dynamic threshold to extract defect points again, and merge the preliminary weld and the burst point according to the collection formula as the entire weld area. Step 54, extract the welding position, establish a feature model of the welding position and calculate the feature vector of the candidate area of ​​the welding position, use principal component analysis to reduce the dimension of the feature vector, and use support vector machine to classify the welding position and texture to obtain the specific welding position.

7. The automobile taillight welding method according to claim 4, characterized in that: The quality inspection after welding builds a welding defect recognition model and judges the quality problems of the weld, and adopts a welding quality inspection model of multi-modal tensor fusion; First, each modal subnet extracts different modal features, and then multimodal feature fusion is realized by tensor fusion. Next, the attention weight on the tensor is calculated through the tensor attention module, and finally the welding quality prediction value is obtained by the feature re-extraction structure.

8. The automobile taillight welding method according to claim 7, characterized in that: For welding single-mode data, the welding pool shape, pool temperature, and current and voltage characteristics are extracted through three subnets respectively; The shape and temperature features of the welding pool are extracted by using two single-side feature extraction networks with the same structure to obtain the shape feature vectors of the front and back sides of the welding pool. Then, the shape feature vectors of the welding pool are obtained by splicing the feature channels and passing through the fully connected layer. Extraction of welding current and voltage features: A set of welding current and voltage sequences are divided into 10 time steps and sent into LSTM. The output feature vector is reduced in dimension through full connection. After obtaining two feature vectors, feature concatenation and full connection are also used to obtain the welding current and voltage features.

9. The automobile taillight welding method according to claim 4, characterized in that: The residual stress detection in the post-weld quality detection first collects the sound during the welding process, filters the collected sound signals through a filter, and finally extracts the sound signal features and inputs them into the detection model to determine the residual stress.

Images