New energy battery aluminum wire welding quality detection method and device

Abstract

The present application relates to the technical field of welding quality detection, and particularly relates to a new energy battery aluminum wire welding quality detection method, comprising: obtaining a brightness map and a height map of a target to be detected; using a target recognition model to recognize a welding point and a first welding arc from the brightness map to obtain a welding point position and a first welding arc position; according to the brightness map, the height map, the welding point position and the first welding arc position, obtaining a reference surface corresponding to a first solder pad and a second reference surface corresponding to a second solder pad, measuring a first welding arc height, measuring a second welding arc height, measuring a welding point height, and measuring a welding point width; if the first welding arc height, the second welding arc height, the welding point width and the welding point height are within the corresponding standard range, respectively, then the welding quality is determined to be qualified, otherwise, the welding quality is determined to be unqualified; without human participation, human error is eliminated, and the detection efficiency and accuracy are improved.

Description

Technical Field

[0001] This invention relates to the field of welding quality inspection technology, and in particular to a method and apparatus for inspecting the welding quality of aluminum wires for new energy batteries. Background Technology

[0002] The application fields of new energy power batteries are becoming increasingly widespread, not only in new energy vehicles, but also in grid energy storage and home energy storage. The welding part of aluminum wire in the battery has quality inspection requirements. In the past, the conventional inspection method was manual inspection, but manual inspection is prone to errors and is also affected by subjectivity, resulting in low efficiency.

[0003] In the fully automatic ultrasonic aluminum wire welding machine welding quality inspection system disclosed in application number CN104020222A, the inspection process is automated by acquiring an image of the target to be inspected and then processing the image according to the texture and contour of the image, which improves the inspection efficiency. However, the accuracy of the inspection results is not high due to the low accuracy of image recognition.

[0004] Therefore, this invention proposes a method and device for detecting the welding quality of aluminum wires in new energy batteries. Summary of the Invention

[0005] The purpose of this invention is to provide a method for detecting the welding quality of aluminum wires in new energy batteries, addressing the shortcomings of existing technologies. This method aims to solve the problem of low accuracy in detection results due to the low precision of image recognition in existing technologies.

[0006] The present invention achieves the above objectives through the following technical solutions:

[0007] A method for inspecting the welding quality of aluminum wire in new energy batteries, comprising:

[0008] Preparation steps: Obtain the brightness map and height map of the target to be detected;

[0009] Identification steps: The solder joint and the first welding arc are identified from the brightness map using a target recognition model to obtain the positions of the solder joint and the first welding arc, respectively;

[0010] Measurement steps: Based on the brightness map, the height map, the solder joint position and the first welding arc position, obtain the first reference plane corresponding to the first solder pad and the second reference plane corresponding to the second solder pad, and measure the first welding arc height, the second welding arc height, the solder joint height and the solder joint width;

[0011] Quality judgment steps: If the height of the first welding arc, the height of the second welding arc, the width of the weld spot, and the height of the weld spot are all within the corresponding standard ranges, the welding quality is judged to be qualified; otherwise, the welding quality is judged to be unqualified.

[0012] Furthermore, the structure of the target recognition model includes:

[0013] The system comprises convolutional layers, pooling layers, activation functions, and a Fire module. The Fire module includes a compression layer and a dilation layer. The compression layer uses a 1×1 convolutional kernel to compress the feature map to obtain a compressed feature map. The dilation layer dilates the compressed feature map by using a 1×1 convolutional kernel and a 3×3 convolutional kernel.

[0014] Furthermore, the target recognition model is pre-trained before performing the recognition step, and the pre-training process includes:

[0015] Image annotation sub-steps: In the first predetermined number of brightness images, the solder joints are individually annotated to obtain a solder joint image set; in the second predetermined number of brightness images, the first welding arc is individually annotated to obtain a first welding arc image set.

[0016] Image processing sub-step: Input the weld point image set and the first welding arc image set into the target recognition model to obtain the output result;

[0017] The parameter update sub-step involves calculating the error of the output result. If the error is greater than a predetermined error, the model parameters are updated based on the error. After the update, the image processing step is performed again. If the error is less than the predetermined error, the training is complete.

