Vision-based battery electrode casting quality inspection method and system

Abstract

The present invention relates to the field of battery electrode quality detection, and in particular to a vision-based battery electrode cast welding quality detection method and system. The method comprises: collecting the timing cast welding control parameters of the battery electrode and the cast welding area image after the cast welding is completed, wherein the timing cast welding control parameters include current, voltage, temperature, cast welding position and cast welding speed; calculating the difference between the timing cast welding control parameters of the battery electrode and the standard timing cast welding control parameters, and determining whether there are abnormal cast welding points, wherein the cast welding points are the range of cast welding at any time in the timing cast welding control parameters; if there are abnormal cast welding points, then according to the abnormal cast welding points, extracting the suspected defect area image from the cast welding area image after the cast welding is completed; inputting the suspected defect area image into the trained quality detection model to determine whether the battery electrode cast welding quality is qualified. The present invention effectively solves the problem that the battery electrode cast welding quality cannot be detected quickly and accurately.

Description

Technical Field

[0001] The present invention relates to the field of battery electrode quality detection, and more specifically, to a vision-based battery electrode casting quality detection method and system. Background Art

[0002] With the rapid development of the battery manufacturing industry, battery electrode cast welding technology has been widely used. In the process of battery electrode cast welding, unreasonable settings of the cast welding equipment or the influence of cast welding environmental factors may cause defects in the cast welding of the battery electrode, resulting in unqualified cast welding quality. The existing battery electrode cast welding quality detection method is to perform threshold segmentation on the welding image, extract the defective area, and then evaluate the welding quality based on the defective area. However, due to the complexity of the welding scene, the threshold segmentation is easy to misjudge the noise, shadow or weld edge area in the image as a defective area, resulting in reduced accuracy in the evaluation of welding quality.

[0003] At present, a Chinese patent document with the announcement number CN116664579B and the name "A battery pack welding quality image detection method" discloses a method for welding quality detection by collecting welding images, filtering and enhancing the images, binarizing the enhanced images using a segmentation threshold and performing least squares fitting, and inputting the processed data into a defect detection network model.

[0004] The above method of detecting the welding quality of the battery pack through images processes the entire welding area, and the detection efficiency is relatively low. In addition, the welding quality cannot be accurately detected without combining welding parameters. Summary of the invention

[0005] In order to solve the above-mentioned problem of being unable to quickly and accurately detect the quality of battery electrode cast welding, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a vision-based battery electrode cast welding quality detection method, comprising: collecting the timing cast welding control parameters of the battery electrode and the cast welding area image after the cast welding is completed, the timing cast welding control parameters including current, voltage, temperature, cast welding position and cast welding speed, the cast welding position is represented by the three-dimensional coordinates of the cast welding point at the current moment; calculating the difference between the timing cast welding control parameters of the battery electrode and the standard timing cast welding control parameters to determine whether there are abnormal cast welding points; if there are abnormal cast welding points, obtaining the three-dimensional coordinates of all abnormal cast welding points, and selecting the continuous abnormal points that are continuously distributed among the abnormal cast welding points; simplifying the three-dimensional coordinates of the continuous abnormal points, retaining the coordinates of the X-axis and Y-axis; calculating the minimum circumscribed rectangle of the range composed of continuous abnormal points, and determining the area corresponding to the minimum circumscribed rectangle in the cast welding area image as a suspected defect area image; inputting the suspected defect area image into a trained quality detection model, and determining whether the battery electrode cast welding quality is qualified in response to whether the cast welding quality of the suspected defect area is qualified.

[0007] By calculating the difference between the timing cast welding control parameters of the battery electrode and the standard timing cast welding control parameters, it is determined whether there are abnormal cast welding points, and normal cast welding points are quickly excluded, reducing the amount of calculation for subsequent operations. Then, based on the abnormal cast welding points, images of suspected defective areas are extracted and tested to accurately detect whether the quality of battery electrode cast welding is qualified.

[0008] Preferably, the calculation of the difference between the timing cast welding control parameters of the battery electrode and the standard timing cast welding control parameters to determine whether there is an abnormal cast welding point specifically includes: for any moment i in the timing cast welding control parameters of the battery electrode, if at least one of the following conditions is met, there is an abnormal cast welding point, and the three-dimensional coordinates of the abnormal cast welding point are the three-dimensional coordinates of the cast welding position corresponding to the moment i: the difference between the current corresponding to the moment i and the standard current is greater than the current fluctuation value, the difference between the voltage and the standard voltage is greater than the voltage fluctuation value, the difference between the temperature and the standard temperature is greater than the temperature fluctuation value, the difference between the cast welding speed and the standard cast welding speed is greater than the cast welding speed fluctuation value, and the distance between the cast welding position and the standard cast welding position is greater than the position deviation threshold.

