Aluminum alloy GTAW imitation welder welding speed control system based on multiple perception

Abstract

The invention relates to the field of GTAW welding, in particular to an aluminum alloy GTAW imitation welder welding speed control system and method based on multiple perception. The aluminum alloy GTAW imitation welder welding speed control system and method based on multiple perceptions are provided for solving the problems that the welding seam forming quality is unstable and adjustment depends on artificial experience due to the fact that the welding speed is not matched with the dynamically-changing welding state. A GTAW robot is used for welding; a visual sensor is used for shooting molten pool boundary and molten pool rear images; capturing the three-dimensional shape of the molten pool by using a three-line structure light sensor; beveled edge width information is obtained in real time through a single-line structured light sensor; detecting arc sound by using a sound sensor; and the industrial personal computer analyzes and processes the obtained data. On the basis of accurate analysis of dynamic characteristics of the molten pool, control over welding seam quality forming is finally achieved.

Description

TECHNICAL FIELD

[0001] The application relates to the field of aluminum alloy GTAW welding and is an aluminum alloy GTAW welding speed control system based on multiple perceptions. BACKGROUND

[0002] Aluminum alloys are widely used in the fields of aerospace, rail transit, new energy vehicles and other high-end equipment manufacturing due to their high specific strength and good corrosion resistance. As a main method for high-quality welding, GTAW (gas tungsten arc welding) is often used for the connection of key structures of aluminum alloys. The limitations of fixed program control and the shortcomings of single sensor technology limit the real-time perception and feedback capability of the welding process, and the welding speed cannot be adjusted in real time according to the molten pool state, groove assembly gap change and arc sound, so that the welding quality is difficult to keep stable and still needs manual intervention. The application discloses an aluminum alloy GTAW welding speed control system based on multiple perceptions, which can simulate the multi-sensory perception and real-time decision-making capability of skilled welders, realizes the self-adaptive and accurate adjustment of the welding speed to the dynamic welding environment, and guarantees the high quality of the weld formation. SUMMARY

[0003] The aluminum alloy GTAW welding speed control system based on multiple perceptions is composed of a GTAW robot, a visual sensor, a three-line structured light sensor, a single-line structured light sensor, a sound sensor and an industrial computer. The GTAW robot is used for controlling the welding gun to weld. The visual sensor is used for taking images of the molten pool boundary and the rear of the molten pool. The three-line structured light sensor is used for capturing the three-dimensional topography of the molten pool. The single-line structured light sensor is used for acquiring the groove width information in real time. The sound sensor is used for detecting the arc sound. The industrial computer is used for analyzing and processing the obtained data. The aluminum alloy GTAW welding speed control system based on multiple perceptions is as shown in Fig. 1. Figure 1 The aluminum alloy GTAW welding speed control system based on multiple perceptions is as shown in Fig. 1.