[0018] Furthermore, the process of obtaining the first reference plane corresponding to the first pad and the second reference plane corresponding to the second pad in the measurement step includes:

[0019] Sub-step for segmenting the pad region: Set a segmentation threshold based on the pixel values ​​of the first pad and the second pad, and segment the first pad region and the second pad region in the brightness map according to the segmentation threshold;

[0020] The sub-step for obtaining the height value set is as follows: Obtain the first height value set and the second height value set corresponding to the first pad area and the second pad area from the height map;

[0021] The sub-step for selecting the set of valid values ​​is as follows: sort the height values ​​in the first set of height values ​​and the second set of height values ​​in ascending order, and then take the height values ​​of the first predetermined ascending interval to obtain the first set of valid values ​​and the second set of valid values.

[0022] Plane fitting sub-step: Perform plane fitting on the first set of effective values ​​and the second set of effective values ​​respectively to obtain the first reference plane and the second reference plane.

[0023] Furthermore, the planar fitting sub-step includes:

[0024] Let the equation of the plane be: Ax + By + Cz + D = 0, and let the constraint equation be: A 2 +B2 +C 2 =1,

[0025] When the plane equation represents the first reference plane, the coefficients A, B, C, and D satisfy the condition that the sum of the squares of the distances from the points in the first set of effective values ​​to the first reference plane is minimized.

[0026] The sum of the squares of the distances from the points in the first set of valid values ​​to the first reference plane is denoted by o1, and satisfies the following formula:

[0027]

[0028] Among them, is D 1i It is any point (x) in the first set of valid values. i ,y i ,z i The distance from the first reference plane;

[0029] When the plane equation represents the second reference plane, the coefficients A, B, C, and D satisfy the condition that the sum of the squares of the distances from the points in the second set of effective values ​​to the second reference plane is minimized.

[0030] The sum of the squares of the distances from the points in the second set of valid values ​​to the second reference plane is denoted by o2 and satisfies the following formula:

[0031]

[0032] Among them, D 2i It is any point (x) in the second set of valid values. i ,y i ,z i The distance from the second reference plane.

[0033] Furthermore, the process of measuring the second welding arc height in the measurement step includes:

[0034] In the height diagram, the reference plane that is higher between the first reference plane and the second reference plane is selected as the second welding arc reference plane;

[0035] In the height map, height values ​​above the first predetermined height range of the second welding arc reference plane are selected and sorted in ascending order to obtain the second welding arc height dataset. The average value of the height values ​​in the second predetermined ascending interval of the second welding arc height dataset is calculated as the second welding arc height.

[0036] Furthermore, the process of measuring the height of the first welding arc in the measurement step includes:

[0037] Step for selecting the first welding arc reference surface: Determine whether the planar area corresponding to the first reference surface in the brightness diagram contains the position of the first welding arc. If so, the first reference surface is used as the first welding arc reference surface; otherwise, the second reference surface is used as the first welding arc reference surface.

[0038] Sub-step for dividing the first welding arc region: Select a region with a first preset pixel area as the first welding arc region in the height map, centered on the first welding arc position;

[0039] The sub-step for calculating the first welding arc height is as follows: Select height values ​​within a second predetermined height range above the first welding arc reference plane in the first welding arc region and sort them in ascending order to obtain the first welding arc height dataset. Calculate the average value of the height values ​​in the third predetermined ascending interval of the first welding arc height dataset as the first welding arc height.

[0040] Furthermore, the process of measuring the height of the solder joint in the above measurement steps includes:

[0041] Step for selecting the solder joint reference surface: Determine whether the plane area corresponding to the first reference surface in the brightness map contains the solder joint position. If so, use the first reference surface as the solder joint reference surface; otherwise, use the second reference surface as the solder joint reference surface.

[0042] Sub-step for dividing the solder joint area: Select an area with a second preset pixel area as the solder joint area in the height map, centered on the location of the solder joint;

[0043] The sub-step for calculating solder joint height is as follows: Select height values ​​within a third predetermined height range above the solder joint reference plane in the solder joint region and sort them in ascending order to obtain a solder joint height dataset. Select the height values ​​of the fourth predetermined ascending interval in the solder joint height dataset and calculate the average value as the solder joint height.