[0009] Preferably, the selecting of continuously distributed continuous abnormal points among the abnormal cast welds includes: among at least three abnormal cast welds, when the maximum value of the time intervals corresponding to the abnormal cast welds at any two adjacent moments is less than or equal to a time interval threshold, the at least three abnormal cast welds are continuous abnormal points.

[0010] By classifying abnormal cast welds into continuous abnormal points and individual abnormal points, defects in the cast welding process can be analyzed more accurately. The identification of continuous abnormal points helps to find systematic problems, such as equipment failure or improper operation, while individual abnormal points generally indicate sporadic problems. This classification improves the accuracy and efficiency of defect detection.

[0011] Preferably, the method for obtaining the time interval threshold is specifically as follows: obtaining a historical cast welding image, manually marking its actual defect area; initializing the time interval threshold, extracting the image of the suspected defect area, and calculating ,in For similarity, Indicates the suspected defect area extracted corresponding to the time interval threshold, Represents the actual defect area manually marked; use the annealing algorithm to optimize and update the time interval threshold, obtain the historical cast welding image again according to the updated value, and repeat the above steps until the fluctuation range of X for m consecutive times is less than the preset fluctuation value. At this time, the corresponding time interval threshold is the final time interval threshold.

[0012] By calculating the ratio of the manually marked suspected defect areas to the extracted suspected defect areas, the time interval threshold can be accurately adjusted to ensure a higher defect detection accuracy.

[0013] Preferably, the optimizing and updating the time interval threshold using an annealing algorithm comprises:

[0014] The initial temperature is preset, the temperature is linearly reduced, and the time interval threshold is iterated using a domain search method.

[0015] The use of linear cooling and neighborhood search methods can avoid falling into local optimality. In the high temperature stage, a larger solution space can be quickly explored, while in the low temperature stage, a better solution is found through detailed neighborhood search, thereby finding the optimal time interval threshold in a shorter time.

[0016] Preferably, the quality detection model is: a convolution-back propagation neural network model.

[0017] Preferably, the training process of the convolution-back propagation neural network model includes: collecting historical cast welding defect area images and corresponding quality labels as samples, and the quality labels are divided into qualified and unqualified; constructing a convolution-back propagation neural network model, using the convolutional neural network to extract the image features of the historical cast welding defect area, inputting the image features into the back propagation neural network, outputting a qualified or unqualified result, using a cross entropy loss function to measure the difference between the output result and the quality label, and using the samples to train the constructed convolution-back propagation neural network model to obtain a trained quality detection model.

[0018] The use of a convolution-back propagation neural network model for cast welding quality inspection has significant advantages. First, by extracting the features of historical defect area images, complex image patterns can be effectively identified and detection accuracy can be improved. Second, the cross entropy loss function is used to measure the difference between the output and the actual quality label, allowing the model to converge quickly during training and optimize performance. Ultimately, this method can achieve automated and accurate battery electrode cast welding quality inspection.

[0019] In a second aspect, the present invention also provides a vision-based battery electrode cast welding quality inspection system, comprising: a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-mentioned vision-based battery electrode cast welding quality inspection method.

[0020] The beneficial effect of the present invention is that by calculating the difference between the timing cast welding control parameters of the battery electrode and the standard timing cast welding control parameters, it is determined whether there are abnormal cast welding points, and the normal cast welding points are quickly excluded, reducing the calculation amount of subsequent operations, and then the minimum circumscribed rectangle is extracted according to the abnormal cast welding points as the image of the suspected defect area and detected. By extracting the abnormal cast welding points and constructing the minimum circumscribed rectangle, the suspected defect area can be analyzed centrally, thereby improving the accuracy and efficiency of detecting the cast welding quality of the battery electrode. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0022] Figure 1 It is a flow chart of a method for detecting the quality of battery electrode casting welding based on vision provided by an embodiment of the present invention;

[0023] Figure 2 The present invention provides a block diagram of a vision-based battery electrode cast welding quality detection system. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0025] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0026] Figure 1 The flowchart of a method for detecting the quality of battery electrode casting welding based on vision provided by an embodiment of the present invention comprises the following steps:

[0027] S101, collecting the sequential cast welding control parameters of the battery electrode and the cast welding area image after the cast welding is completed, wherein the sequential cast welding control parameters include current, voltage, temperature, cast welding position and cast welding speed, and the cast welding position is represented by the three-dimensional coordinates of the cast welding point at the current moment;

[0028] Current is one of the key parameters that affect the cast welding process. The right current ensures the melting and flow of the welding material. Too high a current may cause overheating and defects, while too low a current may result in poor welding or cold joints. Voltage and current work together in the welding process. The right voltage helps maintain a stable welding arc and heat input, thus ensuring the quality of the weld. Too high or too low a voltage will affect the shape of the molten pool and the bonding of the weld. Temperature monitoring is also critical to the quality of cast welding. By monitoring the temperature of the welding area in real time, overheating or too rapid cooling can be avoided, ensuring the proper fluidity and bonding of the weld metal. The cast welding position is generally represented by the three-dimensional coordinates of the cast welding point at the current moment, and the cast welding speed is represented by the spatial Euclidean geometric distance between the cast welding points at the current moment and the next moment. For example, in a battery electrode cast welding process, cast welding is required for 3 seconds, and the collection period is 1 second. Then, the cast welding range collected at 1 second is the first cast welding point, the cast welding range collected at 2 seconds is the second cast welding point, and the cast welding range collected at 3 seconds is the third cast welding point. The cast welding position can be represented by the three-dimensional coordinates of each cast welding point. In some embodiments, if the range of the cast welding points is large, it can also be represented by the centroid coordinates of the cast welding points. For the cast welding speed, the spatial Euclidean distance is the straight-line distance between two points in space. For example, the cast welding position of the cast welding point at 1 second is , the casting position of the casting welding point at 2 seconds is , then the spatial Euclidean distance cm, that is, the casting welding speed at 1 second is .

[0029] S102, calculating the difference between the sequential casting welding control parameters of the battery electrode and the standard sequential casting welding control parameters to determine whether there are abnormal casting welding points;

[0030] In some embodiments, the calculation of the difference between the timing cast welding control parameters of the battery electrode and the standard timing cast welding control parameters to determine whether there is an abnormal cast welding point specifically includes: for any time i in the timing cast welding control parameters of the battery electrode, if at least one of the following conditions is met, there is an abnormal cast welding point, and the three-dimensional coordinates of the abnormal cast welding point are the three-dimensional coordinates of the cast welding position corresponding to time i: the difference between the current corresponding to time i and the standard current is greater than the current fluctuation value, the difference between the voltage and the standard voltage is greater than the voltage fluctuation value, the difference between the temperature and the standard temperature is greater than the temperature fluctuation value, the difference between the cast welding speed and the standard cast welding speed is greater than the cast welding speed fluctuation value, and the distance between the cast welding position and the standard cast welding position is greater than the position deviation threshold. The standard timing cast welding control parameters are generally obtained by collecting the timing cast welding control parameters of the battery electrode when the quality of the cast welding is qualified and statistically analyzing them to obtain the standard timing cast welding control parameters of the battery electrode during cast welding and the fluctuation values ​​allowed at each moment for each control parameter.

[0031] For example, the standard timing casting welding control parameter at time i is the standard current Standard voltage , standard temperature Standard casting speed , standard cast welding position , current fluctuation value , voltage fluctuation value , temperature fluctuation value , casting welding speed fluctuation value , position deviation threshold cm. The time sequence casting welding control parameters collected at time i are current ,Voltage ,temperature Casting welding speed , Casting welding position Obviously, the above conditions satisfy , and cast welding position With standard cast welding position The distance is Two conditions, so the casting weld point corresponding to time i is an abnormal casting weld point.

[0032] S103, if there are abnormal cast welds, obtain the three-dimensional coordinates of all abnormal cast welds, select the continuously distributed continuous abnormal points among the abnormal cast welds; simplify the three-dimensional coordinates of the continuous abnormal points, retain the coordinates of the X-axis and the Y-axis; calculate the minimum circumscribed rectangle of the range formed by the continuous abnormal points, and determine the area corresponding to the minimum circumscribed rectangle in the cast weld area image as the suspected defect area image;

[0033] Generally, if the abnormal weld spot is a single abnormal point, it will not affect the weld quality of the battery electrode. If the abnormal weld spot is a continuous abnormal point, it indicates that the quality of the battery electrode may be poor. In some embodiments, the selecting of continuous abnormal points distributed continuously among the abnormal weld spots includes: among at least three abnormal weld spots, when the maximum value of the time intervals corresponding to the abnormal weld spots at any two adjacent moments is less than or equal to the time interval threshold, the at least three abnormal weld spots are continuous abnormal points.