[0004] An aluminum alloy GTAW welding speed detection method based on molten pool edge detection is used. The molten pool image is obtained through the visual sensor, and the welding state is judged according to the definition of the molten pool boundary. The definition of the molten pool boundary refers to the clear degree of the boundary line between the liquid molten pool and the surrounding solid base material. The average gradient amplitude of the molten pool boundary region is obtained by using the sobel operator and the original image convolution, and the higher the value is, the sharper the molten pool boundary is. The average gradient amplitude of the molten pool boundary region is as shown in formula (1). The formula (1) is as shown in Fig. 2. The formula (1) is as shown in Fig. 2. The formula (2) is as shown in Fig. 3. The formula (3) is as shown in Fig. 4. The formula (4) is as shown in Fig. 5. The formula (5) is as shown in Fig. 6. The formula (6) is as shown in Fig. 7. The formula (7) is as shown in Fig. 8. The formula (8) is as shown in Fig. 9. The formula (9) is as shown in Fig. 10. The formula (10) is as shown in Fig. 11. The formula (11) is as shown in Fig. 12. The formula (12) is as shown in Fig. 13. The formula (13) is as shown in Fig. 14. The formula (14) is as shown in Fig. 15. The formula (15) is as shown in Fig. 16. The formula (16) is as shown in Fig. 17. The formula (17) is as shown in Fig. 18. The formula (18) is as shown in Fig. 19. The formula (19) is as shown in Fig. 20. The formula (20) is as shown in Fig. 21. The formula (21) is as shown in Fig. 22. The formula (22) is as shown in Fig. 23. The formula (23) is as shown in Fig. 24. The formula (24) is as shown in Fig. 25. The formula (25) is as shown in Fig. 26. The formula (26) is as shown in Fig. 27. The formula (27) is as shown in Fig. 28. The formula (28) is as shown in Fig. 29. The formula (29) is as shown in Fig. 30. The formula (30) is as shown in Fig. 31. The formula (31) is as shown in Fig. 32. The formula (32) is as shown in Fig. 33. The formula (33) is as shown in Fig. 34. The formula (34) is as shown in Fig. 35. The formula (35) is as shown in Fig. 36. The formula (36) is as shown in Fig. 37. The formula (37) is as shown in Fig. 38. The formula (38) is as shown in Fig. 39. The formula (39) is as shown in Fig. 40. The formula (40) is as shown in Fig. 41. The formula (41) is as shown in Fig. 42. The formula (42) is as shown in Fig. 43. The formula (43) is as shown in Fig. 44. The formula (44) is as shown in Fig. 45. The formula (45) is as shown in Fig. 46. The formula (46) is as shown in Fig. 47. The formula (47) is as shown in Fig. 48. The formula (48) is as shown in Fig. 49. The formula (49) is as shown in Fig. 50. The formula (50) is as shown in Fig. 51. The formula (51) is as shown in Fig. 52. The formula (52) is as shown in Fig. 53. The formula (53) is as shown in Fig. 54. The formula (54) is as shown in Fig. 55. The formula (55) is as shown in Fig. 56. The formula (56) is as shown in Fig. 57. The formula (57) is as shown in Fig. 58. The formula (58) is as shown in Fig. 59. The formula (59) is as shown in Fig. 60. The formula (60) is as shown in Fig. 61. The formula (61) is as shown in Fig. 62. The formula (62) is as shown in Fig. 63. The formula (63) is as shown in Fig. 64. The formula (64) is as shown in Fig. 65. The formula (65) is as shown in Fig. 66. The formula (66) is as shown in Fig. 67. The formula (67) is as shown in Fig. 68. The formula (68) is as shown in Fig. 69. The formula (69) is as shown in Fig. 70. The formula (70) is as shown in Fig. 71. The formula (71) is as shown in Fig. 72. The formula (72) is as shown in Fig. 73. The formula (73) is as shown in Fig. 74. The formula (74) is as shown in Fig. 75. The formula (75) is as shown in Fig. 76. The formula (76) is as shown in Fig. 77. The formula (77) is as shown in Fig. 78. The formula (78) is as shown in Fig. 79. The formula (79) is as shown in Fig. 80. The formula (80) is as shown in Fig. 81. The formula (81) is as shown in Fig. 82. The formula (82) is as shown in Fig. 83. The formula (83) is as shown in Fig. 84. The formula (84) is as shown in Fig. 85. The formula (85) is as shown in Fig. 86. The formula (86) is as shown in Fig. 87. The formula (87) is as shown in Fig. 88. The formula (88) is as shown in Fig. 89. The formula (89) is as shown in Fig. 90. The formula (90) is as shown in Fig. 91. The formula (91) is as shown in Fig. 92. The formula (92) is as shown in Fig. 93. The formula (93) is as shown in Fig. 94. The formula (94) is as shown in Fig. 95. The formula (95) is as shown in Fig. 96. The formula (96) is as shown in Fig. 97. The formula (97) is as shown in Fig. 98. The formula (98) is as shown in Fig. 99. The formula (99) is as shown in Fig. 100. The formula (100) is as shown in Fig. 101. The It refers to the gradient amplitude of the i-th pixel point within the molten pool boundary region, obtained by dividing the sum of all gradient values within the molten pool boundary region by the total number of pixels; multiple normal lines are calculated within the selected molten pool boundary region to obtain the average transition zone width value. The higher the value, the softer the molten pool boundary; the average transition zone width value is ; the It refers to the average transition zone width of the molten pool boundary region. For each boundary pixel point, search in both the inner and outer sides of the molten pool along the direction perpendicular to the boundary, i.e., the gradient direction, until the gray value stabilizes (the gray difference from the inside of the molten pool is less than the threshold), and record the number of these pixels, which is the transition zone width of this point; the It refers to the single transition zone width of the i-th pixel point within the molten pool boundary region; if k2 and <m1, that is, the average gradient amplitude is on the high side and the average transition zone width value is on the low side, it is determined that the molten pool boundary is sharp and serrated, and the welding speed is fast. At this time, the speed needs to be reduced, and a slight pause of 0.5 - 1 s can be made for adjustment to ensure sufficient fusion between the molten pool and the base material; if <k1 and >m2, that is, the average gradient amplitude is on the low side and the average transition zone width value is on the high side, it is determined that the molten pool boundary is blurred and the welding speed is slow. At this time, the welding speed needs to be gradually increased by 5 - 10 mm / min each time until the boundary becomes clear; if k1 < <k2 and m1 < <m2, that is, the average gradient amplitude and the average transition zone width value are appropriate, it is determined that the clarity of the molten pool boundary is in a normal state, and the transition with the base material is natural, and it is determined that the heat input and the welding speed are matched; the k1 and k2 respectively refer to the minimum preset threshold and the maximum preset threshold of the average gradient amplitude; m1 and m2 respectively refer to the minimum preset threshold and the maximum preset threshold of the average transition zone width value. The schematic diagram of the molten pool boundary is as Figure 3 shown.