[0044] Furthermore, the process of measuring the weld joint width in the measurement step includes:

[0045] Image segmentation step: The SAM algorithm is used to segment the solder joint locations to obtain solder joint images;

[0046] Step to obtain pixel length ratio: Obtain the pixel length ratio corresponding to the solder joint position;

[0047] Steps for obtaining solder joint width: The solder joint width is obtained based on the ratio of the width pixels of the solder joint image to the pixel length.

[0048] A new energy battery aluminum wire welding quality inspection device, applied to the above-mentioned new energy battery aluminum wire welding quality inspection method, includes:

[0049] Preparatory module: Acquires the brightness and height maps of the target to be detected;

[0050] Recognition module: The target recognition model is used to identify the solder joint and the first welding arc from the brightness map to obtain the position of the solder joint and the position of the first welding arc, respectively;

[0051] Measurement module: Based on the height map, the brightness map, the solder joint position and the first welding arc position, obtain the reference plane corresponding to the first solder pad and the second reference plane corresponding to the second solder pad, measure the height of the first welding arc, measure the height of the second welding arc, measure the height of the solder joint, and measure the width of the solder joint;

[0052] Quality assessment module: If the first welding arc height, the second welding arc height, the weld width, and the weld height are all within the corresponding standard ranges, the welding quality is deemed qualified; otherwise, the welding quality is deemed unqualified.

[0053] The beneficial effects of this invention are: by using deep learning algorithms to learn the weld points and welding arcs, the positions of the weld points and welding arcs in the target to be detected are intelligently identified; the height of the welding arc is calculated according to the designed calculation program; and the width of the weld point is calculated based on the obtained pixel length ratio after the image is segmented using an image segmentation algorithm. This eliminates human error without the need for manual intervention and improves detection efficiency and accuracy. Attached Figure Description

[0054] Figure 1 This is a flowchart of a method for inspecting the welding quality of aluminum wires in new energy batteries.

[0055] Figure 2 This is a schematic diagram of a device for detecting the welding quality of aluminum wire in new energy batteries.

[0056] Figure 3 This is a photograph of aluminum wire welding in a new energy battery.

[0057] Figure label:

[0058] 1— Solder joint;

[0059] 2—First welding arc;

[0060] 3—Second welding arc;

[0061] 4—First pad;

[0062] 5 - Second pad. Detailed Implementation

[0063] The present invention will now be described in detail with reference to the accompanying drawings.

[0064] like Figure 3 The image shown is a photograph of the actual aluminum wire welding process in a new energy battery.

[0065] like Figure 1The diagram shows a flowchart of a method for inspecting the welding quality of aluminum wire in new energy batteries, including: preparation steps, identification steps, obtaining a reference surface step, measurement steps, and quality judgment steps. This method uses a convolutional neural network model to learn from an image labeled with weld point 1 and welding arc, obtaining an identification model that can automatically identify the positions of weld point 1 and welding arc. The identification model then identifies the target to be inspected, obtaining the positions of weld point 1 and welding arc. A reference surface is obtained based on the area where the welding pad is located. The heights of weld point 1, the second welding arc 3, and the first welding arc 2 are calculated based on the reference surface and a predetermined calculation program. Finally, the results are compared with standard ranges. If the measurement results are not within the standard range, it indicates a welding quality problem. By using a deep learning model to inspect the welding quality of aluminum wire in new energy batteries, the method achieves the technical effects of improving inspection efficiency and eliminating human error.

[0066] Identification steps: The target recognition model is used to identify solder joint 1 and first welding arc 2 from the brightness map to obtain the position of solder joint and the position of first welding arc respectively;

[0067] In another embodiment, the target recognition model is a lightweight convolutional neural network model; the convolutional neural network model includes convolutional layers, pooling layers, activation functions, and a fire module. The fire module includes a compression layer and a dilation layer. The compression layer uses a 1×1 convolutional kernel to compress the feature map to obtain a compressed feature map. The dilation layer dilates the compressed feature map by using a 1×1 convolutional kernel and a 3×3 convolutional kernel.