[0034] For example, the moments corresponding to the existing four abnormal welds are 1 second, 3 seconds, 5 seconds and 6 seconds respectively, where the maximum time interval corresponding to any two adjacent moments of the abnormal welds is 2 seconds, and the preset time interval threshold for judging consecutive abnormal points is 3 seconds, then the four abnormal welds are consecutive abnormal points. If the moments corresponding to the existing three abnormal welds are 3 seconds, 6 seconds and 10 seconds respectively, and the preset time interval threshold for judging consecutive abnormal points is 2 seconds, for the abnormal weld corresponding to the moment of 6 seconds, its time interval with the abnormal welds corresponding to the moments of 3 seconds and 10 seconds is greater than 2 seconds, so the abnormal weld corresponding to the moment of 6 seconds is a single abnormal point.

[0035] For example, the method of obtaining an image of a suspected defect area is as follows: There are 4 consecutive abnormal points, and their 3D coordinates are , , , , simplify the three-dimensional coordinates, retain the coordinates of the X-axis and Y-axis to obtain , , , , among the simplified coordinates of the four consecutive abnormal points, the minimum X-axis coordinate is min(12,14,14,15)=12, the maximum X-axis coordinate is max(12,14,14,15)=15, the minimum Y-axis coordinate is min(25,24,24,24)=24, and the maximum Y-axis coordinate is max(25,24,24,24)=25. Therefore, the four vertex coordinates of the minimum circumscribed rectangle corresponding to the four consecutive abnormal points can be determined as follows: lower left corner (12, 24), upper left corner (12, 25), lower right corner (15, 24), and upper right corner (15, 25). In the cast welding area image, the area image corresponding to the minimum circumscribed rectangle is the suspected defect area image.

[0036] In addition, the method for obtaining the time interval threshold is specifically as follows: obtaining a historical cast welding image, manually marking its actual defect area; initializing the time interval threshold, extracting the image of the suspected defect area, and calculating ,in For similarity, Indicates the suspected defect area extracted corresponding to the time interval threshold, Represents the actual defect area manually marked; use the annealing algorithm to optimize and update the time interval threshold, obtain the historical cast welding image again according to the updated value, repeat the above steps until the fluctuation range of X for m consecutive times is less than the preset fluctuation value, and the corresponding time interval threshold is the final time interval threshold. In more detail, the process is as follows: obtain a historical cast welding image, mark the actual defect area in the historical cast welding image by manual marking, obtain the actual defect area manually marked, initialize a set time interval threshold, obtain continuous abnormal points in the historical cast welding image according to the set time interval threshold, and then extract the suspected defect area in the historical cast welding image to obtain the extracted suspected defect area corresponding to the set time interval threshold, and calculate ,in is the calculated ratio, Indicates the suspected defect area extracted corresponding to the set time interval threshold. represents the actual defect area annotated manually, represents the overlap area between the suspected defect area and the actual defect area, represents the union of the suspected defect area and the actual defect area, ensuring In the case of The closer it is to 1, the closer the suspected defect area extracted by the set time interval threshold is to the actual defect area manually marked. Then use the annealing algorithm to optimize and update the set time interval threshold, obtain the historical cast welding image again according to the updated value, and repeat the above steps until the calculated X value no longer changes significantly. At this time, the corresponding set time interval threshold is the optimal time interval threshold. The use of the annealing algorithm to optimize and update the time interval threshold includes: presetting the initial temperature, taking linear cooling, and iterating the time interval threshold using the field search method.

[0037] S104, inputting the image of the suspected defect area into a trained quality inspection model, and determining whether the cast welding quality of the battery electrode is qualified in response to whether the cast welding quality of the suspected defect area is qualified.

[0038] In some embodiments, the quality inspection model can be: a convolution-back propagation neural network model. The training process is as follows: collect historical cast welding defect area images and corresponding quality labels as samples, and the quality labels are divided into qualified and unqualified; construct a convolution-back propagation neural network model, use the convolution neural network to extract the image features of the historical cast welding defect area, input the image features into the back propagation neural network, output the result as qualified or unqualified, use the cross entropy loss function to measure the difference between the output result and the quality label, use the sample to train the constructed convolution-back propagation neural network model, and obtain a trained quality inspection model.