[0005] An aluminum alloy GTAW welding speed detection method based on the detection of the molten pool flow trend captures the three-dimensional morphology of the molten pool through a three-line structured light sensor to identify the molten pool flow trend; the capture of the three-dimensional morphology of the molten pool includes the surface height information of the molten pool, the width information of the molten pool, and the contact angle information of the molten pool tail; the three-line structured light sensor projects three structured lights onto the surface of the molten pool, and the camera synchronously takes pictures of the molten pool image containing the light strips. At the same time, the original molten pool image is processed by the Sobel operator to obtain an accurate molten pool boundary. According to the three-dimensional reconstruction algorithm of structured light, the three-dimensional coordinates of all pixels within the molten pool region framed by the Sobel boundary are calculated, and the height direction is taken as the Z-axis, the welding direction is taken as the Y-axis, and the direction perpendicular to the welding direction is taken as the X-axis; the difference between the highest point of the Z-axis of the molten pool surface image and the reference plane is taken as the maximum height of the molten pool surface, and through obtained; the The maximum height of the molten pool surface. Z is the highest Z-axis coordinate value within the molten pool region, and Z0 is the coordinate value of the base material reference plane. The left and right boundaries of the molten pool are located using the Sobel operator. The distance between two points on the boundary in the perpendicular welding direction is taken as the width. Let the coordinates of the left boundary point of the molten pool be (X...). L The coordinates of the right boundary point are (X,y), R If ,y), then the cross-sectional width W y Through W y = X R X L The contact point between the molten pool tail and the base material in the locating molten pool boundary region is calculated. The angle between the two lines obtained by fitting the molten pool tail surface curve and the base material plane at the contact point is the molten pool tail contact angle. Calculated; the aforementioned The normal vector of the molten pool surface. The normal vector of the base material plane is taken; the average value of the parameters in 10 consecutive frames during the welding stabilization phase is used as the reference value, i.e., B. H B W B θ The aforementioned , , The aforementioned , , This refers to the average height, average width, and average contact angle of the molten pool surface; calculate the current frame parameter H. curr W ycurr θ curr Calculate the difference between it and the benchmark value ΔH=H curr B H ΔW y =W curr B W Δθ=θ curr B θ If ΔH>T H ΔW y >T W Δθ>T θT is a preset threshold. If the condition is met for three consecutive frames, the parameter is increased. Regarding the molten pool flow trend in flat welding, if the system detects an increase in molten pool height and width, and an increase in the contact angle between the molten pool tail and the base material, it is determined that the liquid metal on both sides of the molten pool is contracting towards the center, indicating that the welding speed is too fast and the molten pool is not fully spread. In this case, the welding speed needs to be reduced to allow the liquid metal to flow towards the edges. If the molten pool height increases abnormally, the width narrows, and the molten pool tail protrudes, it is determined that the molten pool is bulging in the middle, indicating that the welding speed is too slow, the molten pool exists for too long, and excessive metal is melting. In this case, the welding speed needs to be increased to avoid weld buildup. The installation details of the three-line structured light sensor in the multi-sensor-based aluminum alloy GTAW simulated welder welding speed control system are as follows... Figure 2 As shown.

[0006] A method for detecting the welding speed of aluminum alloy GTAW (Glass-Only Welder) based on the solidification rate of the molten pool is proposed. This method utilizes a visual sensor to acquire high-frequency images of the molten pool behind it. After preprocessing the images to eliminate interference, an edge detection algorithm is used to locate the molten pool boundary and obtain pixel coordinates. The solidification rate is determined by the change in image grayscale values. The image preprocessing involves removing image noise and enhancing the grayscale difference between the molten pool and the background to ensure clear boundaries. The edge detection algorithm uses the Sobel operator to extract continuous and complete molten pool boundaries. The image grayscale values ​​are obtained using the formula Gray = 0.299*R + 0.587*G + 0.114*B, where R, G, and B represent the brightness values ​​of the red, green, and blue channels, respectively. Higher temperatures result in lower grayscale values. The change in grayscale value per unit time is ΔGray. The unit time refers to the time difference between the molten pool's liquid state and its semi-solidified state. This indicates that the temperature drop per unit time is moderate, and the welding speed is moderate; if This indicates a large temperature drop per unit time and a rapid solidification rate of the molten pool, suggesting that the welding speed is too fast. In this case, the welding speed needs to be reduced to extend the heat input time and prevent incomplete penetration at the weld root. If the temperature drop per unit time is small and the solidification rate of the molten pool is slow, then the welding speed is too slow. In this case, the welding speed needs to be increased to prevent the heat-affected zone from becoming too large.