[0068] Specifically, the convolutional neural network model used in this embodiment is more concise and accurate than the traditional convolutional neural network model. By adding a Fire module between the convolutional layer and the pooling layer to compress and then expand the dimension of the feature map, the weight parameters are reduced, giving the network width. This greatly improves the running speed without reducing the model accuracy.

[0069] In one embodiment, the training process of the target recognition model includes: an image annotation step: separately annotating solder joint 1 in a predetermined number of brightness images to obtain a solder joint image set, and separately annotating the first welding arc 2 in a predetermined number of brightness images to obtain a first welding arc image set; an image processing step: inputting the solder joint image set and the first welding arc image set into the target recognition model to obtain an output result; a parameter update step: calculating the error of the output result, and if the error is greater than a predetermined error, updating the parameters of the model according to the error, and after updating, performing the image processing step again, and if the error is less than the predetermined error, the training is complete. Specifically, 3,000 aluminum wire welding images labeled with weld point 1 are input into a convolutional neural network model. The convolutional neural network model extracts the feature map of weld point 1 and obtains the corresponding feature vector through the processes of convolution, pooling, compression and dilation, which is stored in the network nodes. The parameters in the convolutional neural network are continuously corrected and improved based on the learning results. 3,000 aluminum wire welding images labeled with the first welding arc 2 are input into another convolutional neural network model and the same process is performed, thereby training a weld point 1 recognition model and a first welding arc 2 recognition model respectively.

[0070] Steps for obtaining the reference plane: Obtain the first reference plane corresponding to the first pad 4 and the second reference plane corresponding to the second pad 5 from the brightness map and the height map.

[0071] In another embodiment, the step of obtaining the reference plane further includes: a step of segmenting the pad region, a step of obtaining a set of height values, a step of selecting a set of valid values, and a step of plane fitting; specifically, a segmentation threshold is set according to the pixel values ​​of the first pad 4 and the second pad 5, and the first pad region and the second pad region are segmented in the brightness map according to the segmentation threshold; a first set of height values ​​and a second set of height values ​​corresponding to the first pad region and the second pad region are obtained from the height map; the height values ​​in the first set of height values ​​and the second set of height values ​​are sorted in ascending order, and the height values ​​of a first predetermined ascending interval are taken to obtain a first set of valid values ​​and a second set of valid values; plane fitting is performed on the first set of valid values ​​and the second set of valid values ​​to obtain a first reference plane and a second reference plane.

[0072] The plane fitting process for the first and second reference planes is the same, including: setting the plane equation as: Ax + By + Cz + D = 0, and setting the constraint equation as: A 2 +B 2 +C 2=1, the coefficients A, B, C, D in the plane equation corresponding to the first reference plane satisfy the condition: the sum of the squares of the distances from the points in the first set of effective values ​​to the first reference plane is minimized; the coefficients A, B, C, D in the plane equation corresponding to the second reference plane satisfy the condition: the sum of the squares of the distances from the points in the second set of effective values ​​to the second reference plane is minimized, expressed by the formula: Where D i It is any point (x) in the first set of valid values. i ,y i ,z i The distance from the first reference plane, or the distance from any point (x) in the second set of valid values. i ,y i ,z i The distance from the second reference plane.

[0073] Specifically, the pixel values ​​of the pads differ significantly from those of other areas. The pads are located at the center of the target and occupy a large area. In the brightness map, the dense, large-scale area with similar pixel values ​​represents the pad region. Pixel value sets are obtained at the corresponding positions in the height map. Based on the differences in height values, these sets are divided into a first height value set and a second height value set. The first and second height value sets are then combined and sorted in ascending order. Height values ​​within the ascending range of 40%-60% are selected as the first and second effective value sets, respectively. Plane fitting is then performed on both effective value sets to obtain two reference planes. The plane fitting process is the process of finding the most suitable values ​​for A, B, C, and D. The coefficients of the plane fitting equation and the constraint equation are based on... It is understandable that when every point is on the same plane, the sum of the squares of the distances from each point to the plane is minimized. Ideally, all points in the first set of effective values ​​should be on the first reference plane, allowing us to solve for the corresponding plane equation. Similarly, all points in the second set of effective values ​​should be on the second reference plane, allowing us to solve for the corresponding plane equation. The solution principle is as follows:

[0074] The average coordinates of the first set of effective values ​​or the second set of effective values ​​are: And it satisfies equation one:

[0075] Subtracting equation one from the plane fitting equation yields equation two: Let matrix Let matrix Equation 1 can be transformed into MN = 0. In this case, all points in the effective value set are on the plane, and MN = 0 holds true. However, in reality, some points are outside the plane. The purpose of plane fitting is to minimize the sum of the distances from the plane to all points. The objective here is to minimize the product of MN, with the constraint ||N|| = 1. Performing singular value decomposition on M: M = UD1V T Then we have ||MN||=||UD1V T N||=||D1V T N||, where: V T N is a column matrix, and ||V T N||=||N||=1, because the diagonal elements of D1 are singular values. Let the last diagonal element be the smallest singular value, then it is true if and only if: At that time, ||MN||=||UD1V T N||=||D1V T N|| can reach its minimum value, that is, MN is minimized. At this point, we have:

[0076]

[0077] The optimal solution where MN is minimized under the constraint ||N||=1 is: N=(A,B,C)=(v n,1 ,v n,2 ,v n,3 Therefore, the minimum value of o is the minimum eigenvalue of matrix M, and the corresponding eigenvectors are the plane parameters A, B, and C. D can be obtained using the centroid, and the fitted plane equation can be obtained. By combining the pixel value and the height value, the approximate areas of the first pad 4 and the second pad 5 are first segmented, and then points within a predetermined range are extracted and plane fitting is performed to obtain the final first reference plane and the second reference plane. In this way, the areas corresponding to the first pad 4 and the second pad 5 in the target to be detected are accurately transformed into two plane equations, and the concrete pads are transformed into abstract mathematical representations, which facilitates the subsequent calculation of the second arc height, the first arc height, and the solder joint height, and ensures accuracy.

[0078] Measurement steps: The weld joint position is image segmented and measured to obtain the weld joint width; the first weld arc height is obtained from the height map based on the first weld arc position; and the second weld arc height is obtained from the height map based on the first reference plane and the second reference plane.

[0079] In another embodiment, the higher of the first and second reference planes in the height map is selected as the second welding arc reference plane. Height values ​​above a first predetermined height range above the second welding arc reference plane are selected in the height map and sorted in ascending order to obtain a second welding arc height dataset. The average value of the height values ​​in the second predetermined ascending interval of the second welding arc height dataset is calculated as the second welding arc height. Specifically, the first predetermined height range is [H+1mm, H+6mm], and the second predetermined ascending interval is 97%–98%. The 97%–100% interval is not used to eliminate interference from noise at the top of the second welding arc 3 of the aluminum wire. By filtering out most of the interfering height values ​​and then sorting the height values, the average value of the 97%–98% ascending interval is taken as the second welding arc height. This not only detects the height of the second welding arc 3 but also removes noise interference at the top of the second welding arc 3, making the detection results more accurate.

[0080] In another embodiment, the following steps are included: Selecting a reference plane: Determining whether the first welding arc position is within the first reference plane; if so, using the first reference plane as the first welding arc reference plane; otherwise, using the second reference plane as the first welding arc reference plane. Dividing the first welding arc region: Selecting a region with a first preset pixel area as the center in the height map, centered on the first welding arc position, as the first welding arc region. Calculating the first welding arc height: Selecting height values ​​within the first welding arc region that are higher than a second predetermined height range of the first welding arc reference plane and sorting them in ascending order to obtain a first welding arc height dataset; calculating the average value of the height values ​​in the third predetermined ascending interval of the first welding arc height dataset as the first welding arc height. Specifically, since the aluminum wire may be skewed in the horizontal direction, causing the first welding arc 2 to deviate in the horizontal direction, the first welding arc reference plane is set as a rectangular ROI centered on the position of the first welding arc, with a width semi-axis of 60 pixels and a height semi-axis of 20 pixels; the second predetermined height range is [h+0.15mm, h+0.75mm], and the third predetermined ascending interval is 90% to 95%. By first obtaining the position of the first welding arc, and then determining which pad the first welding arc 2 is on, the reference plane corresponding to that pad is used as the first welding arc reference plane, and then adding a certain range of height values ​​upwards to obtain a set, and then sorting the set in ascending order and taking the average of the 90% to 95% of the ascending height values, the height of the first welding arc is accurately obtained, and possible interference factors in the target to be detected are eliminated.