[0039] The present invention determines whether there are abnormal cast welding points by calculating the difference between the timing cast welding control parameters of the battery electrode and the standard timing cast welding control parameters, quickly excludes normal cast welding points, reduces the amount of calculation for subsequent operations, and then extracts images of suspected defective areas based on the abnormal cast welding points and detects them, so as to accurately and quickly detect whether the quality of the battery electrode cast welding is qualified.

[0040] The present invention also provides a visual-based battery electrode casting quality detection system. Figure 2 As shown, the system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a vision-based battery electrode cast welding quality detection method described in the present invention is implemented.

[0041] The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, whose configuration and functions are known in the art and thus will not be described in detail here.

[0042] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0043] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0044] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A visual-based battery electrode casting quality inspection method, characterized in that: include: Collecting the sequential cast welding control parameters of the battery electrode and the cast welding area image after the cast welding is completed, wherein the sequential cast welding control parameters include current, voltage, temperature, cast welding position and cast welding speed, and the cast welding position is represented by the three-dimensional coordinates of the cast welding point at the current moment; Calculating the difference between the sequential casting welding control parameters of the battery electrode and the standard sequential casting welding control parameters to determine whether there are abnormal casting welding points; If there are abnormal cast welds, obtain the three-dimensional coordinates of all abnormal cast welds. Among at least three abnormal cast welds, when the maximum value of the time intervals corresponding to any two adjacent abnormal cast welds is less than or equal to the time interval threshold, the at least three abnormal cast welds are continuous abnormal points. Simplify the three-dimensional coordinates of the continuous abnormal points, and retain the coordinates of the X-axis and the Y-axis. Calculate the minimum bounding rectangle of the range formed by the continuous abnormal points, and determine the area corresponding to the minimum bounding rectangle in the cast weld area image as the suspected defect area image. The method for obtaining the time interval threshold is specifically as follows: Obtain historical cast welding images and manually mark their actual defect areas; Initialize the time interval threshold, extract the image of the suspected defect area, and calculate ,in For similarity, Indicates the suspected defect area extracted corresponding to the time interval threshold, Indicates the actual defect area annotated manually; The time interval threshold is optimized and updated using an annealing algorithm, and the historical cast welding image is acquired again according to the updated value, and the above steps are repeated until the fluctuation range of X for m consecutive times is less than the preset fluctuation value, and the corresponding time interval threshold is the final time interval threshold; The image of the suspected defect area is input into a trained quality inspection model, and in response to whether the cast welding quality of the suspected defect area is qualified, it is determined whether the cast welding quality of the battery electrode is qualified.

2. The visual-based battery electrode cast welding quality detection method according to claim 1, characterized in that: The calculating the difference between the sequential casting welding control parameter of the battery electrode and the standard sequential casting welding control parameter to determine whether there is an abnormal casting welding point specifically includes: For any moment i in the sequential cast welding control parameters of the battery electrode, if at least one of the following conditions is met, there is an abnormal cast welding point, and the three-dimensional coordinates of the abnormal cast welding point are the three-dimensional coordinates of the cast welding position corresponding to the moment i: the difference between the current corresponding to the moment i and the standard current is greater than the current fluctuation value, the difference between the voltage and the standard voltage is greater than the voltage fluctuation value, the difference between the temperature and the standard temperature is greater than the temperature fluctuation value, the difference between the cast welding speed and the standard cast welding speed is greater than the cast welding speed fluctuation value, and the distance between the cast welding position and the standard cast welding position is greater than the position deviation threshold.

3. The visual-based battery electrode casting quality detection method according to claim 1 is characterized in that: The step of optimizing and updating the time interval threshold by using an annealing algorithm includes: The initial temperature is preset, the temperature is linearly reduced, and the time interval threshold is iterated using a domain search method.

4. The visual-based battery electrode cast welding quality detection method according to claim 1, characterized in that: The quality detection model is: a convolution-back propagation neural network model.

5. The visual-based battery electrode casting quality detection method according to claim 4 is characterized in that: The training process of the convolution-back propagation neural network model includes: Collect historical cast welding defect area images and corresponding quality labels as samples, and the quality labels are divided into qualified and unqualified; A convolution-back propagation neural network model is constructed, and the convolutional neural network is used to extract the image features of the historical cast welding defect area. The image features are input into the back propagation neural network, and the output result is qualified or unqualified. The cross entropy loss function is used to measure the difference between the output result and the quality label. The constructed convolution-back propagation neural network model is trained using samples to obtain a trained quality inspection model.

6. A vision-based battery electrode casting quality inspection system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the visual-based battery electrode cast welding quality detection method as described in any one of claims 1 to 5.

Images