[0007] A welding speed detection method for aluminum alloy GTAW (Glass Surface Artificial Welding) based on dynamic detection of weld bevel changes is proposed. The dynamic detection of weld bevel changes involves real-time detection of the bevel gap width, utilizing a front-mounted single-line structured light sensor to acquire bevel width information in real time. The dynamic adaptation of bevel changes occurs when the bevel gap width increases. When the weld pool fills the gap, the welding speed needs to be slowed down while increasing the filler wire amount. Once the weld pool has filled the gap, the original welding speed should be resumed to ensure the weld pool fully melts the base material and prevents the weld pool from widening. When the bevel gap decreases... At this time, it is necessary to increase the welding speed and avoid localized pile-up; the aforementioned This refers to the difference between the current actual gap and the reference value. This refers to the preset threshold value representing the difference between the actual gap and the reference value. The diagram illustrating the bevel gap width is shown below. Figure 4 As shown.

[0008] The aluminum alloy GTAW simulated welder welding speed control system based on multiple sensing according to claim 1 is characterized by: utilizing an aluminum alloy GTAW simulated welder welding speed detection method based on arc sound feedback; the arc sound feedback aluminum alloy GTAW simulated welder welding speed detection method utilizes a sound sensor to collect arc sound signals in real time, performs feature extraction and time-domain analysis after preprocessing; the time-domain analysis refers to analyzing the root mean square value and peak factor; the root mean square value represents the average amplitude intensity of the sound signal within the calculation time period, and its calculation formula is as follows: = The aforementioned This refers to the voltage value of the sound signal; when At this time, the root mean square value is relatively stable and changes gradually, indicating a moderate welding speed; when If the welding heat input is too low or the welding speed is too high, the welding speed needs to be slowed down; when > At that time, if the welding heat input is too high and the welding speed is too slow, the welding speed needs to be increased; , These are the minimum and maximum preset thresholds for the root mean square value, respectively; the peak factor refers to the ratio of the signal peak value to the effective value, and its calculation formula is as follows: The aforementioned This refers to the maximum absolute value of the signal within the analysis frame. This refers to the root mean square value; when At that time, the sound waveform amplitude is uniform, and the welding speed is moderate; when At this time, large amplitude fluctuations and increased high-frequency components indicate that the welding speed is too fast and needs to be slowed down; when When the amplitude fluctuation is small and the overall signal strength is low, it is determined that the welding speed is too slow, and the welding speed needs to be increased and the arc length shortened.

[0009] The weld formation quality in the multi-sensor-based aluminum alloy GTAW simulated welder welding speed control system is controlled by a multi-sensor-based aluminum alloy GTAW simulated welder welding speed control method. This method controls the welding speed using a GTAW welding robot. Visual sensors, three-wire structured light sensors, single-wire structured light sensors, sound sensors, and an industrial control computer are used to detect the average gradient amplitude of the molten pool boundary region. , average width value of the transition zone , the difference ΔH between the molten pool height and the reference value, and the difference ΔW between the molten pool width and the reference value y , the difference Δθ between the contact angle at the tail of the molten pool and the reference value, and the change value of the gray scale value per unit time , , root mean square value of the sound signal and peak factor ; if k1 k2 and m1 m2, ,<o:p>, , it indicates that the welding speed is moderate and the weld forming quality is good; if k2 and m1, , , , , it indicates that the welding speed is too fast and the welding speed needs to be reduced; if <k1 and >m2, , , > , , it indicates that the welding speed is too slow and the welding speed needs to be increased.