[0081] In another embodiment, further, the process of measuring the solder joint height in the measurement step comprises: a reference surface selecting sub-step: determining whether the position of the solder joint is in the first reference surface, if yes, taking the first reference surface as the solder joint reference surface, otherwise, taking the second reference surface as the solder joint reference surface; a solder joint area dividing sub-step: selecting an area with a second preset pixel area in the height map centered on the solder joint position as a solder joint area; a solder joint height calculating sub-step: selecting height values that are higher than the solder joint reference surface within a third predetermined height range in the solder joint area, sorting the height values in ascending order to obtain a solder joint height data set, selecting height values in a fourth predetermined ascending interval from the solder joint height data set, and calculating an average value as the solder joint height. Specifically, the position of the solder joint is obtained by identifying the target to be detected through a solder joint 1 identification model, then a range is set centered on the solder joint 1 according to the reference surface corresponding to the pad where the solder joint 1 is located, that is, the solder joint area, which is a rectangular ROI with a semi-major axis of 60 pixels and a semi-minor axis of 15 pixels. Height values that are higher than the corresponding reference surface within a certain range are selected from the solder joint area to obtain a set, and the range is the third predetermined height range: [h+0.1mm, h+0.35mm]. After sorting the height values in the set in ascending order, the height values in the ascending interval of 40% to 60% are selected, and the average value is calculated as the solder joint height; thus the solder joint height is detected intelligently, which improves the efficiency and eliminates manual errors.

[0082] In another embodiment, the SAM algorithm is used to segment the solder joint position to obtain a solder joint image; obtaining the pixel length ratio corresponding to the position of the solder joint; obtaining the width of the solder joint according to the width pixel of the solder joint image and the pixel length ratio. Specifically, the SAM algorithm is the segment-anything-model, which is used to segment the solder joint 1 from the target to be detected along the contour of the solder joint 1 at the solder joint position, then the size of the solder joint 1 is obtained by dividing the pixels of the segmentation result by the pixel length ratio, the pixel length ratio is the pixel length ratio of the target to be detected, the pixels are pixels of a height map or a brightness map of the target to be detected, and the length is the actual length, thereby detecting the width of the solder joint 1, which has the technical effect of being accurate and efficient.

[0083] Quality determination step: comparing the solder joint width, the solder joint height, the first weld arc height and the second weld arc height with corresponding standard ranges respectively to obtain comparison results, and determining whether the welding quality is qualified according to the comparison results. Specifically, according to the quality standard for aluminum wire welding of new energy batteries, a standard width range for the solder joint 1 is set, a standard height range for the solder joint 1, a standard height range for the first weld arc, and a standard height range for the second weld arc are set, and the measurement results are compared with the above standard ranges. If the measurement result is not within the standard range, it indicates that there is a quality problem in the welding.

[0084] As Figure 2The diagram shows a structural schematic of a welding quality inspection device for aluminum wire in new energy batteries, applied to the aforementioned welding quality inspection method for aluminum wire in new energy batteries. The device includes: a brightness map acquisition module, a height map acquisition module, a target recognition module, and a quality inspection module. Specifically, the brightness map acquisition module and the height map acquisition module are a 2D area array camera and a 3D laser contour sensor, respectively. The target recognition module is an embedded pre-trained convolutional neural network model, capable of recognizing and extracting features from the images transmitted by the 2D area array camera and the 3D laser contour sensor. The quality inspection module includes calculation programs for the second welding arc height, the first welding arc height, the weld point width, and the weld point height, as well as a comparison program with standard ranges. This device, applied to the aforementioned welding quality inspection method for aluminum wire in new energy batteries, intelligently, accurately, and efficiently determines whether welding quality problems exist.