[0010] Advantages of the invention

[0011] The present invention relates to the field of GTAW welding of aluminum alloys, and is a welding speed control system for GTAW welding of aluminum alloys based on multiple sensors. Aiming at the problem that in the GTAW welding of aluminum alloys, the welding speed cannot be adjusted comprehensively and in real time to maintain the weld quality, a welding speed control system for GTAW welding of aluminum alloys based on multiple sensors is proposed. A vision sensor is used to capture images of the molten pool boundary and the rear of the molten pool; a three-line structured light sensor is used to capture the three-dimensional shape of the molten pool; a single-line structured light sensor is used to obtain the groove width information in real time; a sound sensor is used to detect the arc sound; an industrial control computer is used to analyze and process the obtained data. Brief description of the drawings

[0012] Figure 1 is a schematic diagram of a welding speed control system for GTAW welding of aluminum alloys based on multiple sensors.

[0013] In the figure, 1 is an electric control cabinet, 2 is a vision sensor, 3 is a three-line structured light sensor, 4 is a sound sensor, 5 is an industrial control computer, 6 is a single-line structured light sensor, 7 is a wire feeder, 8 is a control cabinet, 9 is a camera, 10 is a camera, and 11 is a welding torch.

[0014] Figure 2Detailed installation diagrams of welding torch, single-wire structured light sensor, three-wire structured light sensor, and sound sensor.

[0015] In the diagram, 3 is a three-wire structured light sensor, 4 is a sound sensor, 6 is a single-wire structured light sensor, 9 is a camera, 10 is a camera, and 11 is a welding torch.

[0016] Figure 3 This is a schematic diagram of the weld bevel.

[0017] In the figure, (a) is a schematic diagram of excessive bevel gap, (b) is a schematic diagram of normal bevel gap, and (c) is a schematic diagram of excessive bevel gap.

[0018] Figure 4 This is a schematic diagram of the boundary of a molten pool in different states.

[0019] In the figure, (a) is a schematic diagram when the boundary of the molten pool is blurred and the welding speed is too slow; (b) is a schematic diagram when the boundary of the molten pool is a soft arc edge and the welding speed is moderate; (c) is a schematic diagram when the boundary of the molten pool is sharp and the welding speed is too fast.

[0020] Figure 5 This is a welding flowchart for a multi-sensor-based aluminum alloy GTAW welder simulation welding speed control system. Detailed Implementation

[0021] To better illustrate the technical solution and beneficial effects of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. The implementation methods of the present invention are not limited thereto.

[0022] Step 1: System Installation and Parameter Setting

[0023] Equipment Installation and Integration: Before GTAW welding begins, the installation position and threshold settings of the clamping device have a significant impact on welding quality. To address this challenge, this invention discloses an installation and setting method for a multi-sensor-based aluminum alloy GTAW welder-simulation welding speed control system. Before welding begins, a vision sensor is installed behind the welding torch, aligned at a certain angle with the molten pool and its tail area, to capture images of the molten pool. A three-line structured light sensor is installed behind the welding torch, aligned with the molten pool area, to project and capture three structured light stripes on the surface of the molten pool. A single-line structured light sensor is installed in front of the welding torch to project a single line of laser light onto the bevel to be welded in real time, measuring the bevel gap. A sound sensor is installed near the welding torch, pointing towards the arc area, to collect arc sound signals. Specific installation details are as follows... Figure 1 As shown. The intrinsic and extrinsic parameters of the vision sensor and structured light sensor are calibrated to determine the transformation relationship between image coordinates and world coordinates. During the stabilization phase at the beginning of welding, the system automatically collects data, calculates and stores the baseline values ​​of various parameters, such as the average height H of the molten pool surface. avg, average width W yavg , average tail contact angle θ avg , root mean square value of arc sound during stable welding and peak factor The empirical ranges. According to aluminum alloy materials, thicknesses, and process tests, thresholds for various criteria are preset, such as k1, k2, m1, m2, T, x1, x2, C1, C2, etc.

[0024] Step 2: Real-time control during welding

[0025] After welding starts, the system enters a closed-loop control cycle of "pre-judgment - fine-tuning - verification". It is necessary to perform real-time multi-sensor synchronous data acquisition. To address this problem, the present invention discloses a method for setting and recording a welding speed control system for an aluminum alloy GTAW welding robot based on multiple perceptions. The vision sensor takes high-frequency images of the molten pool and the rear solidification area; the three-line structured light sensor projects three laser lines synchronously and takes images to obtain a molten pool image with deformed light strips; the single-line structured light sensor continuously projects a single laser line towards the front groove and takes a light strip image to calculate the real-time gap width ΔG w ; the sound sensor continuously collects arc audio signals. The industrial control computer performs parallel data processing and analysis. The industrial control computer performs real-time processing on the collected multi-source data and extracts key features: analyzing the clarity of the molten pool boundary, using the Sobel operator to perform edge detection on the molten pool image, and calculating the average gradient magnitude G avg and the average transition zone width W avg . According to the comparison of G avg and W avg with the thresholds k1, k2, m1, m2, judge the speed state, as shown in Figure 4 . If k1 < < k2 and m1 < < m2, and the molten pool boundary image is as shown in Figure (b), then the molten pool boundary is normal; if < k1 and > m2, and the molten pool boundary image is as shown in Figure (a), the molten pool boundary is blurred; if k2 and < m1, and the molten pool boundary image is as shown in Figure (c), the molten pool boundary is sharp. Analyze the three-dimensional morphology and flow trend of the molten pool, process the three-line structured light image, and calculate the maximum height H curr of the molten pool in the current frame, the width W ycurr of a specific cross-section, and the tail contact angle θ curr . Calculate the differences ΔH, ΔW y , Δθ from the reference values. Judge the flow trend of the molten pool based on the difference trend, and then judge whether the speed is appropriate. Analyze the image of the area behind the molten pool that is in the process of solidification

[0026] Track the change in grayscale value of a specific point or region over time and calculate the absolute value of grayscale change per unit time. ,according to With threshold T Gray1 ,T Gray2 By comparing the solidification rate, the welding speed can be indirectly determined. The centerline of the light stripe is extracted from the light stripe image, and the width of the light stripe at the bevel is calculated, which is the bevel gap width G. w Calculate the difference ΔG between it and the reference gap. w ,and Comparison, specific details as follows Figure 3 As shown, if As shown in Figure (a), the bevel gap is too large, and the welding speed needs to be slowed down; if As shown in Figure (c), the bevel gap is too small, requiring a faster welding speed; the audio signal is filtered and segmented into frames. The root mean square value X of each frame is calculated. rms and peak factor C f According to X rms and C f Whether it is in a stable range determines the stability of the electric arc and the state of heat input.

[0027] Step 3: Multi-information fusion decision-making and speed control

[0028] The industrial control computer integrates the independent judgments from the five analysis modules mentioned above, employs a weighted fusion strategy to arrive at a final comprehensive judgment regarding the current welding speed, and then sends a welding speed adjustment command to the GTAW robot. The GTAW robot controller receives the command from the industrial control computer and adjusts the movement speed of its travel axis in real time, changing the movement speed of the welding torch relative to the workpiece. After the speed change, the state of the molten pool changes accordingly, all sensors immediately collect new data, and the next control cycle begins.

Claims

1. A multi-sensor-based aluminum alloy GTAW (Gas-to-Welder) welding speed control system for controlling weld formation quality, characterized in that: The aforementioned multi-sensor-based aluminum alloy GTAW simulated welder welding speed control system consists of a GTAW robot, a vision sensor, a three-wire structured light sensor, a single-wire structured light sensor, a sound sensor, and an industrial control computer. The GTAW robot is used to control the welding torch for welding. The vision sensor is used to capture images of the molten pool boundary and behind the molten pool. The three-wire structured light sensor is used to capture the three-dimensional shape of the molten pool. The single-wire structured light sensor is used to acquire bevel width information in real time. The sound sensor is used to detect arc sound. The industrial control computer is used for the analysis and processing of the acquired data.

2. The aluminum alloy GTAW welding speed control system based on multiple sensing according to claim 1, characterized in that: Aluminum alloy GTAW welding speed detection method based on molten pool edge detection; the aluminum alloy GTAW welding speed detection method of molten pool edge detection obtains the molten pool image through a vision sensor and judges the welding state based on the clarity of the molten pool boundary; the clarity of the molten pool boundary refers to the clarity of the dividing line between the liquid molten pool and the surrounding solid base material; the average gradient amplitude of the molten pool boundary region is obtained by convolving the sobel operator with the original image, and the higher the value, the sharper the molten pool boundary; the average gradient amplitude of the molten pool boundary region is ; the refers to the average gradient amplitude within the molten pool boundary region, the N refers to the total number of pixels within the boundary region, and the refers to the gradient amplitude of the i-th pixel point within the molten pool boundary region, which is obtained by dividing the sum of all gradient values within the molten pool boundary region by the total number of pixels; multiple normal lines are calculated within the selected molten pool boundary region to obtain the average transition zone width value, and the higher the value, the softer the molten pool boundary; the average transition zone width value is ; the refers to the average transition zone width of the molten pool boundary region. By searching both inside and outside the molten pool along the direction perpendicular to the boundary (i.e., the gradient direction) for each boundary pixel point until the gray value stabilizes (i.e., the gray difference from the inside of the molten pool is less than the threshold), the number of these pixels is recorded, which is the transition zone width of this point; the refers to the single transition zone width of the i-th pixel point within the molten pool boundary region; if k2 and <m1, that is, the average gradient amplitude is on the high side and the average transition zone width value is on the low side, it is determined that the molten pool boundary is sharp and serrated, and the welding speed is fast. At this time, the speed needs to be reduced, and a slight pause of 0.5 - 1 s can be made for adjustment to ensure full fusion of the molten pool and the base material; if <k1 and >m2, that is, the average gradient amplitude is on the low side and the average transition zone width value is on the high side, it is determined that the molten pool boundary is blurred and the welding speed is slow. At this time, the welding speed needs to be gradually increased by a range of 5 - 10 mm / min until the boundary becomes clear; if k1 < <k2 and m1 < <m2, that is, the average gradient amplitude and the average transition zone width value are moderate, it is determined that the clarity of the molten pool boundary is in a normal state, and the transition with the base material is natural, and it is determined that the heat input and the welding speed are matched; the k1 and k2 respectively refer to the minimum preset threshold and the maximum preset threshold of the average gradient amplitude; the m1 and m2 respectively refer to the minimum preset threshold and the maximum preset threshold of the average transition zone width value.