[0085] In one embodiment, the brightness map acquisition module is used to acquire a planar brightness map of the target to be detected; specifically, a 2D area scan camera is used to acquire the brightness map of the target to be detected. The height map acquisition module is used to acquire a height map that matches the planar brightness map; specifically, a 3D laser contour sensor is used to acquire the height map of the target to be detected. The target recognition module uses a pre-trained lightweight convolutional neural network to identify the positions of weld point 1 and the first welding arc 2 from the planar brightness map; specifically, the convolutional neural network is trained with 3000 images labeled with weld point 1 and 3000 images labeled with the first welding arc 2, respectively, and the target to be detected is recognized after training. The quality detection module is used to measure the height of the second welding arc, the height of the first welding arc, the height of the weld point, and the width of the weld point according to a predetermined program, and compare them with the corresponding standard range to obtain the quality detection result. Specifically, the calculation process is set by the program, which obtains input data from the brightness map, the height map, and the recognition result of the convolutional neural network. After the program is executed, the detection result is obtained, thereby achieving the technical effect of intelligent automatic detection of the welding quality of aluminum wire in new energy batteries, which is highly efficient and eliminates human error.

[0086] In summary, the present invention possesses the excellent characteristics described above, which enhances its effectiveness in use compared to previous technologies, making it a highly practical product.

[0087] The above description is only a preferred embodiment of the present invention. For those skilled in the art, there will be changes in the specific implementation and application scope based on the ideas of the present invention. The content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for inspecting the welding quality of aluminum wire in new energy batteries, characterized in that, include: Preparation steps: Obtain the brightness map and height map of the target to be detected; Identification steps: The solder joint and the first welding arc are identified from the brightness map using a target recognition model to obtain the positions of the solder joint and the first welding arc, respectively; Measurement steps: Based on the brightness map, the height map, the solder joint position and the first welding arc position, obtain the first reference plane corresponding to the first solder pad and the second reference plane corresponding to the second solder pad, and measure the first welding arc height, the second welding arc height, the solder joint height and the solder joint width; Quality judgment steps: If the height of the first welding arc, the height of the second welding arc, the height of the weld point, and the width of the weld point are all within the corresponding standard ranges, the welding quality is judged to be qualified; otherwise, the welding quality is judged to be unqualified. The process of obtaining the first reference plane corresponding to the first pad and the second reference plane corresponding to the second pad in the measurement step includes: Sub-step for segmenting the pad region: Set a segmentation threshold based on the pixel values ​​of the first pad and the second pad, and segment the first pad region and the second pad region in the brightness map according to the segmentation threshold; The sub-step for obtaining the height value set is as follows: Obtain the first height value set and the second height value set corresponding to the first pad area and the second pad area from the height map; The sub-step for selecting the set of valid values ​​is as follows: sort the height values ​​in the first set of height values ​​and the second set of height values ​​in ascending order, and then take the height values ​​of the first predetermined ascending interval to obtain the first set of valid values ​​and the second set of valid values. Plane fitting sub-step: Perform plane fitting on the first set of effective values ​​and the second set of effective values ​​respectively to obtain the first reference plane and the second reference plane; The process of measuring the height of the second welding arc in the measurement step includes: Step for selecting a reference surface: Determine whether the height value of the first reference surface is greater than the height value of the second reference surface. If so, select the first reference surface as the second welding arc reference surface; otherwise, select the second reference surface as the second welding arc reference surface. The sub-step for calculating the second welding arc height is as follows: Select the height values ​​in the height map that are higher than the first predetermined height range of the second welding arc reference plane and sort them in ascending order to obtain the second welding arc height dataset. Select the height values ​​in the second predetermined ascending interval of the second welding arc height dataset and calculate the average value as the second welding arc height. The process of measuring the height of the first welding arc in the measurement step includes: Step for selecting the first welding arc reference surface: Determine whether the planar area corresponding to the first reference surface in the brightness diagram contains the position of the first welding arc. If so, the first reference surface is used as the first welding arc reference surface; otherwise, the second reference surface is used as the first welding arc reference surface. Sub-step for dividing the first welding arc region: Select a region with a first preset pixel area as the first welding arc region in the height map, centered on the first welding arc position; The sub-step for calculating the first welding arc height is as follows: Select the height values ​​in the first welding arc region that are higher than the second predetermined height range of the first welding arc reference plane and sort them in ascending order to obtain the first welding arc height dataset. Select the height values ​​in the third predetermined ascending interval of the first welding arc height dataset and calculate the average value as the first welding arc height. The process of measuring the height of the solder joint in the measurement step includes: Step for selecting the solder joint reference surface: Determine whether the plane area corresponding to the first reference surface in the brightness map contains the solder joint position. If so, use the first reference surface as the solder joint reference surface; otherwise, use the second reference surface as the solder joint reference surface. Sub-step for dividing the solder joint area: Select an area with a second preset pixel area as the solder joint area in the height map, centered on the location of the solder joint; The sub-step for calculating solder joint height is as follows: Select height values ​​within a third predetermined height range above the solder joint reference plane in the solder joint region and sort them in ascending order to obtain a solder joint height dataset. Select the height values ​​of the fourth predetermined ascending interval in the solder joint height dataset and calculate the average value as the solder joint height.