3. The aluminum alloy GTAW welding speed control system based on multiple sensing according to claim 1, characterized in that: A welder speed detection method based on GTAW (Glass Surface Arrangement) for aluminum alloys is used. A three-line structured light sensor captures the three-dimensional morphology of the molten pool to identify its flow trend. The captured 3D morphology includes molten pool surface height, width, and tail contact angle information. The three-line structured light sensor projects three structured beams onto the molten pool surface, while a camera simultaneously captures images of the molten pool including the beams. The original molten pool image is then processed using the Sobel operator to obtain a precise molten pool boundary. Based on a structured light 3D reconstruction algorithm, the 3D coordinates of all pixels within the Sobel-bounded molten pool region are calculated, with the height direction as the Z-axis, the welding direction as the Y-axis, and the perpendicular direction to the welding direction as the X-axis. The difference between the highest point of the Z-axis on the molten pool surface image and the reference plane is taken as the maximum height of the molten pool surface. Obtained; The stated The maximum height of the molten pool surface. Z is the highest Z-axis coordinate value within the molten pool region, and Z0 is the coordinate value of the base material reference plane. The left and right boundaries of the molten pool are located using the Sobel operator. The distance between two points on the boundary in the perpendicular welding direction is taken as the width. Let the coordinates of the left boundary point of the molten pool be (X...). L The coordinates of the right boundary point are (X, y), R If ,y), then the cross-sectional width W y Through W y =X R X L The contact point between the molten pool tail and the base material in the locating molten pool boundary region is calculated. The angle between the two lines obtained by fitting the molten pool tail surface curve and the base material plane at the contact point is the molten pool tail contact angle. Calculated; the aforementioned The normal vector of the molten pool surface. The normal vector of the base material plane is taken; the average value of the parameters in 10 consecutive frames during the welding stabilization phase is used as the reference value, i.e., B. H B W B θ The aforementioned , , The aforementioned , , This refers to the average height, average width, and average contact angle of the molten pool surface; calculate the current frame parameter H. curr W ycurr θ curr Calculate the difference between it and the benchmark value ΔH=H curr B H ΔW y =W curr B W Δθ=θ curr B θ If ΔH>T H ΔW y >T W Δθ>T θ T is a preset threshold. If the condition is met for 3 consecutive frames, the determination parameter is increased. In the flat welding state, if the system detects that the height and width of the molten pool have increased and the contact angle between the tail of the molten pool and the base material has increased, it is determined that the liquid metal on both sides of the molten pool is shrinking towards the middle. That is, it is determined that the welding speed is too fast and the molten pool has not spread sufficiently. At this time, the welding speed needs to be reduced to allow the liquid metal to flow towards the edge. If the height of the molten pool increases abnormally, the width becomes narrower, and the tail of the molten pool protrudes, it is determined that the middle of the molten pool is bulging. That is, it is determined that the welding speed is too slow, the molten pool exists for too long, and the metal is excessively melted. At this time, the welding speed needs to be increased to avoid the weld seam from piling up.