2. The method for detecting the welding quality of aluminum wire in new energy batteries according to claim 1, characterized in that, The structure of the target recognition model includes: The system consists of convolutional layers, pooling layers, activation functions, and a fire module. The fire module includes a compression layer and a dilation layer. The compression layer employs... The convolutional kernel performs dimensionality compression on the feature map to obtain a compressed feature map. The dilation layer then passes the compressed feature map through... convolution kernel and The convolution kernel is used to expand the dimension.

3. The method for detecting the welding quality of aluminum wire in new energy batteries according to claim 1, characterized in that, The identification step or the preparation step is preceded by a model training step, which includes: Image annotation sub-steps: In the first predetermined number of brightness images, the solder joints are individually annotated to obtain a solder joint image set; in the second predetermined number of brightness images, the first welding arc is individually annotated to obtain a first welding arc image set. Image processing sub-step: Input the weld point image set and the first welding arc image set into the target recognition model to obtain the output result; Parameter update sub-step: Calculate the error of the output result. If the error is greater than a predetermined error, update the parameters of the model according to the error. After the update, perform the image processing sub-step again. If the error is less than the predetermined error, the training is complete.

4. The method for detecting the welding quality of aluminum wire in new energy batteries according to claim 1, characterized in that, The plane fitting sub-step includes: Let the plane equation be: , The constraint equations are defined as follows: , When the plane equation represents the first reference plane, the coefficients... The condition is met when the sum of the squares of the distances from points in the first set of valid values ​​to the first reference plane is minimized. The sum of the squares of the distances from the points in the first set of valid values ​​to the first reference plane is used as... This means that the following formula is satisfied: ， in, It is any point in the first set of valid values. Distance to the first reference plane; When the plane equation represents the second reference plane, the coefficients... The condition is met when the sum of the squares of the distances from points in the second set of valid values ​​to the second reference plane is minimized. The sum of the squares of the distances from the points in the second set of effective values ​​to the second reference plane is used This means that the following formula is satisfied: ， in, It is any point in the second set of valid values. The distance to the second reference plane.

5. The method for detecting the welding quality of aluminum wire in new energy batteries according to claim 1, characterized in that, The process of measuring the width of the solder joint in the measurement step includes: Image segmentation sub-step: The SAM algorithm is used to segment the solder joint positions to obtain solder joint images; The sub-step for obtaining the pixel length ratio is as follows: Obtain the pixel length ratio corresponding to the solder joint position; The sub-step for obtaining the solder joint width is as follows: The solder joint width is obtained based on the ratio of the width pixels of the solder joint image to the pixel length.

6. A welding quality inspection device for aluminum wires in new energy batteries, applied to the welding quality inspection method for aluminum wires in new energy batteries according to any one of claims 1-5, characterized in that, include: Preparatory module: Acquires the brightness and height maps of the target to be detected; Recognition module: The target recognition model is used to identify the solder joint and the first welding arc from the brightness map to obtain the position of the solder joint and the position of the first welding arc, respectively; Measurement module: Based on the brightness map, the height map, the solder joint position and the first welding arc position, obtain the first reference plane corresponding to the first solder pad and the second reference plane corresponding to the second solder pad, and measure the first welding arc height, the second welding arc height, the solder joint height and the solder joint width; Quality judgment module: If the first welding arc height, the second welding arc height, the weld width, and the weld height are all within the corresponding standard ranges, the welding quality is judged to be qualified; otherwise, the welding quality is judged to be unqualified.

Images