4. The aluminum alloy GTAW welding speed control system based on multiple sensing according to claim 1, characterized in that: A method for detecting the welding speed of aluminum alloy GTAW (Glass-Only Welder) based on the solidification rate of the molten pool is proposed. This method utilizes a visual sensor to acquire high-frequency images of the molten pool behind it. After preprocessing the images to eliminate interference, an edge detection algorithm is used to locate the molten pool boundary and obtain pixel coordinates. The solidification rate is determined by the change in image grayscale values. The image preprocessing involves removing image noise and enhancing the grayscale difference between the molten pool and the background to ensure clear boundaries. The edge detection algorithm uses the Sobel operator to extract continuous and complete molten pool boundaries. The image grayscale values ​​are obtained using the formula Gray = 0.299*R + 0.587*G + 0.114*B, where R, G, and B represent the brightness values ​​of the red, green, and blue channels, respectively. Higher temperatures result in lower grayscale values. The change in grayscale value per unit time is ΔGray. The unit time refers to the time difference between the molten pool's liquid state and its semi-solidified state. This indicates that the temperature drop per unit time is moderate, and the welding speed is moderate; if This indicates a large temperature drop per unit time and a rapid solidification rate of the molten pool, suggesting that the welding speed is too fast. In this case, the welding speed needs to be reduced to extend the heat input time and prevent incomplete penetration at the weld root. If the temperature drop per unit time is small and the solidification rate of the molten pool is slow, then the welding speed is too slow. In this case, the welding speed needs to be increased to prevent the heat-affected zone from becoming too large.

5. The aluminum alloy GTAW welding speed control system based on multiple sensing according to claim 1, characterized in that: A welding speed detection method for aluminum alloy GTAW (Glass Surface Artificial Welding) based on dynamic detection of weld bevel changes is proposed. The dynamic detection of weld bevel changes involves real-time detection of the bevel gap width, utilizing a front-mounted single-line structured light sensor to acquire bevel width information in real time. The dynamic adaptation of bevel changes occurs when the bevel gap width increases. When the weld pool fills the gap, the welding speed needs to be slowed down while increasing the filler wire amount. Once the weld pool has filled the gap, the original welding speed should be resumed to ensure the weld pool fully melts the base material and prevents the weld pool from widening. When the bevel gap decreases... At this time, it is necessary to increase the welding speed and avoid localized pile-up; the aforementioned This refers to the difference between the current actual gap and the reference value. This refers to the preset threshold for the difference between the actual gap and the reference value.

6. The aluminum alloy GTAW welding speed control system based on multiple sensing according to claim 1, characterized in that: A method for detecting the welding speed of aluminum alloy GTAW (Glass Wire Arcade Welding) based on arc sound feedback is presented. This method utilizes a sound sensor to collect arc sound signals in real time, performs preprocessing, feature extraction, and time-domain analysis. The time-domain analysis refers to analyzing the root mean square (RMS) value and peak factor. The RMS value represents the average amplitude intensity of the sound signal within the calculation time period, and its calculation formula is as follows: = The aforementioned This refers to the voltage value of the sound signal; when At this time, the root mean square value is relatively stable and changes gradually, indicating a moderate welding speed; when If the welding heat input is too low or the welding speed is too high, the welding speed needs to be slowed down; when > At that time, if the welding heat input is too high and the welding speed is too slow, the welding speed needs to be increased; , These are the minimum and maximum preset thresholds for the root mean square value, respectively; the peak factor refers to the ratio of the signal peak value to the effective value, and its calculation formula is as follows: The aforementioned This refers to the maximum absolute value of the signal within the analysis frame. This refers to the root mean square value; when At that time, the sound waveform amplitude is uniform, and the welding speed is moderate; when At this time, large amplitude fluctuations and increased high-frequency components indicate that the welding speed is too fast and needs to be slowed down; when When the amplitude fluctuation is small and the overall signal strength is low, it is determined that the welding speed is too slow, and the welding speed needs to be increased and the arc length shortened.

7. The aluminum alloy GTAW welding speed control system based on multiple sensing according to claim 1, characterized in that: In the welding speed control system of aluminum alloy GTAW welding robot based on multi-sensing, the weld forming quality is controlled by the welding speed control method of aluminum alloy GTAW welding robot based on multi-sensing; the welding speed control method of aluminum alloy GTAW welding robot based on multi-sensing realizes the control of welding speed through GTAW welding robot; a vision sensor, a three-line structured light sensor, a single-line structured light sensor, a sound sensor and an industrial control computer are used to detect the average gradient amplitude of the molten pool boundary area , the average width value of the transition zone , the difference ΔH between the molten pool height and the reference value, the difference ΔW between the molten pool width and the reference value y , the difference Δθ between the contact angle at the tail of the molten pool and the reference value, the change value of the gray value per unit time , , the root mean square value of the sound signal and the peak factor ; if k1 k2 and m1 m2, , , , it means that the welding speed is moderate and the weld forming quality is good; if k2 and m1,​​​​​​​​​​​​​​​​​​​​​​

Images